Include Krakatau and enjarify resources directly
This commit is contained in:
Nico Mexis
parent
8cbe4301f7
commit
a4b452de6a
Binary file not shown.
Binary file not shown.
27
pom.xml
27
pom.xml
|
|
@ -5,6 +5,13 @@
|
|||
<artifactId>bytecodeviewer</artifactId>
|
||||
<version>2.9.23</version>
|
||||
|
||||
<properties>
|
||||
<java.version>8</java.version>
|
||||
<maven.compiler.target>${java.version}</maven.compiler.target>
|
||||
<maven.compiler.source>${java.version}</maven.compiler.source>
|
||||
<project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
|
||||
</properties>
|
||||
|
||||
<dependencies>
|
||||
<dependency>
|
||||
<groupId>com.android</groupId>
|
||||
|
|
@ -109,13 +116,6 @@
|
|||
<artifactId>dx</artifactId>
|
||||
<version>1.16</version>
|
||||
</dependency>
|
||||
<dependency>
|
||||
<groupId>enjarify</groupId>
|
||||
<artifactId>enjarify</artifactId>
|
||||
<version>3</version>
|
||||
<scope>system</scope>
|
||||
<systemPath>${project.basedir}/libs/enjarify-3.jar</systemPath>
|
||||
</dependency>
|
||||
<dependency>
|
||||
<groupId>org.jboss.windup.decompiler</groupId>
|
||||
<artifactId>decompiler-fernflower</artifactId>
|
||||
|
|
@ -155,13 +155,6 @@
|
|||
<artifactId>jgraphx</artifactId>
|
||||
<version>3.4.1.3</version>
|
||||
</dependency>
|
||||
<dependency>
|
||||
<groupId>krakatau</groupId>
|
||||
<artifactId>krakatau</artifactId>
|
||||
<version>11</version>
|
||||
<scope>system</scope>
|
||||
<systemPath>${project.basedir}/libs/Krakatau-11.jar</systemPath>
|
||||
</dependency>
|
||||
<dependency>
|
||||
<groupId>org.objenesis</groupId>
|
||||
<artifactId>objenesis</artifactId>
|
||||
|
|
@ -220,8 +213,8 @@
|
|||
<artifactId>maven-compiler-plugin</artifactId>
|
||||
<version>3.8.1</version>
|
||||
<configuration>
|
||||
<source>8</source>
|
||||
<target>8</target>
|
||||
<source>${java.version}</source>
|
||||
<target>${java.version}</target>
|
||||
</configuration>
|
||||
</plugin>
|
||||
<plugin>
|
||||
|
|
@ -229,7 +222,7 @@
|
|||
<artifactId>maven-javadoc-plugin</artifactId>
|
||||
<version>3.2.0</version>
|
||||
<configuration>
|
||||
<source>8</source>
|
||||
<source>${java.version}</source>
|
||||
</configuration>
|
||||
</plugin>
|
||||
<plugin>
|
||||
|
|
|
|||
|
|
@ -0,0 +1,8 @@
|
|||
root = true
|
||||
|
||||
[*.py]
|
||||
charset = utf-8
|
||||
indent_style = space
|
||||
indent_size = 4
|
||||
insert_final_newline = true
|
||||
end_of_line = lf
|
||||
|
|
@ -0,0 +1,2 @@
|
|||
* text=auto
|
||||
*.test binary
|
||||
|
|
@ -0,0 +1,5 @@
|
|||
*.pyc
|
||||
*.pyo
|
||||
test*.py
|
||||
Krakatau/plugins/*
|
||||
tests/.cache
|
||||
|
|
@ -0,0 +1,295 @@
|
|||
Krakatau Assembly Syntax
|
||||
For a list of previous changes to the assembly syntax, see changelog.txt
|
||||
|
||||
Note: This documents the officially supported syntax of the assembler. The assembler accepts some files that don't fully conform to this syntax, but this behavior may change without warning in the future.
|
||||
|
||||
Lexical structure
|
||||
Comments: Comments begin with a semicolon and go to the end of the line. Since no valid tokens start with a semicolon, there is no ambiguity. Comments are ignored during parsing.
|
||||
|
||||
Whitespace: At least one consecutive space or tab character
|
||||
|
||||
Lines that are empty except for whitespace or a comment are ignored. Many grammar productions require certain parts to be separated by a newline (LF/CRLF/CR). This is represented below by the terminal EOL. Due to the rules above, EOL can represent an optional comment, followed by a newline, followed by any number of empty/comment lines. There are no line continuations.
|
||||
|
||||
Integer, Long, Float, and Double literals use the same syntax as Java with a few differences:
|
||||
* No underscores are allowed
|
||||
* Doubles cannot be suffixed with d or D.
|
||||
* Decimal floating point literals with values that can’t be represented exactly in the target type aren’t guaranteed to round the same way as in Java. If the exact value is significant, you should use a hexidecimal floating point literal.
|
||||
* If a decimal point is present, there must be at least one digit before and after it (0.5 is ok, but .5 is not. 5.0 is ok but 5. is not).
|
||||
* A leading plus or minus sign is allowed.
|
||||
* Only decimal and hexadecimal literals are allowed (no binary or octal)
|
||||
* For doubles, special values can be represented by +Infinity, -Infinity, +NaN, and -NaN (case insensitive). For floats, these should be suffixed by f or F.
|
||||
* NaNs with a specific binary representation can be represented by suffixing with the hexadecimal value in angle brackets. For example, -NaN<0x7ff0123456789abc> or +NaN<0xFFABCDEF>f
|
||||
|
||||
Note: NaN requires a leading sign, even though it is ignored. This is to avoid ambiguity with WORDs. The binary representation of a NaN with no explicit representation may be any valid encoding of NaN. If you care about the binary representation in the classfile, you should specify it explicitly as described above.
|
||||
|
||||
String literals use the same syntax as Java string literals with the following exceptions
|
||||
* Non printable and non-ascii characters, including tabs, are not allowed. These can be represented by escape sequences as usual. For example \t means tab.
|
||||
* Either single or double quotes can be used. If single quotes are used, double quotes can appear unescaped inside the string and vice versa.
|
||||
* There are three additional types of escape sequences allowed: \xDD, \uDDDD, and \UDDDDDDDD where D is a hexadecimal digit. The later two are only allowed in unicode strings (see below). In the case of \U, the digits must correspond to a number less than 0x00110000. \x represents a byte or code point up to 255. \u represents a code point up to 65535. \U represents a code point up to 1114111 (0x10FFFF), which will be split into a surrogate pair when encoded if it is above 0xFFFF.
|
||||
* There are two types of string literals - bytes and unicode. Unicode strings are the default and represent a sequence of code points which will be MUTF8 encoded when written to the classfile. A byte string, represented by prefixing with b or B, represents a raw sequence of bytes which will be written unchanged. For example, "\0" is encoded to a two byte sequence while b"\0" puts an actual null byte in the classfile (which is invalid, but potentially useful for testing).
|
||||
|
||||
Reference: The classfile format has a large number of places where an index is made into the constant pool or bootstrap methods table. The assembly format allows you to specify the definition inline, and the assembler will automatically add an entry as appropriate and fill in the index. However, this isn’t acceptable in cases where the exact binary layout is important or where a definition is large and you want to refer to it many times without copying the definition each time.
|
||||
|
||||
For the first case, there are numeric references, designated by a decimal integer in square brackets with no leading zeroes. For example, [43] refers to the index 43 in the constant pool. For the second case, there are symbolic references, which is a sequence of lowercase ascii, digits, and underscores inside square brackets, not beginning with a digit. For example, [foo_bar4].
|
||||
|
||||
Bootstrap method references are the same except preceded by "bs:". For example, [bs:43] or [bs:foo_bar4]. These are represented by the terminal BSREF. Bootstrap method references are only used in very specific circumstances so you probably won’t need them. All other references are constant pool references and have no prefix, designated by the terminal CPREF.
|
||||
|
||||
Note: Constant pools and bootstrap method tables are class-specific. So definitions inside one class do not affect any other classes assembled from the same source file.
|
||||
|
||||
Labels refer to a position within a method’s bytecode. The assembler will automatically fill in each label with the calculated numerical offset. Labels consist of a capital L followed by ascii letters, digits, and underscores. A label definition (LBLDEF) is a label followed by a colon (with no space). For example, "LSTART:". Label uses are included in the WORD token type defined below since they don’t have a colon.
|
||||
|
||||
Note: Labels refer to positions in the bytecode of the enclosing Code attribute where they appear. They may not appear outside of a Code attribute.
|
||||
|
||||
Word: A string beginning with a word_start_character, followed by zero or more word_start_character or word_rest_characters. Furthermore, if the first character is a [, it must be followed by another [ or a capital letter (A-Z).
|
||||
|
||||
word_start_character: a-z, A-Z, _, $, (, <, [
|
||||
word_rest_character: 0-9, ), >, /, ;, *, +, -
|
||||
|
||||
Words are used to specify names, identifiers, descriptors, and so on. If you need to specify a name that can’t be represented as a word (such as using forbidden characters), a string literal can be used instead. Words are represented in the grammar by the terminal WORD.
|
||||
|
||||
For example, 42 is not a valid word because it begins with a digit. A class named 42 can be defined as follows:
|
||||
|
||||
.class "42"
|
||||
|
||||
In addition, when used in a context following flags, words cannot be any of the possible flag names. These are currently public, private, protected, static, final, super, synchronized, open, transitive, volatile, bridge, static_phase, transient, varargs, native, interface, abstract, strict, synthetic, annotation, enum, module, and mandated. In addition, strictfp is disallowed to avoid confusion. So if you wanted to have a string field named bridge, you’d have to do
|
||||
|
||||
.field "bridge" Ljava/lang/String;
|
||||
|
||||
Format of grammar rules
|
||||
Nonterminals are specified in lowercase. Terminals with a specific value required are specified in quotes. e.g. "Foo" means that the exact text Foo (case sensitive) has to appear at that point. Terminals that require a value of a given token type are represented in all caps, e.g. EOL, INT_LITERAL, FLOAT_LITERAL, LONG_LITERAL, DOUBLE_LITERAL, STRING_LITERAL, CPREF, BSREF, WORD, LBLDEF.
|
||||
|
||||
*, +, ?, |, and () have their usual meanings in regular expressions.
|
||||
|
||||
Common constant rules
|
||||
s8: INT_LITERAL
|
||||
u8: INT_LITERAL
|
||||
s16: INT_LITERAL
|
||||
u16: INT_LITERAL
|
||||
s32: INT_LITERAL
|
||||
u32: INT_LITERAL
|
||||
|
||||
ident: WORD | STRING_LITERAL
|
||||
utfref: CPREF | ident
|
||||
clsref: CPREF | ident
|
||||
natref: CPREF | ident utfref
|
||||
fmimref: CPREF | fmim_tagged_const
|
||||
bsref: BSREF | bsnotref
|
||||
invdynref: CPREF | invdyn_tagged_const
|
||||
|
||||
handlecode: "getField" | "getStatic" | "putField" | "putStatic" | "invokeVirtual" | "invokeStatic" | "invokeSpecial" | "newInvokeSpecial" | "invokeInterface"
|
||||
mhandlenotref: handlecode (CPREF | fmim_tagged_const)
|
||||
mhandleref: CPREF | mhandlenotref
|
||||
|
||||
cmhmt_tagged_const: "Class" utfref | "MethodHandle" mhandlenotref | "MethodType" utfref
|
||||
ilfds_tagged_const: "Integer" INT_LITERAL | "Float" FLOAT_LITERAL | "Long" LONG_LITERAL | "Double" DOUBLE_LITERAL | "String" STRING_LITERAL
|
||||
simple_tagged_const: "Utf8" ident | "NameAndType" utfref utfref
|
||||
fmim_tagged_const: ("Field" | "Method" | "InterfaceMethod") clsref natref
|
||||
invdyn_tagged_const: "InvokeDynamic" bsref natref
|
||||
|
||||
ref_or_tagged_const_ldc: CPREF | cmhmt_tagged_const | ilfds_tagged_const
|
||||
ref_or_tagged_const_all: ref_or_tagged_ldconst | simple_tagged_const | fmim_tagged_const | invdyn_tagged_const
|
||||
|
||||
bsnotref: mhandlenotref ref_or_tagged_const_ldc* ":"
|
||||
ref_or_tagged_bootstrap: BSREF | "Bootstrap" bsnotref
|
||||
|
||||
Note: The most deeply nested possible valid constant is 6 levels (InvokeDynamic -> Bootstrap -> MethodHandle -> Method -> NameAndType -> Utf8). It is possible to create a more deeply nested constant definitions in this grammar by using references with invalid types, but the assembler may reject them.
|
||||
|
||||
ldc_rhs: CPREF | INT_LITERAL | FLOAT_LITERAL | LONG_LITERAL | DOUBLE_LITERAL | STRING_LITERAL | cmhmt_tagged_const
|
||||
|
||||
flag: "public" | "private" | "protected" | "static" | "final" | "super" | "synchronized" | "volatile" | "bridge" | "transient" | "varargs" | "native" | "interface" | "abstract" | "strict" | "synthetic" | "annotation" | "enum" | "mandated"
|
||||
Basic assembly structure
|
||||
|
||||
assembly_file: EOL? class_definition*
|
||||
class_definition: version? class_start class_item* class_end
|
||||
|
||||
version: ".version" u16 u16 EOL
|
||||
class_start: class_directive super_directive interface_directive*
|
||||
class_directive: ".class" flag* clsref EOL
|
||||
super_directive: ".super" clsref EOL
|
||||
interface_directive: ".implements" clsref EOL
|
||||
class_end: ".end" "class" EOL
|
||||
|
||||
class_item: const_def | bootstrap_def | field_def | method_def | attribute
|
||||
|
||||
const_def: ".const" CPREF "=" ref_or_tagged_const_all EOL
|
||||
bootstrap_def: ".bootstrap" BSREF "=" ref_or_tagged_bootstrap EOL
|
||||
|
||||
Note: If the right hand side is a reference, the left hand side must be a symbolic reference. For example, the following two are valid.
|
||||
.const [foo] = [bar]
|
||||
.const [foo] = [42]
|
||||
|
||||
While these are not valid.
|
||||
.const [42] = [foo]
|
||||
.const [42] = [32]
|
||||
|
||||
field_def: ".field" flag* utfref utfref initial_value? field_attributes? EOL
|
||||
initial_value: "=" ldc_rhs
|
||||
field_attributes: ".fieldattributes" EOL attribute* ".end" ".fieldattributes"
|
||||
|
||||
method_def: method_start (method_body | legacy_method_body) method_end
|
||||
method_start: ".method" flag* utfref ":" utfref EOL
|
||||
method_end: ".end" "method" EOL
|
||||
|
||||
method_body: attribute*
|
||||
legacy_method_body: limit_directive+ code_body
|
||||
limit_directive: ".limit" ("stack" | "locals") u16 EOL
|
||||
|
||||
Attributes
|
||||
attribute: (named_attribute | generic_attribute) EOL
|
||||
generic_attribute: ".attribute" utfref length_override? attribute_data
|
||||
length_override: "length" u32
|
||||
attribute_data: named_attribute | STRING_LITERAL
|
||||
|
||||
named_attribute: annotation_default | bootstrap_methods | code | constant_value | deprecated | enclosing_method | exceptions | inner_classes | line_number_table | local_variable_table | local_variable_type_table | method_parameters | runtime_annotations | runtime_visible_parameter_annotations | runtime_visible_type_annotations | signature | source_debug_extension | source_file | stack_map_table | synthetic
|
||||
|
||||
annotation_default: ".annotationdefault" element_value
|
||||
|
||||
bootstrap_methods: ".bootstrapmethods"
|
||||
Note: The content of a BootstrapMethods attribute is automatically filled in based on the implicitly and explicitly defined bootstrap methods in the class. If this attribute’s contents are nonempty and the attribute isn’t specified explicitly, one will be added implicitly. This means that you generally don’t have to specify it. It’s only useful if you care about the exact binary layout of the classfile.
|
||||
|
||||
|
||||
code: code_start code_body code_end
|
||||
code_start: ".code" "stack" code_limit_t "locals" code_limit_t EOL
|
||||
code_limit_t: u8 | u16
|
||||
code_end: ".end" "code"
|
||||
|
||||
Note: A Code attribute can only appear as a method attribute. This means that they cannot be nested.
|
||||
|
||||
constant_value: ".constantvalue" ldc_rhs
|
||||
|
||||
deprecated: ".deprecated"
|
||||
|
||||
enclosing_method: ".enclosing" "method" clsref natref
|
||||
|
||||
exceptions: ".exceptions" clsref*
|
||||
|
||||
inner_classes: ".innerclasses" EOL inner_classes_item* ".end" "innerclasses"
|
||||
inner_classes_item: cpref cpref utfref flag* EOL
|
||||
|
||||
line_number_table: ".linenumbertable" EOL line_number* ".end" "linenumbertable"
|
||||
line_number: label u16 EOL
|
||||
|
||||
local_variable_table: ".localvariabletable" EOL local_variable* ".end" "localvariabletable"
|
||||
local_variable: u16 "is" utfref utfref code_range EOL
|
||||
|
||||
local_variable_type_table: ".localvariabletypetable" EOL local_variable_type* ".end" "localvariabletypetable"
|
||||
local_variable_type: u16 "is" utfref utfref code_range EOL
|
||||
|
||||
method_parameters: ".methodparameters" EOL method_parameter_item* ".end" "methodparameters"
|
||||
method_parameter_item: utfref flag* EOL
|
||||
|
||||
runtime_annotations: ".runtime" visibility (normal_annotations | parameter_annotations | type_annotations) ".end" "runtime"
|
||||
visibility: "visible" | "invisible"
|
||||
normal_annotations: "annotations" EOL annotation_line*
|
||||
parameter_annotations: "paramannotations" EOL parameter_annotation_line*
|
||||
type_annotations: "typeannotations" EOL type_annotation_line*
|
||||
|
||||
|
||||
signature: ".signature" utfref
|
||||
|
||||
source_debug_extension: ".sourcedebugextension" STRING_LITERAL
|
||||
|
||||
source_file: ".sourcefile" utfref
|
||||
|
||||
stack_map_table: ".stackmaptable"
|
||||
|
||||
Note: The content of a StackMapTable attribute is automatically filled in based on the stack directives in the enclosing code attribute. If this attribute’s contents are nonempty and the attribute isn’t specified explicitly, one will be added implicitly. This means that you generally don’t have to specify it. It’s only useful if you care about the exact binary layout of the classfile.
|
||||
|
||||
Note: The StackMapTable attribute depends entirely on the .stack directives specified. Krakatau will not calculate a new stack map for you from bytecode that does not have any stack information. If you want to do this, you should try using ASM.
|
||||
|
||||
synthetic: ".synthetic"
|
||||
|
||||
Code
|
||||
|
||||
code_body: (instruction_line | code_directive)* attribute*
|
||||
code_directive: catch_directive | stack_directive | ".noimplicitstackmap"
|
||||
|
||||
catch_directive: ".catch" clsref code_range "using" label EOL
|
||||
code_range: "from" label "to" label
|
||||
|
||||
stack_directive: ".stack" stackmapitem EOL
|
||||
stackmapitem: stackmapitem_simple | stackmapitem_stack1 | stackmapitem_append | stackmapitem_full
|
||||
stackmapitem_simple: "same" | "same_extended" | "chop" INT_LITERAL
|
||||
stackmapitem_stack1: ("stack_1" | "stack_1_extended") verification_type
|
||||
stackmapitem_append: "append" vt1to3
|
||||
vt1to3: verification_type verification_type? verification_type?
|
||||
stackmapitem_full: "full" EOL "locals" vtlist "stack" vtlist ".end" "stack"
|
||||
vtlist: verification_type* EOL
|
||||
|
||||
verification_type: "Top" | "Integer" | "Float" | "Double" | "Long" | "Null" | "UninitializedThis" | "Object" clsref | "Uninitialized" label
|
||||
|
||||
instruction_line: (LBLDEF | LBLDEF? instruction) EOL
|
||||
instruction: simple_instruction | complex_instruction
|
||||
simple_instruction: op_none | op_short u8 | op_iinc u8 s8 | op_bipush s8 | op_sipush s16 | op_lbl label | op_fmim fmimref | on_invint fmimref u8? | op_invdyn invdynref | op_cls clsref | op_cls_int clsref u8 | op_ldc ldc_rhs
|
||||
|
||||
op_none: "nop" | "aconst_null" | "iconst_m1" | "iconst_0" | "iconst_1" | "iconst_2" | "iconst_3" | "iconst_4" | "iconst_5" | "lconst_0" | "lconst_1" | "fconst_0" | "fconst_1" | "fconst_2" | "dconst_0" | "dconst_1" | "iload_0" | "iload_1" | "iload_2" | "iload_3" | "lload_0" | "lload_1" | "lload_2" | "lload_3" | "fload_0" | "fload_1" | "fload_2" | "fload_3" | "dload_0" | "dload_1" | "dload_2" | "dload_3" | "aload_0" | "aload_1" | "aload_2" | "aload_3" | "iaload" | "laload" | "faload" | "daload" | "aaload" | "baload" | "caload" | "saload" | "istore_0" | "istore_1" | "istore_2" | "istore_3" | "lstore_0" | "lstore_1" | "lstore_2" | "lstore_3" | "fstore_0" | "fstore_1" | "fstore_2" | "fstore_3" | "dstore_0" | "dstore_1" | "dstore_2" | "dstore_3" | "astore_0" | "astore_1" | "astore_2" | "astore_3" | "iastore" | "lastore" | "fastore" | "dastore" | "aastore" | "bastore" | "castore" | "sastore" | "pop" | "pop2" | "dup" | "dup_x1" | "dup_x2" | "dup2" | "dup2_x1" | "dup2_x2" | "swap" | "iadd" | "ladd" | "fadd" | "dadd" | "isub" | "lsub" | "fsub" | "dsub" | "imul" | "lmul" | "fmul" | "dmul" | "idiv" | "ldiv" | "fdiv" | "ddiv" | "irem" | "lrem" | "frem" | "drem" | "ineg" | "lneg" | "fneg" | "dneg" | "ishl" | "lshl" | "ishr" | "lshr" | "iushr" | "lushr" | "iand" | "land" | "ior" | "lor" | "ixor" | "lxor" | "i2l" | "i2f" | "i2d" | "l2i" | "l2f" | "l2d" | "f2i" | "f2l" | "f2d" | "d2i" | "d2l" | "d2f" | "i2b" | "i2c" | "i2s" | "lcmp" | "fcmpl" | "fcmpg" | "dcmpl" | "dcmpg" | "ireturn" | "lreturn" | "freturn" | "dreturn" | "areturn" | "return" | "arraylength" | "athrow" | "monitorenter" | "monitorexit"
|
||||
|
||||
op_short: "iload" | "lload" | "fload" | "dload" | "aload" | "istore" | "lstore" | "fstore" | "dstore" | "astore" | "ret"
|
||||
|
||||
op_iinc: "iinc"
|
||||
op_bipush: "bipush"
|
||||
op_sipush: "sipush"
|
||||
|
||||
op_lbl: "ifeq" | "ifne" | "iflt" | "ifge" | "ifgt" | "ifle" | "if_icmpeq" | "if_icmpne" | "if_icmplt" | "if_icmpge" | "if_icmpgt" | "if_icmple" | "if_acmpeq" | "if_acmpne" | "goto" | "jsr" | "ifnull" | "ifnonnull" | "goto_w" | "jsr_w"
|
||||
|
||||
op_fmim: "getstatic" | "putstatic" | "getfield" | "putfield" | "invokevirtual" | "invokespecial" | "invokestatic"
|
||||
on_invint: "invokeinterface"
|
||||
op_invdyn: "invokedynamic"
|
||||
|
||||
op_cls: "new" | "anewarray" | "checkcast" | "instanceof"
|
||||
op_cls_int: "multianewarray"
|
||||
|
||||
op_ldc: "ldc" | "ldc_w" | "ldc2_w"
|
||||
|
||||
complex_instruction: ins_newarr | ins_lookupswitch | ins_tableswitch | ins_wide
|
||||
|
||||
ins_newarr: "newarray" nacode
|
||||
nacode: "boolean" | "char" | "float" | "double" | "byte" | "short" | "int" | "long"
|
||||
|
||||
ins_lookupswitch: "lookupswitch" EOL luentry* defaultentry
|
||||
luentry: s32 ":" label EOL
|
||||
defaultentry: "default:" label
|
||||
|
||||
ins_tableswitch: "tableswitch" s32 EOL tblentry* defaultentry
|
||||
tblentry: label EOL
|
||||
|
||||
ins_wide: "wide" (op_short u16 | op_iinc u16 s16)
|
||||
|
||||
label: WORD
|
||||
Annotations
|
||||
element_value_line: element_value EOL
|
||||
element_value: primtag ldc_rhs | "string" utfref | "class" utfref | "enum" utfref utfref | element_value_array | "annotation" annotation_contents annotation_end
|
||||
primtag: "byte" | "char" | "double" | "int" | "float" | "long" | "short" | "boolean"
|
||||
element_value_array: "array" EOL element_value_line* ".end" "array"
|
||||
|
||||
annotation_line: annotation EOL
|
||||
annotation: ".annotation" annotation_contents annotation_end
|
||||
annotation_contents: utfref key_ev_line*
|
||||
key_ev_line: utfref "=" element_value_line
|
||||
annotation_end: ".end" "annotation"
|
||||
|
||||
parameter_annotation_line: parameter_annotation EOL
|
||||
parameter_annotation: ".paramannotation" EOL annotation_line* ".end" "paramannotation"
|
||||
|
||||
type_annotation_line: type_annotation EOL
|
||||
type_annotation: ".typeannotation" u8 target_info EOL target_path EOL type_annotation_rest
|
||||
target_info: type_parameter_target | supertype_target | type_parameter_bound_target | empty_target | method_formal_parameter_target | throws_target | localvar_target | catch_target | offset_target | type_argument_target
|
||||
|
||||
type_parameter_target: "typeparam" u8
|
||||
supertype_target: "super" u16
|
||||
type_parameter_bound_target: "typeparambound" u8 u8
|
||||
empty_target: "empty"
|
||||
method_formal_parameter_target: "methodparam" u8
|
||||
throws_target: "throws" u16
|
||||
|
||||
localvar_target: "localvar" EOL localvarrange* ".end" "localvar"
|
||||
localvarrange: (code_range | "nowhere") u16 EOL
|
||||
|
||||
catch_target: "catch" u16
|
||||
offset_target: "offset" label
|
||||
type_argument_target: "typearg" label u8
|
||||
|
||||
target_path: ".typepath" EOL type_path_segment* ".end" "typepath"
|
||||
type_path_segment: u8 u8 EOL
|
||||
|
||||
type_annotation_rest: annotation_contents ".end" "typeannotation"
|
||||
|
|
@ -0,0 +1,12 @@
|
|||
2018.01.26:
|
||||
Add .noimplicitstackmap
|
||||
|
||||
2016.10.18: (Backwards incompatible)
|
||||
String element values take utfref, rather than ldcrhs
|
||||
|
||||
2016.06.10:
|
||||
Allow +, -, * to be used in WORDS, except at the beginning of a WORD. This allows all signature values to be specified without quotes.
|
||||
Allow argument count of invokeinterface to be omitted when descriptor is specified inline.
|
||||
|
||||
2015.12.23: (Backwards incompatible)
|
||||
Complete rewrite of the assembler. To many changes to list.
|
||||
|
|
@ -0,0 +1,246 @@
|
|||
from .pool import Pool, utf
|
||||
from .writer import Writer, Label
|
||||
|
||||
|
||||
def writeU16Count(data, error, objects, message):
|
||||
count = len(objects)
|
||||
if count >= 1<<16:
|
||||
error('Maximum {} count is {}, found {}.'.format(message, (1<<16)-1, count), objects[-1].tok)
|
||||
data.u16(count)
|
||||
|
||||
class Code(object):
|
||||
def __init__(self, tok, short):
|
||||
self.tok = tok
|
||||
self.short = short
|
||||
self.locals = self.stack = 0
|
||||
|
||||
self.bytecode = Writer()
|
||||
self.exceptions = Writer()
|
||||
self.exceptcount = 0
|
||||
|
||||
self.stackdata = Writer()
|
||||
self.stackcount = 0
|
||||
self.stackcountpos = self.stackdata.ph16()
|
||||
self.laststackoff = -1 # first frame doesn't subtract 1 from offset
|
||||
|
||||
self.stackmaptable = None
|
||||
self.dont_generate_stackmap = False
|
||||
self.attributes = []
|
||||
|
||||
self.labels = {}
|
||||
self.maxcodelen = (1<<16 if short else 1<<32) - 1
|
||||
|
||||
def labeldef(self, lbl, error):
|
||||
if lbl.sym in self.labels:
|
||||
error('Duplicate label definition', lbl.tok,
|
||||
'Previous definition here:', self.labels[lbl.sym][0])
|
||||
self.labels[lbl.sym] = lbl.tok, self.bytecode.pos
|
||||
|
||||
def catch(self, ref, fromlbl, tolbl, usinglbl):
|
||||
self.exceptcount += 1
|
||||
self.exceptions.lbl(fromlbl, 0, 'u16')
|
||||
self.exceptions.lbl(tolbl, 0, 'u16')
|
||||
self.exceptions.lbl(usinglbl, 0, 'u16')
|
||||
self.exceptions.ref(ref)
|
||||
|
||||
def assembleNoCP(self, data, error):
|
||||
bytecode = self.bytecode
|
||||
|
||||
if self.short:
|
||||
data.u8(self.stack), data.u8(self.locals), data.u16(len(bytecode))
|
||||
else:
|
||||
data.u16(self.stack), data.u16(self.locals), data.u32(len(bytecode))
|
||||
data += bytecode
|
||||
|
||||
data.u16(self.exceptcount)
|
||||
data += self.exceptions
|
||||
|
||||
if self.stackmaptable is None and self.stackcount > 0 and not self.dont_generate_stackmap:
|
||||
# Use arbitrary token in case we need to report errors
|
||||
self.stackmaptable = Attribute(self.tok, b'StackMapTable')
|
||||
self.attributes.append(self.stackmaptable)
|
||||
|
||||
if self.stackmaptable:
|
||||
self.stackdata.setph16(self.stackcountpos, self.stackcount)
|
||||
self.stackmaptable.data = self.stackdata
|
||||
|
||||
writeU16Count(data, error, self.attributes, 'attribute')
|
||||
for attr in self.attributes:
|
||||
attr.assembleNoCP(data, error)
|
||||
return data.fillLabels(self.labels, error)
|
||||
|
||||
class Attribute(object):
|
||||
def __init__(self, tok, name, length=None):
|
||||
assert tok
|
||||
if isinstance(name, bytes):
|
||||
name = utf(tok, name)
|
||||
|
||||
self.tok = tok
|
||||
self.name = name
|
||||
self.length = length
|
||||
self.data = Writer()
|
||||
|
||||
def assembleNoCP(self, data, error):
|
||||
length = len(self.data) if self.length is None else self.length
|
||||
if length >= 1<<32:
|
||||
error('Maximum attribute data length is {} bytes, got {} bytes.'.format((1<<32)-1, length), self.tok)
|
||||
|
||||
data.ref(self.name)
|
||||
data.u32(length)
|
||||
data += self.data
|
||||
return data
|
||||
|
||||
class Method(object):
|
||||
def __init__(self, tok, access, name, desc):
|
||||
self.tok = tok
|
||||
self.access = access
|
||||
self.name = name
|
||||
self.desc = desc
|
||||
self.attributes = []
|
||||
|
||||
def assembleNoCP(self, data, error):
|
||||
data.u16(self.access)
|
||||
data.ref(self.name)
|
||||
data.ref(self.desc)
|
||||
|
||||
writeU16Count(data, error, self.attributes, 'attribute')
|
||||
for attr in self.attributes:
|
||||
attr.assembleNoCP(data, error)
|
||||
return data
|
||||
|
||||
class Field(object):
|
||||
def __init__(self, tok, access, name, desc):
|
||||
self.tok = tok
|
||||
self.access = access
|
||||
self.name = name
|
||||
self.desc = desc
|
||||
self.attributes = []
|
||||
|
||||
def assembleNoCP(self, data, error):
|
||||
data.u16(self.access)
|
||||
data.ref(self.name)
|
||||
data.ref(self.desc)
|
||||
|
||||
writeU16Count(data, error, self.attributes, 'attribute')
|
||||
for attr in self.attributes:
|
||||
attr.assembleNoCP(data, error)
|
||||
return data
|
||||
|
||||
class Class(object):
|
||||
def __init__(self):
|
||||
self.version = 49, 0
|
||||
self.access = self.this = self.super = None
|
||||
|
||||
self.interfaces = []
|
||||
self.fields = []
|
||||
self.methods = []
|
||||
self.attributes = []
|
||||
|
||||
self.useshortcodeattrs = False
|
||||
self.bootstrapmethods = None
|
||||
self.pool = Pool()
|
||||
|
||||
def _getName(self):
|
||||
cpool = self.pool.cp
|
||||
clsind = self.this.resolved_index
|
||||
if not cpool.slots.get(clsind):
|
||||
return None
|
||||
|
||||
if cpool.slots[clsind].type != 'Class':
|
||||
return None
|
||||
|
||||
utfind = cpool.slots[clsind].refs[0].resolved_index
|
||||
if utfind not in cpool.slots:
|
||||
return None
|
||||
|
||||
return cpool.slots[utfind].data
|
||||
|
||||
def _assembleNoCP(self, error):
|
||||
beforepool = Writer()
|
||||
afterpool = Writer()
|
||||
|
||||
beforepool.u32(0xCAFEBABE)
|
||||
beforepool.u16(self.version[1])
|
||||
beforepool.u16(self.version[0])
|
||||
|
||||
afterpool.u16(self.access)
|
||||
afterpool.ref(self.this)
|
||||
afterpool.ref(self.super)
|
||||
writeU16Count(afterpool, error, self.interfaces, 'interface')
|
||||
for i in self.interfaces:
|
||||
afterpool.ref(i)
|
||||
|
||||
writeU16Count(afterpool, error, self.fields, 'field')
|
||||
for field in self.fields:
|
||||
field.assembleNoCP(afterpool, error)
|
||||
|
||||
writeU16Count(afterpool, error, self.methods, 'method')
|
||||
for method in self.methods:
|
||||
method.assembleNoCP(afterpool, error)
|
||||
|
||||
attrcountpos = afterpool.ph16()
|
||||
afterbs = Writer()
|
||||
|
||||
data = afterpool
|
||||
for attr in self.attributes:
|
||||
if attr is self.bootstrapmethods:
|
||||
# skip writing this attr for now and switch to after bs stream
|
||||
data = afterbs
|
||||
else:
|
||||
attr.assembleNoCP(data, error)
|
||||
|
||||
return beforepool, afterpool, afterbs, attrcountpos
|
||||
|
||||
def assemble(self, error):
|
||||
beforepool, afterpool, afterbs, attrcountpos = self._assembleNoCP(error)
|
||||
|
||||
self.pool.cp.freezedefs(self.pool, error)
|
||||
self.pool.bs.freezedefs(self.pool, error)
|
||||
|
||||
# afterpool is the only part that can contain ldcs
|
||||
assert not beforepool.refu8phs
|
||||
assert not afterbs.refu8phs
|
||||
for _, ref in afterpool.refu8phs:
|
||||
ind = ref.resolve(self.pool, error)
|
||||
if ind >= 256:
|
||||
error("Ldc references too many distinct constants in this class. If you don't want to see this message again, use ldc_w instead of ldc everywhere.", ref.tok)
|
||||
|
||||
beforepool.fillRefs(self.pool, error)
|
||||
afterpool.fillRefs(self.pool, error)
|
||||
afterbs.fillRefs(self.pool, error)
|
||||
|
||||
# Figure out if we need to add an implicit BootstrapMethods attribute
|
||||
self.pool.resolveIDBSRefs(error)
|
||||
if self.bootstrapmethods is None and self.pool.bs.slots:
|
||||
assert len(afterbs) == 0
|
||||
# Use arbitrary token in case we need to report errors
|
||||
self.bootstrapmethods = Attribute(self.this.tok, b'BootstrapMethods')
|
||||
self.attributes.append(self.bootstrapmethods)
|
||||
if self.bootstrapmethods is not None:
|
||||
self.bootstrapmethods.name.resolve(self.pool, error)
|
||||
assert len(self.bootstrapmethods.data) == 0
|
||||
|
||||
if len(self.attributes) >= 1<<16:
|
||||
error('Maximum class attribute count is 65535, found {}.'.format(count), self.attributes[-1].tok)
|
||||
afterpool.setph16(attrcountpos, len(self.attributes))
|
||||
|
||||
|
||||
cpdata, bsmdata = self.pool.write(error)
|
||||
assert len(bsmdata) < (1 << 32)
|
||||
|
||||
data = beforepool
|
||||
data += cpdata
|
||||
data += afterpool
|
||||
|
||||
if self.bootstrapmethods is not None:
|
||||
self.bootstrapmethods.data = bsmdata
|
||||
self.bootstrapmethods.assembleNoCP(data, error)
|
||||
data.fillRefs(self.pool, error)
|
||||
data += afterbs
|
||||
else:
|
||||
assert len(afterbs) == 0
|
||||
|
||||
name = self._getName()
|
||||
if name is None:
|
||||
error('Invalid reference for class name.', self.this.tok)
|
||||
return name, data.toBytes()
|
||||
|
|
@ -0,0 +1,12 @@
|
|||
_handle_types = 'getField getStatic putField putStatic invokeVirtual invokeStatic invokeSpecial newInvokeSpecial invokeInterface'.split()
|
||||
handle_codes = dict(zip(_handle_types, range(1,10)))
|
||||
handle_rcodes = {v:k for k,v in handle_codes.items()}
|
||||
|
||||
newarr_rcodes = ([None]*4) + 'boolean char float double byte short int long'.split()
|
||||
newarr_codes = dict(zip('boolean char float double byte short int long'.split(), range(4,12)))
|
||||
|
||||
vt_rcodes = ['Top','Integer','Float','Double','Long','Null','UninitializedThis','Object','Uninitialized']
|
||||
vt_codes = {k:i for i,k in enumerate(vt_rcodes)}
|
||||
|
||||
et_rtags = dict(zip(map(ord, 'BCDFIJSZsec@['), 'byte char double float int long short boolean string enum class annotation array'.split()))
|
||||
et_tags = {v:k for k,v in et_rtags.items()}
|
||||
|
|
@ -0,0 +1,843 @@
|
|||
from __future__ import print_function
|
||||
|
||||
import collections
|
||||
import math
|
||||
import re
|
||||
|
||||
from ..classfileformat import classdata, mutf8
|
||||
from ..classfileformat.reader import Reader, TruncatedStreamError
|
||||
from ..util.thunk import thunk
|
||||
|
||||
from . import codes, token_regexes
|
||||
from . import flags
|
||||
from .instructions import OPNAMES, OP_CLS, OP_FMIM, OP_LBL, OP_NONE, OP_SHORT
|
||||
|
||||
MAX_INLINE_SIZE = 300
|
||||
MAX_INDENT = 20
|
||||
WORD_REGEX = re.compile(token_regexes.WORD + r'\Z')
|
||||
|
||||
PREFIXES = {'Utf8': 'u', 'Class': 'c', 'String': 's', 'Field': 'f', 'Method': 'm', 'InterfaceMethod': 'im', 'NameAndType': 'nat', 'MethodHandle': 'mh', 'MethodType': 'mt', 'InvokeDynamic': 'id'}
|
||||
|
||||
class DisassemblyError(Exception):
|
||||
pass
|
||||
|
||||
def reprbytes(b):
|
||||
# repr will have b in python3 but not python2
|
||||
return 'b' + repr(b).lstrip('b')
|
||||
|
||||
def isword(s):
|
||||
try:
|
||||
s = s.decode('ascii')
|
||||
except UnicodeDecodeError:
|
||||
return False
|
||||
return WORD_REGEX.match(s) and s not in flags.FLAGS
|
||||
|
||||
def format_string(s):
|
||||
try:
|
||||
u = mutf8.decode(s)
|
||||
except UnicodeDecodeError:
|
||||
print('Warning, invalid utf8 data!')
|
||||
else:
|
||||
if mutf8.encode(u) == s:
|
||||
return repr(u).lstrip('u')
|
||||
return reprbytes(s)
|
||||
|
||||
def make_signed(x, bits):
|
||||
if x >= (1 << (bits - 1)):
|
||||
x -= 1 << bits
|
||||
return x
|
||||
|
||||
class StackMapReader(object):
|
||||
def __init__(self):
|
||||
self.stream = None
|
||||
self.tag = -1
|
||||
self.pos = -1
|
||||
self.count = 0
|
||||
self.valid = True
|
||||
|
||||
def setdata(self, r):
|
||||
if self.stream is None:
|
||||
self.stream = r
|
||||
self.count = self.u16() + 1
|
||||
self.parseNextPos()
|
||||
else:
|
||||
# Multiple StackMapTable attributes in same Code attribute
|
||||
self.valid = False
|
||||
|
||||
def parseNextPos(self):
|
||||
self.count -= 1
|
||||
if self.count > 0:
|
||||
self.tag = tag = self.u8()
|
||||
|
||||
if tag <= 127: # same and stack_1
|
||||
delta = tag % 64
|
||||
else: # everything else has 16bit delta field
|
||||
delta = self.u16()
|
||||
self.pos += delta + 1
|
||||
|
||||
def u8(self):
|
||||
try:
|
||||
return self.stream.u8()
|
||||
except TruncatedStreamError:
|
||||
self.valid = False
|
||||
return 0
|
||||
|
||||
def u16(self):
|
||||
try:
|
||||
return self.stream.u16()
|
||||
except TruncatedStreamError:
|
||||
self.valid = False
|
||||
return 0
|
||||
|
||||
|
||||
class ReferencePrinter(object):
|
||||
def __init__(self, clsdata, roundtrip):
|
||||
self.roundtrip = roundtrip
|
||||
|
||||
self.cpslots = clsdata.pool.slots
|
||||
for attr in clsdata.getattrs(b'BootstrapMethods'):
|
||||
self.bsslots = classdata.BootstrapMethodsData(attr.stream()).slots
|
||||
break
|
||||
else:
|
||||
self.bsslots = []
|
||||
|
||||
# CP index 0 should always be a raw reference. Additionally, there is one case where exact
|
||||
# references are significant due to a bug in the JVM. In the InnerClasses attribute,
|
||||
# specifying the same index for inner and outer class will fail verification, but specifying
|
||||
# different indexes which point to identical class entries will pass. In this case, we force
|
||||
# references to those indexes to be raw, so they don't get merged and break the class.
|
||||
self.forcedraw = set()
|
||||
for attr in clsdata.getattrs(b'InnerClasses'):
|
||||
r = attr.stream()
|
||||
for _ in range(r.u16()):
|
||||
inner, outer, _, _ = r.u16(), r.u16(), r.u16(), r.u16()
|
||||
if inner != outer and clsdata.pool.getclsutf(inner) == clsdata.pool.getclsutf(outer):
|
||||
self.forcedraw.add(inner)
|
||||
self.forcedraw.add(outer)
|
||||
self.explicit_forcedraw = self.forcedraw.copy()
|
||||
|
||||
# For invalid cp indices, just output raw ref instead of throwing (including 0)
|
||||
for i, slot in enumerate(self.cpslots):
|
||||
if slot.tag is None:
|
||||
self.forcedraw.add(i)
|
||||
self.forcedraw.update(range(len(self.cpslots), 65536))
|
||||
|
||||
self.used = set()
|
||||
self.encoded = {}
|
||||
self.utfcounts = {}
|
||||
|
||||
def _float_or_double(self, x, nmbits, nebits, suffix, nanfmt):
|
||||
nbits = nmbits + nebits + 1
|
||||
assert nbits % 32 == 0
|
||||
|
||||
sbit, ebits, mbits = x >> (nbits - 1), (x >> nmbits) % (1 << nebits), x % (1 << nmbits)
|
||||
if ebits == (1 << nebits) - 1:
|
||||
result = 'NaN' if mbits else 'Infinity'
|
||||
if self.roundtrip and mbits:
|
||||
result += nanfmt.format(x)
|
||||
elif ebits == 0 and mbits == 0:
|
||||
result = '0.0'
|
||||
else:
|
||||
ebias = (1 << (nebits - 1)) - 1
|
||||
exponent = ebits - ebias - nmbits
|
||||
mantissa = mbits
|
||||
if ebits > 0:
|
||||
mantissa += 1 << nmbits
|
||||
else:
|
||||
exponent += 1
|
||||
|
||||
if self.roundtrip:
|
||||
result = '0x{:X}p{}'.format(mantissa, exponent)
|
||||
else:
|
||||
result = repr(math.ldexp(mantissa, exponent))
|
||||
return '+-'[sbit] + result + suffix
|
||||
|
||||
def _int(self, x): return str(make_signed(x, 32))
|
||||
def _long(self, x): return str(make_signed(x, 64)) + 'L'
|
||||
def _float(self, x): return self._float_or_double(x, 23, 8, 'f', '<0x{:08X}>')
|
||||
def _double(self, x): return self._float_or_double(x, 52, 11, '', '<0x{:016X}>')
|
||||
|
||||
def _encode_utf(self, ind, wordok=True):
|
||||
try:
|
||||
return self.encoded[ind][wordok]
|
||||
except KeyError:
|
||||
s = self.cpslots[ind].data
|
||||
string = format_string(s)
|
||||
word = s.decode() if isword(s) else string
|
||||
self.encoded[ind] = [string, word]
|
||||
return word if wordok else string
|
||||
|
||||
def rawref(self, ind, isbs=False):
|
||||
return '[{}{}]'.format('bs:' if isbs else '', ind)
|
||||
|
||||
def symref(self, ind, isbs=False):
|
||||
self.used.add((ind, isbs))
|
||||
if isbs:
|
||||
return '[bs:_{}]'.format(ind)
|
||||
|
||||
prefix = PREFIXES.get(self.cpslots[ind].tag, '_')
|
||||
return '[{}{}]'.format(prefix, ind)
|
||||
|
||||
def ref(self, ind, isbs=False):
|
||||
if self.roundtrip or not isbs and ind in self.forcedraw:
|
||||
return self.rawref(ind, isbs)
|
||||
return self.symref(ind, isbs)
|
||||
|
||||
def _ident(self, ind, wordok=True):
|
||||
if self.cpslots[ind].tag == 'Utf8':
|
||||
val = self._encode_utf(ind, wordok=wordok)
|
||||
if len(val) < MAX_INLINE_SIZE:
|
||||
if len(val) < 50 or self.utfcounts.get(ind, 0) < 10:
|
||||
self.utfcounts[ind] = 1 + self.utfcounts.get(ind, 0)
|
||||
return val
|
||||
|
||||
def utfref(self, ind):
|
||||
if self.roundtrip or ind in self.forcedraw:
|
||||
return self.rawref(ind)
|
||||
temp = self._ident(ind)
|
||||
if temp is not None:
|
||||
return temp
|
||||
return self.symref(ind)
|
||||
|
||||
def clsref(self, ind, tag='Class'):
|
||||
assert tag in 'Class Module Package'.split()
|
||||
if self.roundtrip or ind in self.forcedraw:
|
||||
return self.rawref(ind)
|
||||
if self.cpslots[ind].tag == tag:
|
||||
ind2 = self.cpslots[ind].refs[0]
|
||||
temp = self._ident(ind2)
|
||||
if temp is not None:
|
||||
return temp
|
||||
return self.symref(ind)
|
||||
|
||||
def natref(self, ind):
|
||||
if self.roundtrip or ind in self.forcedraw:
|
||||
return self.rawref(ind)
|
||||
if self.cpslots[ind].tag == 'NameAndType':
|
||||
ind2, ind3 = self.cpslots[ind].refs
|
||||
temp = self._ident(ind2)
|
||||
if temp is not None:
|
||||
return temp + ' ' + self.utfref(ind3)
|
||||
return self.symref(ind)
|
||||
|
||||
def fmimref(self, ind):
|
||||
if self.roundtrip or ind in self.forcedraw:
|
||||
return self.rawref(ind)
|
||||
if self.cpslots[ind].tag in ['Field', 'Method', 'InterfaceMethod']:
|
||||
ind2, ind3 = self.cpslots[ind].refs
|
||||
return ' '.join([self.cpslots[ind].tag, self.clsref(ind2), self.natref(ind3)])
|
||||
return self.symref(ind)
|
||||
|
||||
def mhnotref(self, ind):
|
||||
slot = self.cpslots[ind]
|
||||
return codes.handle_rcodes[slot.data] + ' ' + self.taggedref(slot.refs[0], allowed=['Field', 'Method', 'InterfaceMethod'])
|
||||
|
||||
def taggedconst(self, ind):
|
||||
slot = self.cpslots[ind]
|
||||
if slot.tag == 'Utf8':
|
||||
parts = [self._encode_utf(ind)]
|
||||
elif slot.tag == 'Int':
|
||||
parts = [self._int(slot.data)]
|
||||
elif slot.tag == 'Float':
|
||||
parts = [self._float(slot.data)]
|
||||
elif slot.tag == 'Long':
|
||||
parts = [self._long(slot.data)]
|
||||
elif slot.tag == 'Double':
|
||||
parts = [self._double(slot.data)]
|
||||
elif slot.tag in ['Class', 'String', 'MethodType', 'Module', 'Package']:
|
||||
parts = [self.utfref(slot.refs[0])]
|
||||
elif slot.tag in ['Field', 'Method', 'InterfaceMethod']:
|
||||
parts = [self.clsref(slot.refs[0]), self.natref(slot.refs[1])]
|
||||
elif slot.tag == 'NameAndType':
|
||||
parts = [self.utfref(slot.refs[0]), self.utfref(slot.refs[1])]
|
||||
elif slot.tag == 'MethodHandle':
|
||||
parts = [self.mhnotref(ind)]
|
||||
elif slot.tag == 'InvokeDynamic':
|
||||
parts = [self.bsref(slot.refs[0]), self.natref(slot.refs[1])]
|
||||
parts.insert(0, slot.tag)
|
||||
return ' '.join(parts)
|
||||
|
||||
def taggedref(self, ind, allowed=None):
|
||||
if self.roundtrip or ind in self.forcedraw:
|
||||
return self.rawref(ind)
|
||||
|
||||
if allowed is None or self.cpslots[ind].tag in allowed:
|
||||
temp = self.taggedconst(ind)
|
||||
if len(temp) < MAX_INLINE_SIZE:
|
||||
return temp
|
||||
return self.symref(ind)
|
||||
|
||||
def ldcrhs(self, ind):
|
||||
if self.roundtrip or ind in self.forcedraw:
|
||||
return self.rawref(ind)
|
||||
slot = self.cpslots[ind]
|
||||
t = slot.tag
|
||||
|
||||
if t == 'Int':
|
||||
return self._int(slot.data)
|
||||
elif slot.tag == 'Float':
|
||||
return self._float(slot.data)
|
||||
elif slot.tag == 'Long':
|
||||
return self._long(slot.data)
|
||||
elif slot.tag == 'Double':
|
||||
return self._double(slot.data)
|
||||
elif t == 'String':
|
||||
ind2 = self.cpslots[ind].refs[0]
|
||||
temp = self._ident(ind2, wordok=False)
|
||||
if temp is not None:
|
||||
return temp
|
||||
return self.symref(ind)
|
||||
return self.taggedref(ind, allowed=['Class', 'MethodHandle', 'MethodType'])
|
||||
|
||||
def bsnotref(self, ind, tagged=False):
|
||||
slot = self.bsslots[ind]
|
||||
parts = []
|
||||
if tagged:
|
||||
parts.append('Bootstrap')
|
||||
|
||||
if tagged and self.roundtrip:
|
||||
parts.append(self.rawref(slot.refs[0]))
|
||||
else:
|
||||
parts.append(self.mhnotref(slot.refs[0]))
|
||||
for bsarg in slot.refs[1:]:
|
||||
parts.append(self.taggedref(bsarg))
|
||||
parts.append(':')
|
||||
return ' '.join(parts)
|
||||
|
||||
def bsref(self, ind):
|
||||
if self.roundtrip:
|
||||
return self.rawref(ind, isbs=True)
|
||||
return self.bsnotref(ind)
|
||||
|
||||
LabelInfos = collections.namedtuple('LabelInfos', 'defined used')
|
||||
class Disassembler(object):
|
||||
def __init__(self, clsdata, out, roundtrip):
|
||||
self.roundtrip = roundtrip
|
||||
|
||||
self.out = out
|
||||
self.cls = clsdata
|
||||
self.pool = clsdata.pool
|
||||
|
||||
self.indentlevel = 0
|
||||
self.labels = None
|
||||
self.refprinter = ReferencePrinter(clsdata, roundtrip)
|
||||
|
||||
def _getattr(a, obj, name):
|
||||
for attr in obj.attributes:
|
||||
if a.pool.getutf(attr.name) == name:
|
||||
return attr
|
||||
|
||||
def sol(a, text=''):
|
||||
level = min(a.indentlevel, MAX_INDENT) * 4
|
||||
text += ' ' * (level - len(text))
|
||||
a.out(text)
|
||||
|
||||
def eol(a): a.out('\n')
|
||||
def val(a, s): a.out(s + ' ')
|
||||
def int(a, x): a.val(str(x))
|
||||
|
||||
def lbl(a, x):
|
||||
a.labels.used.add(x)
|
||||
a.val('L{}'.format(x))
|
||||
|
||||
def try_lbl(a, x):
|
||||
if a.labels is None or x not in a.labels.defined:
|
||||
raise DisassemblyError()
|
||||
a.lbl(x)
|
||||
###########################################################################
|
||||
def extrablankline(a): a.eol()
|
||||
|
||||
def ref(a, ind, isbs=False): a.val(a.refprinter.ref(ind, isbs))
|
||||
def utfref(a, ind): a.val(a.refprinter.utfref(ind))
|
||||
def clsref(a, ind, tag='Class'): a.val(a.refprinter.clsref(ind, tag))
|
||||
def natref(a, ind): a.val(a.refprinter.natref(ind))
|
||||
def fmimref(a, ind): a.val(a.refprinter.fmimref(ind))
|
||||
def taggedbs(a, ind): a.val(a.refprinter.bsnotref(ind, tagged=True))
|
||||
def taggedconst(a, ind): a.val(a.refprinter.taggedconst(ind))
|
||||
def taggedref(a, ind): a.val(a.refprinter.taggedref(ind))
|
||||
def ldcrhs(a, ind): a.val(a.refprinter.ldcrhs(ind))
|
||||
|
||||
def flags(a, access, names):
|
||||
for i in range(16):
|
||||
if access & (1 << i):
|
||||
a.val(names[1 << i])
|
||||
|
||||
###########################################################################
|
||||
### Top level stuff (class, const defs, fields, methods) ##################
|
||||
def disassemble(a):
|
||||
cls = a.cls
|
||||
a.val('.version'), a.int(cls.version[0]), a.int(cls.version[1]), a.eol()
|
||||
a.val('.class'), a.flags(cls.access, flags.RFLAGS_CLASS), a.clsref(cls.this), a.eol()
|
||||
a.val('.super'), a.clsref(cls.super), a.eol()
|
||||
for ref in cls.interfaces:
|
||||
a.val('.implements'), a.clsref(ref), a.eol()
|
||||
|
||||
for f in cls.fields:
|
||||
a.field(f)
|
||||
|
||||
for m in cls.methods:
|
||||
a.method(m)
|
||||
|
||||
for attr in cls.attributes:
|
||||
a.attribute(attr)
|
||||
|
||||
a.constdefs()
|
||||
a.val('.end class'), a.eol()
|
||||
|
||||
def field(a, f):
|
||||
a.val('.field'), a.flags(f.access, flags.RFLAGS_FIELD), a.utfref(f.name), a.utfref(f.desc)
|
||||
|
||||
attrs = f.attributes[:]
|
||||
cvattr = a._getattr(f, b'ConstantValue')
|
||||
if cvattr and not a.roundtrip:
|
||||
a.val('='), a.ldcrhs(cvattr.stream().u16())
|
||||
attrs.remove(cvattr)
|
||||
|
||||
if attrs:
|
||||
a.val('.fieldattributes'), a.eol()
|
||||
a.indentlevel += 1
|
||||
for attr in attrs:
|
||||
a.attribute(attr)
|
||||
a.indentlevel -= 1
|
||||
a.val('.end fieldattributes')
|
||||
a.eol()
|
||||
|
||||
def method(a, m):
|
||||
a.extrablankline()
|
||||
a.val('.method'), a.flags(m.access, flags.RFLAGS_METHOD), a.utfref(m.name), a.val(':'), a.utfref(m.desc), a.eol()
|
||||
a.indentlevel += 1
|
||||
for attr in m.attributes:
|
||||
a.attribute(attr, in_method=True)
|
||||
a.indentlevel -= 1
|
||||
a.val('.end method'), a.eol()
|
||||
|
||||
def constdefs(a):
|
||||
if a.roundtrip:
|
||||
for ind in range(len(a.refprinter.cpslots)):
|
||||
a.constdef(ind, False)
|
||||
for ind in range(len(a.refprinter.bsslots)):
|
||||
a.constdef(ind, True)
|
||||
else:
|
||||
assert not a.refprinter.used & a.refprinter.forcedraw
|
||||
for ind in sorted(a.refprinter.explicit_forcedraw):
|
||||
a.constdef(ind, False)
|
||||
|
||||
done = set()
|
||||
while len(done) < len(a.refprinter.used):
|
||||
for ind, isbs in sorted(a.refprinter.used - done):
|
||||
a.constdef(ind, isbs)
|
||||
done.add((ind, isbs))
|
||||
|
||||
def constdef(a, ind, isbs):
|
||||
if not isbs and a.refprinter.cpslots[ind].tag is None:
|
||||
return
|
||||
|
||||
a.sol(), a.val('.bootstrap' if isbs else '.const'), a.ref(ind, isbs), a.val('=')
|
||||
if isbs:
|
||||
a.taggedbs(ind)
|
||||
else:
|
||||
a.taggedconst(ind)
|
||||
a.eol()
|
||||
|
||||
###########################################################################
|
||||
### Bytecode ##############################################################
|
||||
def code(a, r):
|
||||
c = classdata.CodeData(r, a.pool, a.cls.version < (45, 3))
|
||||
a.val('.code'), a.val('stack'), a.int(c.stack), a.val('locals'), a.int(c.locals), a.eol()
|
||||
a.indentlevel += 1
|
||||
assert a.labels is None
|
||||
a.labels = LabelInfos(set(), set())
|
||||
|
||||
stackreader = StackMapReader()
|
||||
for attr in c.attributes:
|
||||
if a.pool.getutf(attr.name) == b'StackMapTable':
|
||||
stackreader.setdata(attr.stream())
|
||||
|
||||
rexcepts = c.exceptions[::-1]
|
||||
bcreader = Reader(c.bytecode)
|
||||
while bcreader.size():
|
||||
a.insline_start(bcreader.off, rexcepts, stackreader)
|
||||
a.instruction(bcreader)
|
||||
a.insline_start(bcreader.off, rexcepts, stackreader), a.eol()
|
||||
|
||||
badlbls = a.labels.used - a.labels.defined
|
||||
if badlbls:
|
||||
stackreader.valid = False
|
||||
a.sol('; Labels used by invalid StackMapTable attribute'), a.eol()
|
||||
for pos in sorted(badlbls):
|
||||
a.sol('L{}'.format(pos) + ':'), a.eol()
|
||||
|
||||
if stackreader.stream and (stackreader.stream.size() or stackreader.count > 0):
|
||||
stackreader.valid = False
|
||||
|
||||
if not stackreader.valid:
|
||||
a.sol('.noimplicitstackmap'), a.eol()
|
||||
|
||||
for attr in c.attributes:
|
||||
a.attribute(attr, use_raw_stackmap=not stackreader.valid)
|
||||
|
||||
a.labels = None
|
||||
a.indentlevel -= 1
|
||||
a.sol(), a.val('.end code')
|
||||
|
||||
def insline_start(a, pos, rexcepts, stackreader):
|
||||
while rexcepts and rexcepts[-1].start <= pos:
|
||||
e = rexcepts.pop()
|
||||
a.sol(), a.val('.catch'), a.clsref(e.type), a.val('from'), a.lbl(e.start)
|
||||
a.val('to'), a.lbl(e.end), a.val('using'), a.lbl(e.handler), a.eol()
|
||||
|
||||
if stackreader.count > 0 and stackreader.pos == pos:
|
||||
r = stackreader
|
||||
tag = stackreader.tag
|
||||
a.extrablankline()
|
||||
a.sol(), a.val('.stack')
|
||||
if tag <= 63:
|
||||
a.val('same')
|
||||
elif tag <= 127:
|
||||
a.val('stack_1'), a.verification_type(r)
|
||||
elif tag == 247:
|
||||
a.val('stack_1_extended'), a.verification_type(r)
|
||||
elif tag < 251:
|
||||
a.val('chop'), a.int(251 - tag)
|
||||
elif tag == 251:
|
||||
a.val('same_extended')
|
||||
elif tag < 255:
|
||||
a.val('append')
|
||||
for _ in range(tag - 251):
|
||||
a.verification_type(r)
|
||||
else:
|
||||
a.val('full')
|
||||
a.indentlevel += 1
|
||||
|
||||
a.eol(), a.sol(), a.val('locals')
|
||||
for _ in range(r.u16()):
|
||||
a.verification_type(r)
|
||||
a.eol(), a.sol(), a.val('stack')
|
||||
for _ in range(r.u16()):
|
||||
a.verification_type(r)
|
||||
|
||||
a.indentlevel -= 1
|
||||
a.eol(), a.sol(), a.val('.end stack')
|
||||
a.eol()
|
||||
stackreader.parseNextPos()
|
||||
|
||||
a.sol('L{}'.format(pos) + ':')
|
||||
a.labels.defined.add(pos)
|
||||
|
||||
def verification_type(a, r):
|
||||
try:
|
||||
tag = codes.vt_rcodes[r.u8()]
|
||||
except IndexError:
|
||||
r.valid = False
|
||||
a.val('Top')
|
||||
return
|
||||
|
||||
a.val(tag)
|
||||
if tag == 'Object':
|
||||
a.clsref(r.u16())
|
||||
elif tag == 'Uninitialized':
|
||||
a.lbl(r.u16())
|
||||
|
||||
def instruction(a, r):
|
||||
pos = r.off
|
||||
op = OPNAMES[r.u8()]
|
||||
a.val(op)
|
||||
|
||||
if op in OP_LBL:
|
||||
a.lbl(pos + (r.s32() if op.endswith('_w') else r.s16()))
|
||||
elif op in OP_SHORT:
|
||||
a.int(r.u8())
|
||||
elif op in OP_CLS:
|
||||
a.clsref(r.u16())
|
||||
elif op in OP_FMIM:
|
||||
a.fmimref(r.u16())
|
||||
elif op == 'invokeinterface':
|
||||
a.fmimref(r.u16()), a.int(r.u8()), r.u8()
|
||||
elif op == 'invokedynamic':
|
||||
a.taggedref(r.u16()), r.u16()
|
||||
elif op in ['ldc', 'ldc_w', 'ldc2_w']:
|
||||
a.ldcrhs(r.u8() if op == 'ldc' else r.u16())
|
||||
elif op == 'multianewarray':
|
||||
a.clsref(r.u16()), a.int(r.u8())
|
||||
elif op == 'bipush':
|
||||
a.int(r.s8())
|
||||
elif op == 'sipush':
|
||||
a.int(r.s16())
|
||||
elif op == 'iinc':
|
||||
a.int(r.u8()), a.int(r.s8())
|
||||
elif op == 'wide':
|
||||
op2 = OPNAMES[r.u8()]
|
||||
a.val(op2), a.int(r.u16())
|
||||
if op2 == 'iinc':
|
||||
a.int(r.s16())
|
||||
elif op == 'newarray':
|
||||
a.val(codes.newarr_rcodes[r.u8()])
|
||||
elif op == 'tableswitch':
|
||||
r.getRaw((3-pos) % 4)
|
||||
default = pos + r.s32()
|
||||
low, high = r.s32(), r.s32()
|
||||
|
||||
a.int(low), a.eol()
|
||||
a.indentlevel += 1
|
||||
for _ in range(high - low + 1):
|
||||
a.sol(), a.lbl(pos + r.s32()), a.eol()
|
||||
a.sol(), a.val('default'), a.val(':'), a.lbl(default), a.eol()
|
||||
a.indentlevel -= 1
|
||||
elif op == 'lookupswitch':
|
||||
r.getRaw((3-pos) % 4)
|
||||
default = pos + r.s32()
|
||||
|
||||
a.eol()
|
||||
a.indentlevel += 1
|
||||
for _ in range(r.s32()):
|
||||
a.sol(), a.int(r.s32()), a.val(':'), a.lbl(pos + r.s32()), a.eol()
|
||||
a.sol(), a.val('default'), a.val(':'), a.lbl(default), a.eol()
|
||||
a.indentlevel -= 1
|
||||
else:
|
||||
assert op in OP_NONE
|
||||
a.eol()
|
||||
|
||||
###########################################################################
|
||||
### Attributes ############################################################
|
||||
def attribute(a, attr, in_method=False, use_raw_stackmap=False):
|
||||
name = a.pool.getutf(attr.name)
|
||||
if not a.roundtrip and name in (b'BootstrapMethods', b'StackMapTable'):
|
||||
return
|
||||
|
||||
# a.extrablankline()
|
||||
a.sol()
|
||||
isnamed = False
|
||||
if a.roundtrip or name is None:
|
||||
isnamed = True
|
||||
a.val('.attribute'), a.utfref(attr.name)
|
||||
if attr.wronglength:
|
||||
a.val('length'), a.int(attr.length)
|
||||
|
||||
if name == b'Code' and in_method:
|
||||
a.code(attr.stream())
|
||||
elif name == b'BootstrapMethods' and a.cls.version >= (51, 0):
|
||||
a.val('.bootstrapmethods')
|
||||
elif name == b'StackMapTable' and not use_raw_stackmap:
|
||||
a.val('.stackmaptable')
|
||||
elif a.attribute_fallible(name, attr):
|
||||
pass
|
||||
else:
|
||||
print('Nonstandard attribute', name[:70], len(attr.raw))
|
||||
if not isnamed:
|
||||
a.val('.attribute'), a.utfref(attr.name)
|
||||
a.val(reprbytes(attr.raw))
|
||||
|
||||
a.eol()
|
||||
|
||||
def attribute_fallible(a, name, attr):
|
||||
# Temporarily buffer output so we don't get partial output if attribute disassembly fails
|
||||
# in case of failure, we'll fall back to binary output in the caller
|
||||
orig_out = a.out
|
||||
buffer_ = []
|
||||
a.out = buffer_.append
|
||||
|
||||
try:
|
||||
r = attr.stream()
|
||||
if name == b'AnnotationDefault':
|
||||
a.val('.annotationdefault'), a.element_value(r)
|
||||
elif name == b'ConstantValue':
|
||||
a.val('.constantvalue'), a.ldcrhs(r.u16())
|
||||
elif name == b'Deprecated':
|
||||
a.val('.deprecated')
|
||||
elif name == b'EnclosingMethod':
|
||||
a.val('.enclosing method'), a.clsref(r.u16()), a.natref(r.u16())
|
||||
elif name == b'Exceptions':
|
||||
a.val('.exceptions')
|
||||
for _ in range(r.u16()):
|
||||
a.clsref(r.u16())
|
||||
elif name == b'InnerClasses':
|
||||
a.indented_line_list(r, a._innerclasses_item, 'innerclasses')
|
||||
elif name == b'LineNumberTable':
|
||||
a.indented_line_list(r, a._linenumber_item, 'linenumbertable')
|
||||
elif name == b'LocalVariableTable':
|
||||
a.indented_line_list(r, a._localvariabletable_item, 'localvariabletable')
|
||||
elif name == b'LocalVariableTypeTable':
|
||||
a.indented_line_list(r, a._localvariabletable_item, 'localvariabletypetable')
|
||||
elif name == b'MethodParameters':
|
||||
a.indented_line_list(r, a._methodparams_item, 'methodparameters', bytelen=True)
|
||||
elif name == b'Module':
|
||||
a.module_attr(r)
|
||||
elif name == b'ModuleMainClass':
|
||||
a.val('.modulemainclass'), a.clsref(r.u16())
|
||||
elif name == b'ModulePackages':
|
||||
a.val('.modulepackages')
|
||||
for _ in range(r.u16()):
|
||||
a.clsref(r.u16(), tag='Package')
|
||||
elif name in (b'RuntimeVisibleAnnotations', b'RuntimeVisibleParameterAnnotations',
|
||||
b'RuntimeVisibleTypeAnnotations', b'RuntimeInvisibleAnnotations',
|
||||
b'RuntimeInvisibleParameterAnnotations', b'RuntimeInvisibleTypeAnnotations'):
|
||||
a.val('.runtime')
|
||||
a.val('invisible' if b'Inv' in name else 'visible')
|
||||
if b'Type' in name:
|
||||
a.val('typeannotations'), a.eol()
|
||||
a.indented_line_list(r, a.type_annotation_line, 'runtime', False)
|
||||
elif b'Parameter' in name:
|
||||
a.val('paramannotations'), a.eol()
|
||||
a.indented_line_list(r, a.param_annotation_line, 'runtime', False, bytelen=True)
|
||||
else:
|
||||
a.val('annotations'), a.eol()
|
||||
a.indented_line_list(r, a.annotation_line, 'runtime', False)
|
||||
|
||||
elif name == b'Signature':
|
||||
a.val('.signature'), a.utfref(r.u16())
|
||||
elif name == b'SourceDebugExtension':
|
||||
a.val('.sourcedebugextension')
|
||||
a.val(reprbytes(attr.raw))
|
||||
elif name == b'SourceFile':
|
||||
a.val('.sourcefile'), a.utfref(r.u16())
|
||||
elif name == b'Synthetic':
|
||||
a.val('.synthetic')
|
||||
|
||||
# check for extra data in the attribute
|
||||
if r.size():
|
||||
raise DisassemblyError()
|
||||
|
||||
except (TruncatedStreamError, DisassemblyError):
|
||||
a.out = orig_out
|
||||
return False
|
||||
|
||||
a.out = orig_out
|
||||
a.out(''.join(buffer_))
|
||||
return True
|
||||
|
||||
def module_attr(a, r):
|
||||
a.val('.module'), a.clsref(r.u16(), tag='Module'), a.flags(r.u16(), flags.RFLAGS_MOD_OTHER)
|
||||
a.val('version'), a.utfref(r.u16()), a.eol()
|
||||
a.indentlevel += 1
|
||||
|
||||
for _ in range(r.u16()):
|
||||
a.sol(), a.val('.requires'), a.clsref(r.u16(), tag='Module'), a.flags(r.u16(), flags.RFLAGS_MOD_REQUIRES), a.val('version'), a.utfref(r.u16()), a.eol()
|
||||
|
||||
for dir_ in ('.exports', '.opens'):
|
||||
for _ in range(r.u16()):
|
||||
a.sol(), a.val(dir_), a.clsref(r.u16(), tag='Package'), a.flags(r.u16(), flags.RFLAGS_MOD_OTHER)
|
||||
count = r.u16()
|
||||
if count:
|
||||
a.val('to')
|
||||
for _ in range(count):
|
||||
a.clsref(r.u16(), tag='Module')
|
||||
a.eol()
|
||||
|
||||
for _ in range(r.u16()):
|
||||
a.sol(), a.val('.uses'), a.clsref(r.u16()), a.eol()
|
||||
|
||||
for _ in range(r.u16()):
|
||||
a.sol(), a.val('.provides'), a.clsref(r.u16()), a.val('with')
|
||||
for _ in range(r.u16()):
|
||||
a.clsref(r.u16())
|
||||
a.eol()
|
||||
a.indentlevel -= 1
|
||||
a.sol(), a.val('.end module')
|
||||
|
||||
def indented_line_list(a, r, cb, dirname, dostart=True, bytelen=False):
|
||||
if dostart:
|
||||
a.val('.' + dirname), a.eol()
|
||||
a.indentlevel += 1
|
||||
for _ in range(r.u8() if bytelen else r.u16()):
|
||||
a.sol(), cb(r), a.eol()
|
||||
a.indentlevel -= 1
|
||||
if dirname is not None:
|
||||
a.sol(), a.val('.end ' + dirname)
|
||||
|
||||
def _innerclasses_item(a, r): a.clsref(r.u16()), a.clsref(r.u16()), a.utfref(r.u16()), a.flags(r.u16(), flags.RFLAGS_CLASS)
|
||||
def _linenumber_item(a, r): a.try_lbl(r.u16()), a.int(r.u16())
|
||||
def _localvariabletable_item(a, r):
|
||||
start, length, name, desc, ind = r.u16(), r.u16(), r.u16(), r.u16(), r.u16()
|
||||
a.int(ind), a.val('is'), a.utfref(name), a.utfref(desc),
|
||||
a.val('from'), a.try_lbl(start), a.val('to'), a.try_lbl(start + length)
|
||||
def _methodparams_item(a, r): a.utfref(r.u16()), a.flags(r.u16(), flags.RFLAGS_MOD_OTHER)
|
||||
|
||||
###########################################################################
|
||||
### Annotations ###########################################################
|
||||
def annotation_line(a, r):
|
||||
a.val('.annotation'), a.annotation_contents(r), a.sol(), a.val('.end'), a.val('annotation')
|
||||
|
||||
def param_annotation_line(a, r):
|
||||
a.indented_line_list(r, a.annotation_line, 'paramannotation')
|
||||
|
||||
def type_annotation_line(a, r):
|
||||
a.val('.typeannotation')
|
||||
a.indentlevel += 1
|
||||
a.ta_target_info(r) # Note: begins on same line as .typeannotation
|
||||
a.ta_target_path(r)
|
||||
a.sol(), a.annotation_contents(r),
|
||||
a.indentlevel -= 1
|
||||
a.sol(), a.val('.end'), a.val('typeannotation')
|
||||
|
||||
def ta_target_info(a, r):
|
||||
tag = r.u8()
|
||||
a.int(tag)
|
||||
if tag <= 0x01:
|
||||
a.val('typeparam'), a.int(r.u8())
|
||||
elif tag <= 0x10:
|
||||
a.val('super'), a.int(r.u16())
|
||||
elif tag <= 0x12:
|
||||
a.val('typeparambound'), a.int(r.u8()), a.int(r.u8())
|
||||
elif tag <= 0x15:
|
||||
a.val('empty')
|
||||
elif tag <= 0x16:
|
||||
a.val('methodparam'), a.int(r.u8())
|
||||
elif tag <= 0x17:
|
||||
a.val('throws'), a.int(r.u16())
|
||||
elif tag <= 0x41:
|
||||
a.val('localvar'), a.eol()
|
||||
a.indented_line_list(r, a._localvarrange, 'localvar', False)
|
||||
elif tag <= 0x42:
|
||||
a.val('catch'), a.int(r.u16())
|
||||
elif tag <= 0x46:
|
||||
a.val('offset'), a.try_lbl(r.u16())
|
||||
else:
|
||||
a.val('typearg'), a.try_lbl(r.u16()), a.int(r.u8())
|
||||
a.eol()
|
||||
|
||||
def _localvarrange(a, r):
|
||||
start, length, index = r.u16(), r.u16(), r.u16()
|
||||
if start == length == 0xFFFF: # WTF, Java?
|
||||
a.val('nowhere')
|
||||
else:
|
||||
a.val('from'), a.try_lbl(start), a.val('to'), a.try_lbl(start + length)
|
||||
a.int(index)
|
||||
|
||||
def ta_target_path(a, r):
|
||||
a.sol(), a.indented_line_list(r, a._type_path_segment, 'typepath', bytelen=True), a.eol()
|
||||
|
||||
def _type_path_segment(a, r):
|
||||
a.int(r.u8()), a.int(r.u8())
|
||||
|
||||
# The following are recursive and can be nested arbitrarily deep,
|
||||
# so we use generators and a thunk to avoid the Python stack limit.
|
||||
def element_value(a, r): thunk(a._element_value(r))
|
||||
def annotation_contents(a, r): thunk(a._annotation_contents(r))
|
||||
|
||||
def _element_value(a, r):
|
||||
tag = codes.et_rtags.get(r.u8())
|
||||
if tag is None:
|
||||
raise DisassemblyError()
|
||||
a.val(tag)
|
||||
if tag == 'annotation':
|
||||
(yield a._annotation_contents(r)), a.sol(), a.val('.end'), a.val('annotation')
|
||||
elif tag == 'array':
|
||||
a.eol()
|
||||
a.indentlevel += 1
|
||||
for _ in range(r.u16()):
|
||||
a.sol(), (yield a._element_value(r)), a.eol()
|
||||
a.indentlevel -= 1
|
||||
a.sol(), a.val('.end'), a.val('array')
|
||||
elif tag == 'enum':
|
||||
a.utfref(r.u16()), a.utfref(r.u16())
|
||||
elif tag == 'class' or tag == 'string':
|
||||
a.utfref(r.u16())
|
||||
else:
|
||||
a.ldcrhs(r.u16())
|
||||
|
||||
def _annotation_contents(a, r):
|
||||
a.utfref(r.u16()), a.eol()
|
||||
a.indentlevel += 1
|
||||
for _ in range(r.u16()):
|
||||
a.sol(), a.utfref(r.u16()), a.val('='), (yield a._element_value(r)), a.eol()
|
||||
a.indentlevel -= 1
|
||||
|
|
@ -0,0 +1,43 @@
|
|||
_pairs = [
|
||||
('public', 0x0001),
|
||||
('private', 0x0002),
|
||||
('protected', 0x0004),
|
||||
('static', 0x0008),
|
||||
('final', 0x0010),
|
||||
('super', 0x0020),
|
||||
# todo - order attributes properly by context
|
||||
('transitive', 0x0020),
|
||||
('open', 0x0020),
|
||||
('synchronized', 0x0020),
|
||||
('volatile', 0x0040),
|
||||
('static_phase', 0x0040),
|
||||
('bridge', 0x0040),
|
||||
('transient', 0x0080),
|
||||
('varargs', 0x0080),
|
||||
('native', 0x0100),
|
||||
('interface', 0x0200),
|
||||
('abstract', 0x0400),
|
||||
('strict', 0x0800),
|
||||
('synthetic', 0x1000),
|
||||
('annotation', 0x2000),
|
||||
('enum', 0x4000),
|
||||
('module', 0x8000),
|
||||
('mandated', 0x8000),
|
||||
]
|
||||
|
||||
FLAGS = dict(_pairs)
|
||||
# Treat strictfp as flag too to reduce confusion
|
||||
FLAGS['strictfp'] = FLAGS['strict']
|
||||
|
||||
def _make_dict(priority):
|
||||
d = {v:k for k,v in reversed(_pairs)}
|
||||
# ensure that the specified flags have priority
|
||||
for flag in priority.split():
|
||||
d[FLAGS[flag]] = flag
|
||||
return d
|
||||
|
||||
RFLAGS_CLASS = _make_dict('super module')
|
||||
RFLAGS_FIELD = _make_dict('volatile transient')
|
||||
RFLAGS_METHOD = _make_dict('synchronized bridge varargs')
|
||||
RFLAGS_MOD_REQUIRES = _make_dict('transitive static_phase mandated')
|
||||
RFLAGS_MOD_OTHER = _make_dict('open mandated')
|
||||
|
|
@ -0,0 +1,18 @@
|
|||
OP_NONE = frozenset(['nop', 'aconst_null', 'iconst_m1', 'iconst_0', 'iconst_1', 'iconst_2', 'iconst_3', 'iconst_4', 'iconst_5', 'lconst_0', 'lconst_1', 'fconst_0', 'fconst_1', 'fconst_2', 'dconst_0', 'dconst_1', 'iload_0', 'iload_1', 'iload_2', 'iload_3', 'lload_0', 'lload_1', 'lload_2', 'lload_3', 'fload_0', 'fload_1', 'fload_2', 'fload_3', 'dload_0', 'dload_1', 'dload_2', 'dload_3', 'aload_0', 'aload_1', 'aload_2', 'aload_3', 'iaload', 'laload', 'faload', 'daload', 'aaload', 'baload', 'caload', 'saload', 'istore_0', 'istore_1', 'istore_2', 'istore_3', 'lstore_0', 'lstore_1', 'lstore_2', 'lstore_3', 'fstore_0', 'fstore_1', 'fstore_2', 'fstore_3', 'dstore_0', 'dstore_1', 'dstore_2', 'dstore_3', 'astore_0', 'astore_1', 'astore_2', 'astore_3', 'iastore', 'lastore', 'fastore', 'dastore', 'aastore', 'bastore', 'castore', 'sastore', 'pop', 'pop2', 'dup', 'dup_x1', 'dup_x2', 'dup2', 'dup2_x1', 'dup2_x2', 'swap', 'iadd', 'ladd', 'fadd', 'dadd', 'isub', 'lsub', 'fsub', 'dsub', 'imul', 'lmul', 'fmul', 'dmul', 'idiv', 'ldiv', 'fdiv', 'ddiv', 'irem', 'lrem', 'frem', 'drem', 'ineg', 'lneg', 'fneg', 'dneg', 'ishl', 'lshl', 'ishr', 'lshr', 'iushr', 'lushr', 'iand', 'land', 'ior', 'lor', 'ixor', 'lxor', 'i2l', 'i2f', 'i2d', 'l2i', 'l2f', 'l2d', 'f2i', 'f2l', 'f2d', 'd2i', 'd2l', 'd2f', 'i2b', 'i2c', 'i2s', 'lcmp', 'fcmpl', 'fcmpg', 'dcmpl', 'dcmpg', 'ireturn', 'lreturn', 'freturn', 'dreturn', 'areturn', 'return', 'arraylength', 'athrow', 'monitorenter', 'monitorexit'])
|
||||
OP_SHORT = frozenset(['iload', 'lload', 'fload', 'dload', 'aload', 'istore', 'lstore', 'fstore', 'dstore', 'astore', 'ret'])
|
||||
OP_LBL = frozenset(['ifeq', 'ifne', 'iflt', 'ifge', 'ifgt', 'ifle', 'if_icmpeq', 'if_icmpne', 'if_icmplt', 'if_icmpge', 'if_icmpgt', 'if_icmple', 'if_acmpeq', 'if_acmpne', 'goto', 'jsr', 'ifnull', 'ifnonnull', 'goto_w', 'jsr_w'])
|
||||
OP_CLS = frozenset(['new', 'anewarray', 'checkcast', 'instanceof'])
|
||||
|
||||
OP_FMIM_TO_GUESS = {
|
||||
'getstatic': 'Field',
|
||||
'putstatic': 'Field',
|
||||
'getfield': 'Field',
|
||||
'putfield': 'Field',
|
||||
'invokevirtual': 'Method',
|
||||
'invokespecial': 'Method',
|
||||
'invokestatic': 'Method',
|
||||
}
|
||||
OP_FMIM = frozenset(OP_FMIM_TO_GUESS)
|
||||
|
||||
OPNAMES = ('nop', 'aconst_null', 'iconst_m1', 'iconst_0', 'iconst_1', 'iconst_2', 'iconst_3', 'iconst_4', 'iconst_5', 'lconst_0', 'lconst_1', 'fconst_0', 'fconst_1', 'fconst_2', 'dconst_0', 'dconst_1', 'bipush', 'sipush', 'ldc', 'ldc_w', 'ldc2_w', 'iload', 'lload', 'fload', 'dload', 'aload', 'iload_0', 'iload_1', 'iload_2', 'iload_3', 'lload_0', 'lload_1', 'lload_2', 'lload_3', 'fload_0', 'fload_1', 'fload_2', 'fload_3', 'dload_0', 'dload_1', 'dload_2', 'dload_3', 'aload_0', 'aload_1', 'aload_2', 'aload_3', 'iaload', 'laload', 'faload', 'daload', 'aaload', 'baload', 'caload', 'saload', 'istore', 'lstore', 'fstore', 'dstore', 'astore', 'istore_0', 'istore_1', 'istore_2', 'istore_3', 'lstore_0', 'lstore_1', 'lstore_2', 'lstore_3', 'fstore_0', 'fstore_1', 'fstore_2', 'fstore_3', 'dstore_0', 'dstore_1', 'dstore_2', 'dstore_3', 'astore_0', 'astore_1', 'astore_2', 'astore_3', 'iastore', 'lastore', 'fastore', 'dastore', 'aastore', 'bastore', 'castore', 'sastore', 'pop', 'pop2', 'dup', 'dup_x1', 'dup_x2', 'dup2', 'dup2_x1', 'dup2_x2', 'swap', 'iadd', 'ladd', 'fadd', 'dadd', 'isub', 'lsub', 'fsub', 'dsub', 'imul', 'lmul', 'fmul', 'dmul', 'idiv', 'ldiv', 'fdiv', 'ddiv', 'irem', 'lrem', 'frem', 'drem', 'ineg', 'lneg', 'fneg', 'dneg', 'ishl', 'lshl', 'ishr', 'lshr', 'iushr', 'lushr', 'iand', 'land', 'ior', 'lor', 'ixor', 'lxor', 'iinc', 'i2l', 'i2f', 'i2d', 'l2i', 'l2f', 'l2d', 'f2i', 'f2l', 'f2d', 'd2i', 'd2l', 'd2f', 'i2b', 'i2c', 'i2s', 'lcmp', 'fcmpl', 'fcmpg', 'dcmpl', 'dcmpg', 'ifeq', 'ifne', 'iflt', 'ifge', 'ifgt', 'ifle', 'if_icmpeq', 'if_icmpne', 'if_icmplt', 'if_icmpge', 'if_icmpgt', 'if_icmple', 'if_acmpeq', 'if_acmpne', 'goto', 'jsr', 'ret', 'tableswitch', 'lookupswitch', 'ireturn', 'lreturn', 'freturn', 'dreturn', 'areturn', 'return', 'getstatic', 'putstatic', 'getfield', 'putfield', 'invokevirtual', 'invokespecial','invokestatic', 'invokeinterface', 'invokedynamic', 'new', 'newarray', 'anewarray', 'arraylength', 'athrow', 'checkcast', 'instanceof', 'monitorenter', 'monitorexit', 'wide', 'multianewarray', 'ifnull', 'ifnonnull', 'goto_w', 'jsr_w')
|
||||
OPNAME_TO_BYTE = {v:i for i, v in enumerate(OPNAMES)}
|
||||
File diff suppressed because it is too large
Load Diff
|
|
@ -0,0 +1,259 @@
|
|||
import collections
|
||||
import itertools
|
||||
|
||||
from .writer import Writer
|
||||
|
||||
TAGS = [None, 'Utf8', None, 'Int', 'Float', 'Long', 'Double', 'Class', 'String', 'Field', 'Method', 'InterfaceMethod', 'NameAndType', None, None, 'MethodHandle', 'MethodType', None, 'InvokeDynamic', 'Module', 'Package']
|
||||
|
||||
class Ref(object):
|
||||
def __init__(self, tok, index=None, symbol=None, type=None, refs=None, data=None, isbs=False):
|
||||
self.tok = tok
|
||||
self.isbs = isbs
|
||||
self.index = index
|
||||
self.symbol = symbol
|
||||
|
||||
assert type == 'Bootstrap' or type in TAGS
|
||||
self.type = type
|
||||
self.refs = refs or []
|
||||
self.data = data
|
||||
|
||||
self.resolved_index = None
|
||||
|
||||
def israw(self): return self.index is not None
|
||||
def issym(self): return self.symbol is not None
|
||||
|
||||
def _deepdata(self, pool, error, defstack=()):
|
||||
if self.issym():
|
||||
return pool.sub(self).getroot(self, error)._deepdata(pool, error, defstack)
|
||||
|
||||
if self.israw():
|
||||
return 'Raw', self.index
|
||||
|
||||
if len(defstack) > 5: # Maximum legitimate depth is 5: ID -> BS -> MH -> F -> NAT -> UTF
|
||||
error_args = ['Constant pool definitions cannot be nested more than 5 deep (excluding raw references).', self.tok]
|
||||
for ref in reversed(defstack):
|
||||
error_args.append('Included from {} ref here:'.format(ref.type))
|
||||
error_args.append(ref.tok)
|
||||
error(*error_args)
|
||||
return self.type, self.data, tuple(ref._deepdata(pool, error, defstack + (self,)) for ref in self.refs)
|
||||
|
||||
def _resolve(self, pool, error):
|
||||
if self.israw():
|
||||
return self.index
|
||||
if self.issym():
|
||||
return pool.sub(self).getroot(self, error).resolve(pool, error)
|
||||
return pool.sub(self).resolvedata(self, error, self._deepdata(pool, error))
|
||||
|
||||
def resolve(self, pool, error):
|
||||
if self.resolved_index is None:
|
||||
self.resolved_index = self._resolve(pool, error)
|
||||
assert self.resolved_index is not None
|
||||
return self.resolved_index
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
prefix = 'bs:' if self.isbs else ''
|
||||
if self.israw():
|
||||
return '[{}{}]'.format(prefix, self.index)
|
||||
elif self.issym():
|
||||
return '[{}{}]'.format(prefix, self.symbol)
|
||||
parts = [self.type] + self.refs
|
||||
if self.data is not None:
|
||||
parts.insert(1, self.data)
|
||||
return ' '.join(map(str, parts))
|
||||
|
||||
def utf(tok, s):
|
||||
assert isinstance(s, bytes)
|
||||
assert len(s) <= 65535
|
||||
return Ref(tok, type='Utf8', data=s)
|
||||
|
||||
def single(type, tok, s):
|
||||
assert type in 'Class String MethodType Module Package'.split()
|
||||
return Ref(tok, type=type, refs=[utf(tok, s)])
|
||||
|
||||
def nat(name, desc):
|
||||
return Ref(name.tok, type='NameAndType', refs=[name, desc])
|
||||
|
||||
def primitive(type, tok, x):
|
||||
assert type in 'Int Long Float Double'.split()
|
||||
return Ref(tok, type=type, data=x)
|
||||
|
||||
class PoolSub(object):
|
||||
def __init__(self, isbs):
|
||||
self.isbs = isbs
|
||||
self.symdefs = {}
|
||||
self.symrootdefs = {}
|
||||
self.slot_def_tokens = {}
|
||||
self.slots = collections.OrderedDict()
|
||||
self.dataToSlot = {}
|
||||
self.narrowcounter = itertools.count()
|
||||
self.widecounter = itertools.count()
|
||||
|
||||
self.dirtyslotdefs = []
|
||||
self.defsfrozen = False
|
||||
if not isbs:
|
||||
self.slots[0] = None
|
||||
|
||||
def adddef(self, lhs, rhs, error):
|
||||
assert not self.defsfrozen
|
||||
assert lhs.israw() or lhs.issym()
|
||||
if lhs.israw():
|
||||
if lhs.index == 0 and not self.isbs:
|
||||
error('Constant pool index must be nonzero', lhs.tok)
|
||||
|
||||
if lhs.index in self.slots:
|
||||
error('Conflicting raw reference definition', lhs.tok,
|
||||
'Conflicts with previous definition:', self.slot_def_tokens[lhs.index])
|
||||
self.slots[lhs.index] = rhs
|
||||
self.slot_def_tokens[lhs.index] = lhs.tok
|
||||
self.dirtyslotdefs.append(lhs.index)
|
||||
assert rhs.type
|
||||
if rhs.type in ('Long', 'Double'):
|
||||
if lhs.index + 1 in self.slots:
|
||||
error('Conflicting raw reference definition', lhs.tok,
|
||||
'Conflicts with previous definition:', self.slot_def_tokens[lhs.index + 1])
|
||||
self.slots[lhs.index + 1] = None
|
||||
self.slot_def_tokens[lhs.index + 1] = lhs.tok
|
||||
else:
|
||||
if lhs.symbol in self.symdefs:
|
||||
error('Duplicate symbolic reference definition', lhs.tok,
|
||||
'Previously defined here:', self.symdefs[lhs.symbol][0])
|
||||
self.symdefs[lhs.symbol] = lhs.tok, rhs
|
||||
|
||||
def freezedefs(self, pool, error): self.defsfrozen = True
|
||||
|
||||
def _getslot(self, iswide):
|
||||
assert self.defsfrozen
|
||||
if iswide:
|
||||
ind = next(self.widecounter)
|
||||
while ind in self.slots or ind + 1 in self.slots:
|
||||
ind = next(self.widecounter)
|
||||
if ind + 1 >= 0xFFFF:
|
||||
return None
|
||||
else:
|
||||
ind = next(self.narrowcounter)
|
||||
while ind in self.slots:
|
||||
ind = next(self.narrowcounter)
|
||||
if ind >= 0xFFFF:
|
||||
return None
|
||||
return ind
|
||||
|
||||
def getroot(self, ref, error):
|
||||
assert self.defsfrozen and ref.issym()
|
||||
|
||||
try:
|
||||
return self.symrootdefs[ref.symbol]
|
||||
except KeyError:
|
||||
stack = []
|
||||
visited = set()
|
||||
while ref.issym():
|
||||
sym = ref.symbol
|
||||
if sym in visited:
|
||||
error_args = ['Circular symbolic reference', ref.tok]
|
||||
for tok in stack[::-1]:
|
||||
error_args.extend(('Included from here:', tok))
|
||||
error(*error_args)
|
||||
stack.append(ref.tok)
|
||||
visited.add(sym)
|
||||
|
||||
if sym not in self.symdefs:
|
||||
error('Undefined symbolic reference', ref.tok)
|
||||
_, ref = self.symdefs[sym]
|
||||
|
||||
for sym in visited:
|
||||
self.symrootdefs[sym] = ref
|
||||
return ref
|
||||
|
||||
def resolvedata(self, ref, error, newdata):
|
||||
try:
|
||||
return self.dataToSlot[newdata]
|
||||
except KeyError:
|
||||
iswide = newdata[0] in ('Long', 'Double')
|
||||
slot = self._getslot(iswide)
|
||||
if slot is None:
|
||||
name = 'bootstrap method' if ref.isbs else 'constant pool'
|
||||
error('Exhausted {} space.'.format(name), ref.tok)
|
||||
|
||||
self.dataToSlot[newdata] = slot
|
||||
self.slots[slot] = ref
|
||||
self.dirtyslotdefs.append(slot)
|
||||
if iswide:
|
||||
self.slots[slot + 1] = None
|
||||
return slot
|
||||
|
||||
def resolveslotrefs(self, pool, error):
|
||||
while len(self.dirtyslotdefs) > 0:
|
||||
i = self.dirtyslotdefs.pop()
|
||||
for ref in self.slots[i].refs:
|
||||
ref.resolve(pool, error)
|
||||
|
||||
def writeconst(self, w, ref, pool, error):
|
||||
t = ref.type
|
||||
w.u8(TAGS.index(t))
|
||||
if t == 'Utf8':
|
||||
w.u16(len(ref.data))
|
||||
w.writeBytes(ref.data)
|
||||
elif t == 'Int' or t == 'Float':
|
||||
w.u32(ref.data)
|
||||
elif t == 'Long' or t == 'Double':
|
||||
w.u64(ref.data)
|
||||
elif t == 'MethodHandle':
|
||||
w.u8(ref.data)
|
||||
w.u16(ref.refs[0].resolve(pool, error))
|
||||
else:
|
||||
for child in ref.refs:
|
||||
w.u16(child.resolve(pool, error))
|
||||
return w
|
||||
|
||||
def writebootstrap(self, w, ref, pool, error):
|
||||
assert ref.type == 'Bootstrap'
|
||||
w.u16(ref.refs[0].resolve(pool, error))
|
||||
w.u16(len(ref.refs)-1)
|
||||
for child in ref.refs[1:]:
|
||||
w.u16(child.resolve(pool, error))
|
||||
return w
|
||||
|
||||
def write(self, pool, error):
|
||||
self.resolveslotrefs(pool, error)
|
||||
self.dirtyslotdefs = None # make sure we don't accidently add entries after size is taken
|
||||
|
||||
size = max(self.slots) + 1 if self.slots else 0
|
||||
dummyentry = b'\1\0\0' # empty UTF8
|
||||
if self.isbs and self.slots:
|
||||
first = next(iter(self.slots.values()))
|
||||
dummyentry = self.writebootstrap(Writer(), first, pool, error).toBytes()
|
||||
|
||||
w = Writer()
|
||||
w.u16(size)
|
||||
for i in range(size):
|
||||
if i not in self.slots:
|
||||
w.writeBytes(dummyentry)
|
||||
continue
|
||||
|
||||
v = self.slots[i]
|
||||
if v is None:
|
||||
continue
|
||||
|
||||
if self.isbs:
|
||||
self.writebootstrap(w, v, pool, error)
|
||||
if len(w) >= (1<<32):
|
||||
error('Maximum BootstrapMethods length is {} bytes.'.format((1<<32)-1), v.tok)
|
||||
else:
|
||||
self.writeconst(w, v, pool, error)
|
||||
return w
|
||||
|
||||
class Pool(object):
|
||||
def __init__(self):
|
||||
self.cp = PoolSub(False)
|
||||
self.bs = PoolSub(True)
|
||||
|
||||
def sub(self, ref): return self.bs if ref.isbs else self.cp
|
||||
|
||||
def resolveIDBSRefs(self, error):
|
||||
for v in self.cp.slots.values():
|
||||
if v is not None and v.type == 'InvokeDynamic':
|
||||
v.refs[0].resolve(self, error)
|
||||
|
||||
def write(self, error):
|
||||
bsmdata = self.bs.write(self, error)
|
||||
cpdata = self.cp.write(self, error)
|
||||
return cpdata, bsmdata
|
||||
|
|
@ -0,0 +1,45 @@
|
|||
DIRECTIVE = r'\.[a-z]+'
|
||||
WORD = r'(?:[a-zA-Z_$\(<]|\[[A-Z\[])[\w$;\/\[\(\)<>*+-]*'
|
||||
FOLLOWED_BY_WHITESPACE = r'(?=\s|\Z)'
|
||||
REF = r'\[[a-z0-9_:]+\]'
|
||||
LABEL_DEF = r'L\w+:'
|
||||
|
||||
COMMENT = r';.*'
|
||||
# Match optional comment and at least one newline, followed by any number of empty/whitespace lines
|
||||
NEWLINES = r'(?:{})?\n\s*'.format(COMMENT)
|
||||
|
||||
HEX_DIGIT = r'[0-9a-fA-F]'
|
||||
ESCAPE_SEQUENCE = r'''\\(?:U00(?:10|0{hd}){hd}{{4}}|u{hd}{{4}}|x{hd}{{2}}|[btnfr'"\\0-7])'''.format(hd=HEX_DIGIT)
|
||||
# See http://stackoverflow.com/questions/430759/regex-for-managing-escaped-characters-for-items-like-string-literals/5455705# 5455705
|
||||
STRING_LITERAL = r'''
|
||||
[bB]?(?:
|
||||
"
|
||||
[^"\n\\]* # any number of unescaped characters
|
||||
(?:{es}[^"\n\\]* # escape sequence followed by 0 or more unescaped
|
||||
)*
|
||||
"
|
||||
|
|
||||
'
|
||||
[^'\n\\]* # any number of unescaped characters
|
||||
(?:{es}[^'\n\\]* # escape sequence followed by 0 or more unescaped
|
||||
)*
|
||||
'
|
||||
)'''.format(es=ESCAPE_SEQUENCE)
|
||||
# For error detection
|
||||
STRING_START = r'''[bB]?(?:"(?:[^"\\\n]|{es})*|'(?:[^'\\\n]|{es})*)'''.format(es=ESCAPE_SEQUENCE)
|
||||
|
||||
# Careful here: | is not greedy so hex must come first
|
||||
INT_LITERAL = r'[+-]?(?:0[xX]{hd}+|[1-9][0-9]*|0)[lL]?'.format(hd=HEX_DIGIT)
|
||||
FLOAT_LITERAL = r'''(?:
|
||||
(?:
|
||||
[-+][Ii][Nn][Ff][Ii][Nn][Ii][Tt][Yy]| # Nan and Inf both have mandatory sign
|
||||
[-+][Nn][Aa][Nn]
|
||||
(?:<0[xX]{hd}+>)? # Optional suffix for nonstandard NaNs
|
||||
)|
|
||||
[-+]?(?:
|
||||
\d+\.\d+(?:[eE][+-]?\d+)?| # decimal float
|
||||
\d+[eE][+-]?\d+| # decimal float with no fraction (exponent mandatory)
|
||||
0[xX]{hd}+(?:\.{hd}+)?[pP][+-]?\d+ # hexidecimal float
|
||||
)
|
||||
)[fF]?
|
||||
'''.format(hd=HEX_DIGIT)
|
||||
|
|
@ -0,0 +1,113 @@
|
|||
from __future__ import print_function
|
||||
|
||||
import collections
|
||||
import re
|
||||
import sys
|
||||
|
||||
from . import token_regexes as res
|
||||
|
||||
class AsssemblerError(Exception):
|
||||
pass
|
||||
|
||||
Token = collections.namedtuple('Token', 'type val pos')
|
||||
|
||||
TOKENS = [
|
||||
('WHITESPACE', r'[ \t]+'),
|
||||
('WORD', res.WORD + res.FOLLOWED_BY_WHITESPACE),
|
||||
('DIRECTIVE', res.DIRECTIVE + res.FOLLOWED_BY_WHITESPACE),
|
||||
('LABEL_DEF', res.LABEL_DEF + res.FOLLOWED_BY_WHITESPACE),
|
||||
('NEWLINES', res.NEWLINES),
|
||||
('REF', res.REF + res.FOLLOWED_BY_WHITESPACE),
|
||||
('COLON', r':' + res.FOLLOWED_BY_WHITESPACE),
|
||||
('EQUALS', r'=' + res.FOLLOWED_BY_WHITESPACE),
|
||||
('INT_LITERAL', res.INT_LITERAL + res.FOLLOWED_BY_WHITESPACE),
|
||||
('DOUBLE_LITERAL', res.FLOAT_LITERAL + res.FOLLOWED_BY_WHITESPACE),
|
||||
('STRING_LITERAL', res.STRING_LITERAL + res.FOLLOWED_BY_WHITESPACE),
|
||||
]
|
||||
REGEX = re.compile('|'.join('(?P<{}>{})'.format(*pair) for pair in TOKENS), re.VERBOSE)
|
||||
# For error detection
|
||||
STRING_START_REGEX = re.compile(res.STRING_START)
|
||||
WORD_LIKE_REGEX = re.compile(r'.\S*')
|
||||
|
||||
MAXLINELEN = 80
|
||||
def formatError(source, filename, message, point, point2):
|
||||
try:
|
||||
start = source.rindex('\n', 0, point) + 1
|
||||
except ValueError:
|
||||
start = 0
|
||||
line_start = start
|
||||
|
||||
try:
|
||||
end = source.index('\n', start) + 1
|
||||
except ValueError: # pragma: no cover
|
||||
end = len(source) + 1
|
||||
|
||||
# Find an 80 char section of the line around the point of interest to display
|
||||
temp = min(point2, point + MAXLINELEN//2)
|
||||
if temp < start + MAXLINELEN:
|
||||
end = min(end, start + MAXLINELEN)
|
||||
elif point >= end - MAXLINELEN:
|
||||
start = max(start, end - MAXLINELEN)
|
||||
else:
|
||||
mid = (point + temp) // 2
|
||||
start = max(start, mid - MAXLINELEN//2)
|
||||
end = min(end, start + MAXLINELEN)
|
||||
point2 = min(point2, end)
|
||||
assert line_start <= start <= point < point2 <= end
|
||||
|
||||
pchars = [' '] * (end - start)
|
||||
for i in range(point - start, point2 - start):
|
||||
pchars[i] = '~'
|
||||
pchars[point - start] = '^'
|
||||
lineno = source[:line_start].count('\n') + 1
|
||||
colno = point - line_start + 1
|
||||
return '{}:{}:{}: {}\n{}\n{}'.format(filename, lineno, colno,
|
||||
message, source[start:end].rstrip('\n'), ''.join(pchars))
|
||||
|
||||
class Tokenizer(object):
|
||||
def __init__(self, source, filename):
|
||||
self.s = source
|
||||
self.pos = 0
|
||||
self.atlineend = True
|
||||
if isinstance(filename, bytes):
|
||||
filename = filename.decode()
|
||||
self.filename = filename.rpartition('/')[-1]
|
||||
|
||||
def error(self, error, *notes):
|
||||
printerr = lambda s: print(s, file=sys.stderr)
|
||||
message, point, point2 = error
|
||||
printerr(formatError(self.s, self.filename, 'error: ' + message, point, point2))
|
||||
for message, point, point2 in notes:
|
||||
printerr(formatError(self.s, self.filename, 'note: ' + message, point, point2))
|
||||
raise AsssemblerError()
|
||||
|
||||
def _nextsub(self):
|
||||
match = REGEX.match(self.s, self.pos)
|
||||
if match is None:
|
||||
if self.atend():
|
||||
return Token('EOF', '', self.pos)
|
||||
else:
|
||||
str_match = STRING_START_REGEX.match(self.s, self.pos)
|
||||
if str_match is not None:
|
||||
self.error(('Invalid escape sequence or character in string literal', str_match.end(), str_match.end()+1))
|
||||
|
||||
match = WORD_LIKE_REGEX.match(self.s, self.pos)
|
||||
return Token('INVALID_TOKEN', match.group(), match.start())
|
||||
assert match.start() == match.pos == self.pos
|
||||
|
||||
self.pos = match.end()
|
||||
return Token(match.lastgroup, match.group(), match.start())
|
||||
|
||||
def next(self):
|
||||
tok = self._nextsub()
|
||||
while tok.type == 'WHITESPACE' or self.atlineend and tok.type == 'NEWLINES':
|
||||
tok = self._nextsub()
|
||||
self.atlineend = tok.type == 'NEWLINES'
|
||||
|
||||
if tok.type == 'INT_LITERAL' and tok.val.lower().endswith('l'):
|
||||
return tok._replace(type='LONG_LITERAL')
|
||||
elif tok.type == 'DOUBLE_LITERAL' and tok.val.lower().endswith('f'):
|
||||
return tok._replace(type='FLOAT_LITERAL')
|
||||
return tok
|
||||
|
||||
def atend(self): return self.pos == len(self.s)
|
||||
|
|
@ -0,0 +1,138 @@
|
|||
import collections
|
||||
import struct
|
||||
|
||||
Label = collections.namedtuple('Label', ['tok', 'sym'])
|
||||
|
||||
class Writer(object):
|
||||
def __init__(self):
|
||||
self.b = bytearray()
|
||||
self.refphs = []
|
||||
self.refu8phs = []
|
||||
self.lblphs = []
|
||||
|
||||
# includes lbl and manual phs but not ref phs
|
||||
self._ph8s = set()
|
||||
self._ph16s = set()
|
||||
self._ph32s = set()
|
||||
|
||||
@property
|
||||
def pos(self): return len(self.b)
|
||||
|
||||
def u8(self, x): self.b.append(x)
|
||||
def s8(self, x): self.b.append(x % 256)
|
||||
def u16(self, x): self.b.extend(struct.pack('>H', x))
|
||||
def s16(self, x): self.b.extend(struct.pack('>h', x))
|
||||
def u32(self, x): self.b.extend(struct.pack('>I', x))
|
||||
def s32(self, x): self.b.extend(struct.pack('>i', x))
|
||||
def u64(self, x): self.b.extend(struct.pack('>Q', x))
|
||||
def writeBytes(self, b): self.b.extend(b)
|
||||
|
||||
def ref(self, ref):
|
||||
self.refphs.append((self.pos, ref))
|
||||
self.u16(0)
|
||||
|
||||
def refu8(self, ref):
|
||||
self.refu8phs.append((self.pos, ref))
|
||||
self.u8(0)
|
||||
|
||||
def ph8(self):
|
||||
pos = self.pos
|
||||
self.u8(0)
|
||||
self._ph8s.add(pos)
|
||||
return pos
|
||||
|
||||
def ph16(self):
|
||||
pos = self.pos
|
||||
self.u16(0)
|
||||
self._ph16s.add(pos)
|
||||
return pos
|
||||
|
||||
def ph32(self):
|
||||
pos = self.pos
|
||||
self.u32(0)
|
||||
self._ph32s.add(pos)
|
||||
return pos
|
||||
|
||||
def lbl(self, lbl, base, dtype):
|
||||
pos = self.ph32() if dtype == 's32' else self.ph16()
|
||||
self.lblphs.append((pos, lbl, base, dtype))
|
||||
|
||||
def lblrange(self, start, end):
|
||||
self.lbl(start, 0, 'u16')
|
||||
self.lbl(end, start, 'u16')
|
||||
|
||||
def setph8(self, pos, x):
|
||||
assert self.b[pos] == 0
|
||||
self.b[pos] = x
|
||||
self._ph8s.remove(pos)
|
||||
|
||||
def setph16(self, pos, x):
|
||||
assert self.b[pos:pos+2] == b'\0\0'
|
||||
self.b[pos:pos+2] = struct.pack('>H', x)
|
||||
self._ph16s.remove(pos)
|
||||
|
||||
def setph32(self, pos, x):
|
||||
assert self.b[pos:pos+4] == b'\0\0\0\0'
|
||||
self.b[pos:pos+4] = struct.pack('>I', x)
|
||||
self._ph32s.remove(pos)
|
||||
|
||||
def _getlbl(self, lbl, labels, error):
|
||||
if lbl.sym not in labels:
|
||||
error('Undefined label', lbl.tok)
|
||||
return labels[lbl.sym][1]
|
||||
|
||||
def fillLabels(self, labels, error):
|
||||
for pos, lbl, base, dtype in self.lblphs:
|
||||
tok = lbl.tok
|
||||
lbl = self._getlbl(lbl, labels, error)
|
||||
|
||||
# base can also be a second label
|
||||
if isinstance(base, Label):
|
||||
base = self._getlbl(base, labels, error)
|
||||
|
||||
offset = lbl - base
|
||||
if dtype == 's16':
|
||||
if not -1<<15 <= offset < 1<<15:
|
||||
error('Label offset must fit in signed 16 bit int. (offset is {})'.format(offset), tok)
|
||||
self.setph16(pos, offset % (1<<16))
|
||||
elif dtype == 'u16':
|
||||
if not 0 <= offset < 1<<16:
|
||||
error('Label offset must fit in unsigned 16 bit int. (offset is {})'.format(offset), tok)
|
||||
self.setph16(pos, offset)
|
||||
elif dtype == 's32':
|
||||
if not -1<<31 <= offset < 1<<31:
|
||||
error('Label offset must fit in signed 32 bit int. (offset is {})'.format(offset), tok)
|
||||
self.setph32(pos, offset % (1<<32))
|
||||
else:
|
||||
assert 0 # pragma: no cover
|
||||
self.lblphs = []
|
||||
return self
|
||||
|
||||
def fillRefs(self, pool, error):
|
||||
for pos, ref in self.refu8phs:
|
||||
self.b[pos] = ref.resolve(pool, error)
|
||||
for pos, ref in self.refphs:
|
||||
self.b[pos:pos+2] = struct.pack('>H', ref.resolve(pool, error))
|
||||
self.refu8phs = []
|
||||
self.refphs = []
|
||||
|
||||
def toBytes(self):
|
||||
assert not self.refphs and not self.refu8phs
|
||||
assert not self._ph8s and not self._ph16s and not self._ph32s
|
||||
return bytes(self.b)
|
||||
|
||||
def __len__(self): return len(self.b)
|
||||
|
||||
def __iadd__(self, other):
|
||||
# Make sure there are no manual placeholders in other
|
||||
assert len(other.lblphs) == len(other._ph8s) + len(other._ph16s) + len(other._ph32s)
|
||||
|
||||
offset = self.pos
|
||||
self.b += other.b
|
||||
self.refphs.extend((pos + offset, ref) for pos, ref in other.refphs)
|
||||
self.refu8phs.extend((pos + offset, ref) for pos, ref in other.refu8phs)
|
||||
self.lblphs.extend((pos + offset, lbl, base, dtype) for pos, lbl, base, dtype in other.lblphs)
|
||||
self._ph8s.update(pos + offset for pos in other._ph8s)
|
||||
self._ph16s.update(pos + offset for pos in other._ph16s)
|
||||
self._ph32s.update(pos + offset for pos in other._ph32s)
|
||||
return self
|
||||
|
|
@ -0,0 +1,21 @@
|
|||
def get_attribute_raw(bytestream, ic_indices):
|
||||
name_ind, length = bytestream.get('>HL')
|
||||
|
||||
# Hotspot does not actually check the attribute length of InnerClasses prior to 49.0
|
||||
# so this case requires special handling. We will keep the purported length of the
|
||||
# attribute so that it can be displayed in the disassembly. For InnerClass attributes
|
||||
# data is actually a (length, bytes) tuple, rather than storing the bytes directly
|
||||
if name_ind in ic_indices:
|
||||
count = bytestream.get('>H', peek=True)
|
||||
data = length, bytestream.getRaw(2+8*count)
|
||||
else:
|
||||
data = bytestream.getRaw(length)
|
||||
|
||||
return name_ind,data
|
||||
|
||||
def get_attributes_raw(bytestream, ic_indices=()):
|
||||
attribute_count = bytestream.get('>H')
|
||||
return [get_attribute_raw(bytestream, ic_indices) for _ in range(attribute_count)]
|
||||
|
||||
def fixAttributeNames(attributes_raw, cpool):
|
||||
return [(cpool.getArgsCheck('Utf8', name_ind), data) for name_ind, data in attributes_raw]
|
||||
|
|
@ -0,0 +1,217 @@
|
|||
from __future__ import division
|
||||
|
||||
from . import opnames
|
||||
|
||||
def parseInstructions(bytestream, isConstructor):
|
||||
data = bytestream
|
||||
assert data.off == 0
|
||||
|
||||
instructions = {}
|
||||
while data.size() > 0:
|
||||
address = data.off
|
||||
inst = getNextInstruction(data, address)
|
||||
|
||||
# replace constructor invocations with synthetic op invokeinit to simplfy things later
|
||||
if inst[0] == opnames.INVOKESPECIAL and isConstructor(inst[1]):
|
||||
inst = (opnames.INVOKEINIT,) + inst[1:]
|
||||
|
||||
instructions[address] = inst
|
||||
assert data.size() == 0
|
||||
return instructions
|
||||
|
||||
simpleOps = {0x00:opnames.NOP, 0x01:opnames.CONSTNULL, 0x94:opnames.LCMP,
|
||||
0xbe:opnames.ARRLEN, 0xbf:opnames.THROW, 0xc2:opnames.MONENTER,
|
||||
0xc3:opnames.MONEXIT, 0x57:opnames.POP, 0x58:opnames.POP2, 0x59:opnames.DUP,
|
||||
0x5a:opnames.DUPX1, 0x5b:opnames.DUPX2, 0x5c:opnames.DUP2,
|
||||
0x5d:opnames.DUP2X1, 0x5e:opnames.DUP2X2, 0x5f:opnames.SWAP}
|
||||
|
||||
singleIndexOps = {0xb2:opnames.GETSTATIC,0xb3:opnames.PUTSTATIC,0xb4:opnames.GETFIELD,
|
||||
0xb5:opnames.PUTFIELD,0xb6:opnames.INVOKEVIRTUAL,0xb7:opnames.INVOKESPECIAL,
|
||||
0xb8:opnames.INVOKESTATIC, 0xbb:opnames.NEW,0xbd:opnames.ANEWARRAY,
|
||||
0xc0:opnames.CHECKCAST,0xc1:opnames.INSTANCEOF}
|
||||
|
||||
def getNextInstruction(data, address):
|
||||
byte = data.get('>B')
|
||||
|
||||
# typecode - B,C,S, and Bool are only used for array types and sign extension
|
||||
A,B,C,D,F,I,L,S = "ABCDFIJS"
|
||||
Bool = "Z"
|
||||
|
||||
if byte in simpleOps:
|
||||
inst = (simpleOps[byte],)
|
||||
elif byte in singleIndexOps:
|
||||
inst = (singleIndexOps[byte], data.get('>H'))
|
||||
elif byte <= 0x11:
|
||||
op = opnames.CONST
|
||||
if byte <= 0x08:
|
||||
t, val = I, byte - 0x03
|
||||
elif byte <= 0x0a:
|
||||
t, val = L, byte - 0x09
|
||||
elif byte <= 0x0d:
|
||||
t, val = F, float(byte - 0x0b)
|
||||
elif byte <= 0x0f:
|
||||
t, val = D, float(byte - 0x0e)
|
||||
elif byte == 0x10:
|
||||
t, val = I, data.get('>b')
|
||||
else:
|
||||
t, val = I, data.get('>h')
|
||||
inst = op, t, val
|
||||
elif byte == 0x12:
|
||||
inst = opnames.LDC, data.get('>B'), 1
|
||||
elif byte == 0x13:
|
||||
inst = opnames.LDC, data.get('>H'), 1
|
||||
elif byte == 0x14:
|
||||
inst = opnames.LDC, data.get('>H'), 2
|
||||
elif byte <= 0x2d:
|
||||
op = opnames.LOAD
|
||||
if byte <= 0x19:
|
||||
t = [I,L,F,D,A][byte - 0x15]
|
||||
val = data.get('>B')
|
||||
else:
|
||||
temp = byte - 0x1a
|
||||
t = [I,L,F,D,A][temp // 4]
|
||||
val = temp % 4
|
||||
inst = op, t, val
|
||||
elif byte <= 0x35:
|
||||
op = opnames.ARRLOAD
|
||||
t = [I,L,F,D,A,B,C,S][byte - 0x2e]
|
||||
inst = (op, t) if t != A else (opnames.ARRLOAD_OBJ,) # split object case into seperate op name to simplify things later
|
||||
elif byte <= 0x4e:
|
||||
op = opnames.STORE
|
||||
if byte <= 0x3a:
|
||||
t = [I,L,F,D,A][byte - 0x36]
|
||||
val = data.get('>B')
|
||||
else:
|
||||
temp = byte - 0x3b
|
||||
t = [I,L,F,D,A][temp // 4]
|
||||
val = temp % 4
|
||||
inst = op, t, val
|
||||
elif byte <= 0x56:
|
||||
op = opnames.ARRSTORE
|
||||
t = [I,L,F,D,A,B,C,S][byte - 0x4f]
|
||||
inst = (op, t) if t != A else (opnames.ARRSTORE_OBJ,) # split object case into seperate op name to simplify things later
|
||||
elif byte <= 0x77:
|
||||
temp = byte - 0x60
|
||||
opt = (opnames.ADD,opnames.SUB,opnames.MUL,opnames.DIV,opnames.REM,opnames.NEG)[temp//4]
|
||||
t = (I,L,F,D)[temp % 4]
|
||||
inst = opt, t
|
||||
elif byte <= 0x83:
|
||||
temp = byte - 0x78
|
||||
opt = (opnames.SHL,opnames.SHR,opnames.USHR,opnames.AND,opnames.OR,opnames.XOR)[temp//2]
|
||||
t = (I,L)[temp % 2]
|
||||
inst = opt, t
|
||||
elif byte == 0x84:
|
||||
inst = opnames.IINC, data.get('>B'), data.get('>b')
|
||||
elif byte <= 0x90:
|
||||
op = opnames.CONVERT
|
||||
pairs = ((I,L),(I,F),(I,D),(L,I),(L,F),(L,D),(F,I),(F,L),(F,D),
|
||||
(D,I),(D,L),(D,F))
|
||||
src_t, dest_t = pairs[byte - 0x85]
|
||||
inst = op, src_t, dest_t
|
||||
elif byte <= 0x93:
|
||||
op = opnames.TRUNCATE
|
||||
dest_t = [B,C,S][byte - 0x91]
|
||||
inst = op, dest_t
|
||||
elif byte <= 0x98:
|
||||
op = opnames.FCMP
|
||||
temp = byte - 0x95
|
||||
t = (F,D)[temp//2]
|
||||
NaN_val = (-1,1)[temp % 2]
|
||||
inst = op, t, NaN_val
|
||||
elif byte <= 0x9e:
|
||||
op = opnames.IF_I
|
||||
cmp_t = ('eq','ne','lt','ge','gt','le')[byte - 0x99]
|
||||
jumptarget = data.get('>h') + address
|
||||
inst = op, cmp_t, jumptarget
|
||||
elif byte <= 0xa4:
|
||||
op = opnames.IF_ICMP
|
||||
cmp_t = ('eq','ne','lt','ge','gt','le')[byte - 0x9f]
|
||||
jumptarget = data.get('>h') + address
|
||||
inst = op, cmp_t, jumptarget
|
||||
elif byte <= 0xa6:
|
||||
op = opnames.IF_ACMP
|
||||
cmp_t = ('eq','ne')[byte - 0xa5]
|
||||
jumptarget = data.get('>h') + address
|
||||
inst = op, cmp_t, jumptarget
|
||||
elif byte == 0xa7:
|
||||
inst = opnames.GOTO, data.get('>h') + address
|
||||
elif byte == 0xa8:
|
||||
inst = opnames.JSR, data.get('>h') + address
|
||||
elif byte == 0xa9:
|
||||
inst = opnames.RET, data.get('>B')
|
||||
elif byte == 0xaa: # Table Switch
|
||||
padding = data.getRaw((3-address) % 4)
|
||||
default = data.get('>i') + address
|
||||
low = data.get('>i')
|
||||
high = data.get('>i')
|
||||
assert high >= low
|
||||
numpairs = high - low + 1
|
||||
offsets = [data.get('>i') + address for _ in range(numpairs)]
|
||||
jumps = zip(range(low, high+1), offsets)
|
||||
inst = opnames.SWITCH, default, jumps
|
||||
elif byte == 0xab: # Lookup Switch
|
||||
padding = data.getRaw((3-address) % 4)
|
||||
default = data.get('>i') + address
|
||||
numpairs = data.get('>i')
|
||||
assert numpairs >= 0
|
||||
pairs = [data.get('>ii') for _ in range(numpairs)]
|
||||
jumps = [(x,(y + address)) for x,y in pairs]
|
||||
inst = opnames.SWITCH, default, jumps
|
||||
elif byte <= 0xb1:
|
||||
op = opnames.RETURN
|
||||
t = (I,L,F,D,A,None)[byte - 0xac]
|
||||
inst = op, t
|
||||
elif byte == 0xb9:
|
||||
op = opnames.INVOKEINTERFACE
|
||||
index = data.get('>H')
|
||||
count, zero = data.get('>B'), data.get('>B')
|
||||
inst = op, index, count, zero
|
||||
elif byte == 0xba:
|
||||
op = opnames.INVOKEDYNAMIC
|
||||
index = data.get('>H')
|
||||
zero = data.get('>H')
|
||||
inst = op, index, zero
|
||||
elif byte == 0xbc:
|
||||
typecode = data.get('>b')
|
||||
types = {4:Bool, 5:C, 6:F, 7:D, 8:B, 9:S, 10:I, 11:L}
|
||||
t = types.get(typecode)
|
||||
inst = opnames.NEWARRAY, t
|
||||
elif byte == 0xc4: # wide
|
||||
realbyte = data.get('>B')
|
||||
if realbyte >= 0x15 and realbyte < 0x1a:
|
||||
t = [I,L,F,D,A][realbyte - 0x15]
|
||||
inst = opnames.LOAD, t, data.get('>H')
|
||||
elif realbyte >= 0x36 and realbyte < 0x3b:
|
||||
t = [I,L,F,D,A][realbyte - 0x36]
|
||||
inst = opnames.STORE, t, data.get('>H')
|
||||
elif realbyte == 0xa9:
|
||||
inst = opnames.RET, data.get('>H')
|
||||
elif realbyte == 0x84:
|
||||
inst = opnames.IINC, data.get('>H'), data.get('>h')
|
||||
else:
|
||||
assert 0
|
||||
elif byte == 0xc5:
|
||||
op = opnames.MULTINEWARRAY
|
||||
index = data.get('>H')
|
||||
dim = data.get('>B')
|
||||
inst = op, index, dim
|
||||
elif byte <= 0xc7:
|
||||
op = opnames.IF_A
|
||||
cmp_t = ('eq','ne')[byte - 0xc6]
|
||||
jumptarget = data.get('>h') + address
|
||||
inst = op, cmp_t, jumptarget
|
||||
elif byte == 0xc8:
|
||||
inst = opnames.GOTO, data.get('>i') + address
|
||||
elif byte == 0xc9:
|
||||
inst = opnames.JSR, data.get('>i') + address
|
||||
else:
|
||||
assert 0
|
||||
return inst
|
||||
|
||||
def printInstruction(instr):
|
||||
if len(instr) == 1:
|
||||
return instr[0]
|
||||
elif len(instr) == 2:
|
||||
return '{}({})'.format(*instr)
|
||||
else:
|
||||
return '{}{}'.format(instr[0], instr[1:])
|
||||
|
|
@ -0,0 +1,107 @@
|
|||
from . import constant_pool, field, method
|
||||
from .attributes_raw import fixAttributeNames, get_attributes_raw
|
||||
|
||||
cp_structFmts = {3: '>i',
|
||||
4: '>i', # floats and doubles internally represented as integers with same bit pattern
|
||||
5: '>q',
|
||||
6: '>q',
|
||||
7: '>H',
|
||||
8: '>H',
|
||||
9: '>HH',
|
||||
10: '>HH',
|
||||
11: '>HH',
|
||||
12: '>HH',
|
||||
15: '>BH',
|
||||
16: '>H',
|
||||
18: '>HH'}
|
||||
|
||||
def get_cp_raw(bytestream):
|
||||
const_count = bytestream.get('>H')
|
||||
assert const_count > 1
|
||||
|
||||
placeholder = None,None
|
||||
pool = [placeholder]
|
||||
|
||||
while len(pool) < const_count:
|
||||
tag = bytestream.get('B')
|
||||
if tag == 1: # utf8
|
||||
length = bytestream.get('>H')
|
||||
data = bytestream.getRaw(length)
|
||||
val = tag, (data,)
|
||||
else:
|
||||
val = tag,bytestream.get(cp_structFmts[tag], True)
|
||||
pool.append(val)
|
||||
# Longs and Doubles take up two spaces in the pool
|
||||
if tag == 5 or tag == 6:
|
||||
pool.append(placeholder)
|
||||
assert len(pool) == const_count
|
||||
return pool
|
||||
|
||||
def get_field_raw(bytestream):
|
||||
flags, name, desc = bytestream.get('>HHH')
|
||||
attributes = get_attributes_raw(bytestream)
|
||||
return flags, name, desc, attributes
|
||||
|
||||
def get_fields_raw(bytestream):
|
||||
count = bytestream.get('>H')
|
||||
return [get_field_raw(bytestream) for _ in range(count)]
|
||||
|
||||
# fields and methods have same raw format
|
||||
get_method_raw = get_field_raw
|
||||
get_methods_raw = get_fields_raw
|
||||
|
||||
class ClassFile(object):
|
||||
flagVals = {'PUBLIC':0x0001,
|
||||
'FINAL':0x0010,
|
||||
'SUPER':0x0020,
|
||||
'INTERFACE':0x0200,
|
||||
'ABSTRACT':0x0400,
|
||||
'SYNTHETIC':0x1000,
|
||||
'ANNOTATION':0x2000,
|
||||
'ENUM':0x4000,
|
||||
|
||||
# These flags are only used for InnerClasses attributes
|
||||
'PRIVATE':0x0002,
|
||||
'PROTECTED':0x0004,
|
||||
'STATIC':0x0008,
|
||||
}
|
||||
|
||||
def __init__(self, bytestream):
|
||||
magic, minor, major = bytestream.get('>LHH')
|
||||
assert magic == 0xCAFEBABE
|
||||
self.version = major,minor
|
||||
|
||||
const_pool_raw = get_cp_raw(bytestream)
|
||||
flags, self.this, self.super = bytestream.get('>HHH')
|
||||
|
||||
interface_count = bytestream.get('>H')
|
||||
self.interfaces_raw = [bytestream.get('>H') for _ in range(interface_count)]
|
||||
|
||||
self.fields_raw = get_fields_raw(bytestream)
|
||||
self.methods_raw = get_methods_raw(bytestream)
|
||||
|
||||
ic_indices = [i for i,x in enumerate(const_pool_raw) if x == (1, ("InnerClasses",))]
|
||||
self.attributes_raw = get_attributes_raw(bytestream, ic_indices)
|
||||
assert bytestream.size() == 0
|
||||
|
||||
self.flags = frozenset(name for name,mask in ClassFile.flagVals.items() if (mask & flags))
|
||||
self.cpool = constant_pool.ConstPool(const_pool_raw)
|
||||
self.name = self.cpool.getArgsCheck('Class', self.this)
|
||||
self.elementsLoaded = False
|
||||
|
||||
self.env = self.supername = None
|
||||
self.fields = self.methods = self.attributes = None
|
||||
if self.super:
|
||||
self.supername = self.cpool.getArgsCheck('Class', self.super)
|
||||
|
||||
def loadElements(self, keepRaw=False):
|
||||
if self.elementsLoaded:
|
||||
return
|
||||
self.fields = [field.Field(m, self, keepRaw) for m in self.fields_raw]
|
||||
self.methods = [method.Method(m, self, keepRaw) for m in self.methods_raw]
|
||||
self.attributes = fixAttributeNames(self.attributes_raw, self.cpool)
|
||||
|
||||
self.fields_raw = self.methods_raw = None
|
||||
if not keepRaw:
|
||||
self.attributes_raw = None
|
||||
self.elementsLoaded = True
|
||||
|
|
@ -0,0 +1,116 @@
|
|||
import collections
|
||||
|
||||
from .reader import Reader
|
||||
|
||||
TAGS = [None, 'Utf8', None, 'Int', 'Float', 'Long', 'Double', 'Class', 'String', 'Field', 'Method', 'InterfaceMethod', 'NameAndType', None, None, 'MethodHandle', 'MethodType', None, 'InvokeDynamic', 'Module', 'Package']
|
||||
|
||||
SlotData = collections.namedtuple('SlotData', ['tag', 'data', 'refs'])
|
||||
ExceptData = collections.namedtuple('ExceptData', ['start', 'end', 'handler', 'type'])
|
||||
|
||||
class ConstantPoolData(object):
|
||||
def __init__(self, r):
|
||||
self.slots = []
|
||||
self._null()
|
||||
|
||||
size = r.u16()
|
||||
while len(self.slots) < size:
|
||||
self._const(r)
|
||||
|
||||
def _null(self):
|
||||
self.slots.append(SlotData(None, None, None))
|
||||
|
||||
def _const(self, r):
|
||||
t = TAGS[r.u8()]
|
||||
data = None
|
||||
refs = []
|
||||
|
||||
if t == 'Utf8':
|
||||
data = r.getRaw(r.u16())
|
||||
elif t == 'Int' or t == 'Float':
|
||||
data = r.u32()
|
||||
elif t == 'Long' or t == 'Double':
|
||||
data = r.u64()
|
||||
elif t == 'MethodHandle':
|
||||
data = r.u8()
|
||||
refs.append(r.u16())
|
||||
elif t in ['Class', 'String', 'MethodType', 'Module', 'Package']:
|
||||
refs.append(r.u16())
|
||||
else:
|
||||
refs.append(r.u16())
|
||||
refs.append(r.u16())
|
||||
self.slots.append(SlotData(t, data, refs))
|
||||
if t in ('Long', 'Double'):
|
||||
self._null()
|
||||
|
||||
def getutf(self, ind):
|
||||
if ind < len(self.slots) and self.slots[ind].tag == 'Utf8':
|
||||
return self.slots[ind].data
|
||||
|
||||
def getclsutf(self, ind):
|
||||
if ind < len(self.slots) and self.slots[ind].tag == 'Class':
|
||||
return self.getutf(self.slots[ind].refs[0])
|
||||
|
||||
class BootstrapMethodsData(object):
|
||||
def __init__(self, r):
|
||||
self.slots = []
|
||||
for _ in range(r.u16()):
|
||||
first = r.u16()
|
||||
argcount = r.u16()
|
||||
refs = [first] + [r.u16() for _ in range(argcount)]
|
||||
self.slots.append(SlotData('Bootstrap', None, refs))
|
||||
|
||||
class CodeData(object):
|
||||
def __init__(self, r, pool, short):
|
||||
if short:
|
||||
self.stack, self.locals, codelen = r.u8(), r.u8(), r.u16()
|
||||
else:
|
||||
self.stack, self.locals, codelen = r.u16(), r.u16(), r.u32()
|
||||
|
||||
self.bytecode = r.getRaw(codelen)
|
||||
self.exceptions = [ExceptData(r.u16(), r.u16(), r.u16(), r.u16()) for _ in range(r.u16())]
|
||||
self.attributes = [AttributeData(r) for _ in range(r.u16())]
|
||||
|
||||
class AttributeData(object):
|
||||
def __init__(self, r, pool=None):
|
||||
self.name, self.length = r.u16(), r.u32()
|
||||
|
||||
# The JVM allows InnerClasses attributes to have a bogus length field,
|
||||
# and hence we must calculate the length from the contents
|
||||
if pool and pool.getutf(self.name) == b'InnerClasses':
|
||||
actual_length = r.copy().u16() * 8 + 2
|
||||
else:
|
||||
actual_length = self.length
|
||||
|
||||
self.raw = r.getRaw(actual_length)
|
||||
self.wronglength = actual_length != self.length
|
||||
|
||||
def stream(self): return Reader(self.raw)
|
||||
|
||||
class FieldData(object):
|
||||
def __init__(self, r):
|
||||
self.access, self.name, self.desc = r.u16(), r.u16(), r.u16()
|
||||
self.attributes = [AttributeData(r) for _ in range(r.u16())]
|
||||
|
||||
class MethodData(object):
|
||||
def __init__(self, r):
|
||||
self.access, self.name, self.desc = r.u16(), r.u16(), r.u16()
|
||||
self.attributes = [AttributeData(r) for _ in range(r.u16())]
|
||||
|
||||
class ClassData(object):
|
||||
def __init__(self, r):
|
||||
magic, minor, major = r.u32(), r.u16(), r.u16()
|
||||
self.version = major, minor
|
||||
|
||||
self.pool = ConstantPoolData(r)
|
||||
|
||||
self.access, self.this, self.super = r.u16(), r.u16(), r.u16()
|
||||
self.interfaces = [r.u16() for _ in range(r.u16())]
|
||||
self.fields = [FieldData(r) for _ in range(r.u16())]
|
||||
self.methods = [MethodData(r) for _ in range(r.u16())]
|
||||
self.attributes = [AttributeData(r, pool=self.pool) for _ in range(r.u16())]
|
||||
# assert r.done()
|
||||
|
||||
def getattrs(self, name):
|
||||
for attr in self.attributes:
|
||||
if self.pool.getutf(attr.name) == name:
|
||||
yield attr
|
||||
|
|
@ -0,0 +1,28 @@
|
|||
import re
|
||||
|
||||
# First alternative handles a single surrogate, in case input string somehow contains unmerged surrogates
|
||||
NONASTRAL_REGEX = re.compile(u'[\ud800-\udfff]|[\0-\ud7ff\ue000-\uffff]+')
|
||||
|
||||
def encode(s):
|
||||
assert not isinstance(s, bytes)
|
||||
b = b''
|
||||
pos = 0
|
||||
while pos < len(s):
|
||||
x = ord(s[pos])
|
||||
if x >= 1<<16:
|
||||
x -= 1<<16
|
||||
high = 0xD800 + (x >> 10)
|
||||
low = 0xDC00 + (x % (1 << 10))
|
||||
b += unichr(high).encode('utf8')
|
||||
b += unichr(low).encode('utf8')
|
||||
pos += 1
|
||||
else:
|
||||
m = NONASTRAL_REGEX.match(s, pos)
|
||||
b += m.group().encode('utf8')
|
||||
pos = m.end()
|
||||
return b.replace(b'\0', b'\xc0\x80')
|
||||
|
||||
# Warning, decode(encode(s)) != s if s contains astral characters, as they are converted to surrogate pairs
|
||||
def decode(b):
|
||||
assert isinstance(b, bytes)
|
||||
return b.replace(b'\xc0\x80', b'\0').decode('utf8')
|
||||
|
|
@ -0,0 +1,47 @@
|
|||
import struct
|
||||
|
||||
|
||||
class TruncatedStreamError(EOFError):
|
||||
pass
|
||||
|
||||
|
||||
class Reader(object):
|
||||
__slots__ = ['d', 'off']
|
||||
def __init__(self, data, off=0):
|
||||
self.d = data
|
||||
self.off = off
|
||||
|
||||
def done(self): return self.off >= len(self.d)
|
||||
def copy(self): return Reader(self.d, self.off)
|
||||
|
||||
def u8(self): return self.get('>B')
|
||||
def s8(self): return self.get('>b')
|
||||
def u16(self): return self.get('>H')
|
||||
def s16(self): return self.get('>h')
|
||||
def u32(self): return self.get('>I')
|
||||
def s32(self): return self.get('>i')
|
||||
def u64(self): return self.get('>Q')
|
||||
|
||||
# binUnpacker functions
|
||||
def get(self, fmt, forceTuple=False, peek=False):
|
||||
size = struct.calcsize(fmt)
|
||||
if self.size() < size:
|
||||
raise TruncatedStreamError()
|
||||
|
||||
val = struct.unpack_from(fmt, self.d, self.off)
|
||||
|
||||
if not peek:
|
||||
self.off += size
|
||||
if not forceTuple and len(val) == 1:
|
||||
val = val[0]
|
||||
return val
|
||||
|
||||
def getRaw(self, num):
|
||||
if self.size() < num:
|
||||
raise TruncatedStreamError()
|
||||
val = self.d[self.off:self.off+num]
|
||||
self.off += num
|
||||
return val
|
||||
|
||||
def size(self):
|
||||
return len(self.d) - self.off
|
||||
|
|
@ -0,0 +1,103 @@
|
|||
import collections
|
||||
import struct
|
||||
|
||||
# ConstantPool stores strings as strings or unicodes. They are automatically
|
||||
# converted to and from modified Utf16 when reading and writing to binary
|
||||
|
||||
# Floats and Doubles are internally stored as integers with the same bit pattern
|
||||
# Since using raw floats breaks equality testing for signed zeroes and NaNs
|
||||
# cpool.getArgs/getArgsCheck will automatically convert them into Python floats
|
||||
|
||||
def decodeStr(s):
|
||||
return s.replace('\xc0\x80','\0').decode('utf8'),
|
||||
def decodeFloat(i):
|
||||
return struct.unpack('>f', struct.pack('>i', i)) # Note: returns tuple
|
||||
def decodeDouble(i):
|
||||
return struct.unpack('>d', struct.pack('>q', i))
|
||||
|
||||
cpoolInfo_t = collections.namedtuple('cpoolInfo_t',
|
||||
['name','tag','recoverArgs'])
|
||||
|
||||
Utf8 = cpoolInfo_t('Utf8',1,
|
||||
(lambda self,s:(s,)))
|
||||
|
||||
Class = cpoolInfo_t('Class',7,
|
||||
(lambda self,n_id:self.getArgs(n_id)))
|
||||
|
||||
NameAndType = cpoolInfo_t('NameAndType',12,
|
||||
(lambda self,n,d:self.getArgs(n) + self.getArgs(d)))
|
||||
|
||||
Field = cpoolInfo_t('Field',9,
|
||||
(lambda self,c_id,nat_id:self.getArgs(c_id) + self.getArgs(nat_id)))
|
||||
|
||||
Method = cpoolInfo_t('Method',10,
|
||||
(lambda self,c_id,nat_id:self.getArgs(c_id) + self.getArgs(nat_id)))
|
||||
|
||||
InterfaceMethod = cpoolInfo_t('InterfaceMethod',11,
|
||||
(lambda self,c_id,nat_id:self.getArgs(c_id) + self.getArgs(nat_id)))
|
||||
|
||||
String = cpoolInfo_t('String',8,
|
||||
(lambda self,n_id:self.getArgs(n_id)))
|
||||
|
||||
Int = cpoolInfo_t('Int',3,
|
||||
(lambda self,s:(s,)))
|
||||
|
||||
Long = cpoolInfo_t('Long',5,
|
||||
(lambda self,s:(s,)))
|
||||
|
||||
Float = cpoolInfo_t('Float',4,
|
||||
(lambda self,s:decodeFloat(s)))
|
||||
|
||||
Double = cpoolInfo_t('Double',6,
|
||||
(lambda self,s:decodeDouble(s)))
|
||||
|
||||
MethodHandle = cpoolInfo_t('MethodHandle',15,
|
||||
(lambda self,t,n_id:(t,)+self.getArgs(n_id)))
|
||||
|
||||
MethodType = cpoolInfo_t('MethodType',16,
|
||||
(lambda self,n_id:self.getArgs(n_id)))
|
||||
|
||||
InvokeDynamic = cpoolInfo_t('InvokeDynamic',18,
|
||||
(lambda self,bs_id,nat_id:(bs_id,) + self.getArgs(nat_id)))
|
||||
|
||||
cpoolTypes = [Utf8, Class, NameAndType, Field, Method, InterfaceMethod,
|
||||
String, Int, Long, Float, Double,
|
||||
MethodHandle, MethodType, InvokeDynamic]
|
||||
name2Type = {t.name:t for t in cpoolTypes}
|
||||
tag2Type = {t.tag:t for t in cpoolTypes}
|
||||
|
||||
class ConstPool(object):
|
||||
def __init__(self, initialData=((None,None),)):
|
||||
self.pool = []
|
||||
self.reserved = set()
|
||||
self.available = set()
|
||||
|
||||
for tag, val in initialData:
|
||||
if tag is None:
|
||||
self.addEmptySlot()
|
||||
else:
|
||||
t = tag2Type[tag]
|
||||
if t.name == 'Utf8':
|
||||
val = decodeStr(*val)
|
||||
self.pool.append((t.name, val))
|
||||
|
||||
def addEmptySlot(self):
|
||||
self.pool.append((None, None))
|
||||
|
||||
def getArgs(self, i):
|
||||
if not (i >= 0 and i<len(self.pool)):
|
||||
raise IndexError('Constant pool index {} out of range'.format(i))
|
||||
if self.pool[i][0] is None:
|
||||
raise IndexError('Constant pool index {} invalid'.format(i))
|
||||
|
||||
name, val = self.pool[i]
|
||||
t = name2Type[name]
|
||||
return t.recoverArgs(self, *val)
|
||||
|
||||
def getArgsCheck(self, typen, index):
|
||||
# if (self.pool[index][0] != typen):
|
||||
# raise KeyError('Constant pool index {} has incorrect type {}'.format(index, typen))
|
||||
val = self.getArgs(index)
|
||||
return val if len(val) > 1 else val[0]
|
||||
|
||||
def getType(self, index): return self.pool[index][0]
|
||||
|
|
@ -0,0 +1,110 @@
|
|||
import os.path
|
||||
import zipfile
|
||||
|
||||
from .classfile import ClassFile
|
||||
from .classfileformat.reader import Reader
|
||||
from .error import ClassLoaderError
|
||||
|
||||
class Environment(object):
|
||||
def __init__(self):
|
||||
self.classes = {}
|
||||
self.path = []
|
||||
self._open = {}
|
||||
|
||||
def addToPath(self, path):
|
||||
self.path.append(path)
|
||||
|
||||
def _getSuper(self, name):
|
||||
return self.getClass(name).supername
|
||||
|
||||
def getClass(self, name, partial=False):
|
||||
try:
|
||||
result = self.classes[name]
|
||||
except KeyError:
|
||||
result = self._loadClass(name)
|
||||
if not partial:
|
||||
result.loadElements()
|
||||
return result
|
||||
|
||||
def isSubclass(self, name1, name2):
|
||||
if name2 == 'java/lang/Object':
|
||||
return True
|
||||
|
||||
while name1 != 'java/lang/Object':
|
||||
if name1 == name2:
|
||||
return True
|
||||
name1 = self._getSuper(name1)
|
||||
return False
|
||||
|
||||
def commonSuperclass(self, name1, name2):
|
||||
a, b = name1, name2
|
||||
supers = {a}
|
||||
while a != b and a != 'java/lang/Object':
|
||||
a = self._getSuper(a)
|
||||
supers.add(a)
|
||||
|
||||
while b not in supers:
|
||||
b = self._getSuper(b)
|
||||
return b
|
||||
|
||||
def isInterface(self, name, forceCheck=False):
|
||||
try:
|
||||
class_ = self.getClass(name, partial=True)
|
||||
return 'INTERFACE' in class_.flags
|
||||
except ClassLoaderError as e:
|
||||
if forceCheck:
|
||||
raise e
|
||||
# If class is not found, assume worst case, that it is a interface
|
||||
return True
|
||||
|
||||
def isFinal(self, name):
|
||||
try:
|
||||
class_ = self.getClass(name, partial=True)
|
||||
return 'FINAL' in class_.flags
|
||||
except ClassLoaderError as e:
|
||||
return False
|
||||
|
||||
def _searchForFile(self, name):
|
||||
name += '.class'
|
||||
for place in self.path:
|
||||
try:
|
||||
archive = self._open[place]
|
||||
except KeyError: # plain folder
|
||||
try:
|
||||
path = os.path.join(place, name)
|
||||
with open(path, 'rb') as file_:
|
||||
return file_.read()
|
||||
except IOError:
|
||||
print 'failed to open', path.encode('utf8')
|
||||
else: # zip archive
|
||||
try:
|
||||
return archive.read(name)
|
||||
except KeyError:
|
||||
pass
|
||||
|
||||
def _loadClass(self, name):
|
||||
print "Loading", name[:70]
|
||||
data = self._searchForFile(name)
|
||||
|
||||
if data is None:
|
||||
raise ClassLoaderError('ClassNotFoundException', name)
|
||||
|
||||
stream = Reader(data=data)
|
||||
new = ClassFile(stream)
|
||||
new.env = self
|
||||
self.classes[new.name] = new
|
||||
return new
|
||||
|
||||
# Context Manager methods to manager our zipfiles
|
||||
def __enter__(self):
|
||||
assert not self._open
|
||||
for place in self.path:
|
||||
if place.endswith('.jar') or place.endswith('.zip'):
|
||||
self._open[place] = zipfile.ZipFile(place, 'r').__enter__()
|
||||
return self
|
||||
|
||||
def __exit__(self, type_, value, traceback):
|
||||
for place in reversed(self.path):
|
||||
if place in self._open:
|
||||
self._open[place].__exit__(type_, value, traceback)
|
||||
del self._open[place]
|
||||
|
|
@ -0,0 +1,12 @@
|
|||
class ClassLoaderError(Exception):
|
||||
def __init__(self, typen=None, data=""):
|
||||
self.type = typen
|
||||
self.data = data
|
||||
|
||||
message = "\n{}: {}".format(typen, data) if typen else data
|
||||
super(ClassLoaderError, self).__init__(message)
|
||||
|
||||
class VerificationError(Exception):
|
||||
def __init__(self, message, data=None):
|
||||
super(VerificationError, self).__init__(message)
|
||||
self.data = data
|
||||
|
|
@ -0,0 +1,29 @@
|
|||
from .attributes_raw import fixAttributeNames
|
||||
|
||||
class Field(object):
|
||||
flagVals = {'PUBLIC':0x0001,
|
||||
'PRIVATE':0x0002,
|
||||
'PROTECTED':0x0004,
|
||||
'STATIC':0x0008,
|
||||
'FINAL':0x0010,
|
||||
'VOLATILE':0x0040,
|
||||
'TRANSIENT':0x0080,
|
||||
'SYNTHETIC':0x1000,
|
||||
'ENUM':0x4000,
|
||||
}
|
||||
|
||||
def __init__(self, data, classFile, keepRaw):
|
||||
self.class_ = classFile
|
||||
cpool = self.class_.cpool
|
||||
|
||||
flags, name_id, desc_id, attributes_raw = data
|
||||
|
||||
self.name = cpool.getArgsCheck('Utf8', name_id)
|
||||
self.descriptor = cpool.getArgsCheck('Utf8', desc_id)
|
||||
self.attributes = fixAttributeNames(attributes_raw, cpool)
|
||||
|
||||
self.flags = set(name for name,mask in Field.flagVals.items() if (mask & flags))
|
||||
self.static = 'STATIC' in self.flags
|
||||
if keepRaw:
|
||||
self.attributes_raw = attributes_raw
|
||||
self.name_id, self.desc_id = name_id, desc_id
|
||||
|
|
@ -0,0 +1,81 @@
|
|||
import math
|
||||
|
||||
INF_MAG = 1, None
|
||||
ZERO_MAG = 0, None
|
||||
|
||||
# Numbers are represented as (sign, (mantissa, exponent))
|
||||
# For finite nonzero values, the float value is sign * mantissa * 2 ^ (exponent - mbits - 1)
|
||||
# Mantissa is normalized to always be within (2 ^ mbits) <= m < (2 ^ mbits + 1) even for subnormal numbers
|
||||
NAN = None,(None,None)
|
||||
INF = 1,INF_MAG
|
||||
NINF = -1,INF_MAG
|
||||
ZERO = 1,ZERO_MAG
|
||||
NZERO = -1,ZERO_MAG
|
||||
|
||||
# Key suitable for sorting finite (normalized) nonzero values
|
||||
sortkey = lambda (s,(m,e)):(s,s*e,s*m)
|
||||
|
||||
# Size info for type - mantissa bits, min exponent, max exponent
|
||||
FLOAT_SIZE = 23,-126,127
|
||||
DOUBLE_SIZE = 52,-1022,1023
|
||||
|
||||
def flog(x):
|
||||
'''returns f such that 2**f <= x < 2**(f+1)'''
|
||||
assert x > 0
|
||||
return len(bin(x))-3
|
||||
|
||||
def roundMag(size, mag):
|
||||
'''Round (unnormalized) magnitude to nearest representable magnitude with ties going to 0 lsb'''
|
||||
mbits, emin, emax = size
|
||||
m, e = mag
|
||||
assert m >= 1
|
||||
f = flog(m)
|
||||
|
||||
if e+f < emin: # subnormal
|
||||
dnmin = emin - mbits
|
||||
if e+f < (dnmin - 1):
|
||||
return ZERO_MAG
|
||||
if e > dnmin:
|
||||
m = m << (e - dnmin)
|
||||
f += (e - dnmin)
|
||||
e = dnmin
|
||||
s = dnmin - e
|
||||
i = m >> s
|
||||
r = (m - (i << s)) * 2
|
||||
h = 1 << s
|
||||
if r > h or r == h and (i&1):
|
||||
i += 1
|
||||
return i, e+s-mbits-1
|
||||
else:
|
||||
if f < mbits:
|
||||
m = m << (mbits - f)
|
||||
f = mbits
|
||||
s = f - mbits
|
||||
if (e+f) > emax:
|
||||
return INF_MAG
|
||||
i = m >> s
|
||||
r = (m - (i << s)) * 2
|
||||
h = 1 << s
|
||||
if r > h or r == h and (i&1):
|
||||
i += 1
|
||||
if i == (1<<mbits):
|
||||
i = i >> 1
|
||||
e += 1
|
||||
if e > emax:
|
||||
return INF_MAG
|
||||
return i, e+s-mbits-1
|
||||
|
||||
def fromRawFloat(size, x):
|
||||
if math.isnan(x):
|
||||
return NAN
|
||||
sign = int(math.copysign(1, x))
|
||||
x = math.copysign(x, 1)
|
||||
|
||||
if math.isinf(x):
|
||||
return sign, INF_MAG
|
||||
elif x == 0.0:
|
||||
return sign, ZERO_MAG
|
||||
else:
|
||||
m, e = math.frexp(x)
|
||||
m = int(m * (1<<(size[0]+1)))
|
||||
return sign, roundMag(size, (m, e))
|
||||
|
|
@ -0,0 +1,68 @@
|
|||
import itertools
|
||||
|
||||
def tarjanSCC(roots, getChildren):
|
||||
"""Return a list of strongly connected components in a graph. If getParents is passed instead of getChildren, the result will be topologically sorted.
|
||||
|
||||
roots - list of root nodes to search from
|
||||
getChildren - function which returns children of a given node
|
||||
"""
|
||||
|
||||
sccs = []
|
||||
indexCounter = itertools.count()
|
||||
index = {}
|
||||
lowlink = {}
|
||||
removed = set()
|
||||
subtree = []
|
||||
|
||||
# Use iterative version to avoid stack limits for large datasets
|
||||
stack = [(node, 0) for node in roots]
|
||||
while stack:
|
||||
current, state = stack.pop()
|
||||
if state == 0: # before recursing
|
||||
if current not in index: # if it's in index, it was already visited (possibly earlier on the current search stack)
|
||||
lowlink[current] = index[current] = next(indexCounter)
|
||||
subtree.append(current)
|
||||
|
||||
stack.append((current, 1))
|
||||
stack.extend((child, 0) for child in getChildren(current) if child not in removed)
|
||||
else: # after recursing
|
||||
children = [child for child in getChildren(current) if child not in removed]
|
||||
for child in children:
|
||||
if index[child] <= index[current]: # backedge (or selfedge)
|
||||
lowlink[current] = min(lowlink[current], index[child])
|
||||
else:
|
||||
lowlink[current] = min(lowlink[current], lowlink[child])
|
||||
assert lowlink[current] <= index[current]
|
||||
|
||||
if index[current] == lowlink[current]:
|
||||
scc = []
|
||||
while not scc or scc[-1] != current:
|
||||
scc.append(subtree.pop())
|
||||
|
||||
sccs.append(tuple(scc))
|
||||
removed.update(scc)
|
||||
return sccs
|
||||
|
||||
def topologicalSort(roots, getParents):
|
||||
"""Return a topological sorting of nodes in a graph.
|
||||
|
||||
roots - list of root nodes to search from
|
||||
getParents - function which returns the parents of a given node
|
||||
"""
|
||||
|
||||
results = []
|
||||
visited = set()
|
||||
|
||||
# Use iterative version to avoid stack limits for large datasets
|
||||
stack = [(node,0) for node in roots]
|
||||
while stack:
|
||||
current, state = stack.pop()
|
||||
if state == 0: # before recursing
|
||||
if current not in visited:
|
||||
visited.add(current)
|
||||
stack.append((current,1))
|
||||
stack.extend((parent,0) for parent in getParents(current))
|
||||
else: # after recursing
|
||||
assert current in visited
|
||||
results.append(current)
|
||||
return results
|
||||
|
|
@ -0,0 +1,734 @@
|
|||
import itertools
|
||||
import math
|
||||
|
||||
from ..ssa import objtypes
|
||||
|
||||
from . import visitor
|
||||
from .stringescape import escapeString
|
||||
|
||||
# Explicitly cast parameters to the desired type in order to avoid potential issues with overloaded methods
|
||||
ALWAYS_CAST_PARAMS = 1
|
||||
|
||||
class VariableDeclarator(object):
|
||||
def __init__(self, typename, identifier): self.typename = typename; self.local = identifier
|
||||
|
||||
def print_(self, printer, print_):
|
||||
return '{} {}'.format(print_(self.typename), print_(self.local))
|
||||
|
||||
def tree(self, printer, tree): return [tree(self.typename), tree(self.local)]
|
||||
|
||||
#############################################################################################################################################
|
||||
|
||||
class JavaStatement(object):
|
||||
expr = None # provide default for subclasses that don't have an expression
|
||||
def getScopes(self): return ()
|
||||
|
||||
def fixLiterals(self):
|
||||
if self.expr is not None:
|
||||
self.expr = self.expr.fixLiterals()
|
||||
|
||||
def addCastsAndParens(self, env):
|
||||
if self.expr is not None:
|
||||
self.expr.addCasts(env)
|
||||
self.expr.addParens()
|
||||
|
||||
class ExpressionStatement(JavaStatement):
|
||||
def __init__(self, expr):
|
||||
self.expr = expr
|
||||
assert expr is not None
|
||||
|
||||
def print_(self, printer, print_): return print_(self.expr) + ';'
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, tree(self.expr)]
|
||||
|
||||
class LocalDeclarationStatement(JavaStatement):
|
||||
def __init__(self, decl, expr=None):
|
||||
self.decl = decl
|
||||
self.expr = expr
|
||||
|
||||
def print_(self, printer, print_):
|
||||
if self.expr is not None:
|
||||
return '{} = {};'.format(print_(self.decl), print_(self.expr))
|
||||
return print_(self.decl) + ';'
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, tree(self.expr), tree(self.decl)]
|
||||
|
||||
def addCastsAndParens(self, env):
|
||||
if self.expr is not None:
|
||||
self.expr.addCasts(env)
|
||||
|
||||
if not isJavaAssignable(env, self.expr.dtype, self.decl.typename.tt):
|
||||
self.expr = makeCastExpr(self.decl.typename.tt, self.expr, fixEnv=env)
|
||||
self.expr.addParens()
|
||||
|
||||
class ReturnStatement(JavaStatement):
|
||||
def __init__(self, expr=None, tt=None):
|
||||
self.expr = expr
|
||||
self.tt = tt
|
||||
|
||||
def print_(self, printer, print_): return 'return {};'.format(print_(self.expr)) if self.expr is not None else 'return;'
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, tree(self.expr)]
|
||||
|
||||
def addCastsAndParens(self, env):
|
||||
if self.expr is not None:
|
||||
self.expr.addCasts(env)
|
||||
if not isJavaAssignable(env, self.expr.dtype, self.tt):
|
||||
self.expr = makeCastExpr(self.tt, self.expr, fixEnv=env)
|
||||
self.expr.addParens()
|
||||
|
||||
class ThrowStatement(JavaStatement):
|
||||
def __init__(self, expr):
|
||||
self.expr = expr
|
||||
def print_(self, printer, print_): return 'throw {};'.format(print_(self.expr))
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, tree(self.expr)]
|
||||
|
||||
class JumpStatement(JavaStatement):
|
||||
def __init__(self, target, isFront):
|
||||
self.label = target.getLabel() if target is not None else None
|
||||
self.keyword = 'continue' if isFront else 'break'
|
||||
|
||||
def print_(self, printer, print_):
|
||||
label = ' ' + self.label if self.label is not None else ''
|
||||
return self.keyword + label + ';'
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, self.keyword, self.label]
|
||||
|
||||
# Compound Statements
|
||||
sbcount = itertools.count()
|
||||
class LazyLabelBase(JavaStatement):
|
||||
# Jumps are represented by arbitrary 'keys', currently just the key of the
|
||||
# original proxy node. Each item has a continueKey and a breakKey representing
|
||||
# the beginning and the point just past the end respectively. breakKey may be
|
||||
# None if this item appears at the end of the function and there is nothing after it.
|
||||
# Statement blocks have a jump key representing where it jumps to if any. This
|
||||
# may be None if the jump is unreachable (such as if there is a throw or return)
|
||||
def __init__(self, labelfunc, begink, endk):
|
||||
self.label, self.func = None, labelfunc
|
||||
self.continueKey = begink
|
||||
self.breakKey = endk
|
||||
self.id = next(sbcount) # For debugging purposes
|
||||
|
||||
def getLabel(self):
|
||||
if self.label is None:
|
||||
self.label = self.func() # Not a bound function!
|
||||
return self.label
|
||||
|
||||
def getLabelPrefix(self): return '' if self.label is None else self.label + ': '
|
||||
# def getLabelPrefix(self): return self.getLabel() + ': '
|
||||
|
||||
# For debugging
|
||||
def __str__(self): # pragma: no cover
|
||||
if isinstance(self, StatementBlock):
|
||||
return 'Sb'+str(self.id)
|
||||
return type(self).__name__[:3]+str(self.id)
|
||||
__repr__ = __str__
|
||||
|
||||
class TryStatement(LazyLabelBase):
|
||||
def __init__(self, labelfunc, begink, endk, tryb, pairs):
|
||||
super(TryStatement, self).__init__(labelfunc, begink, endk)
|
||||
self.tryb, self.pairs = tryb, pairs
|
||||
|
||||
def getScopes(self): return (self.tryb,) + zip(*self.pairs)[1]
|
||||
|
||||
def print_(self, printer, print_):
|
||||
tryb = print_(self.tryb)
|
||||
parts = ['catch({}) {}'.format(print_(x), print_(y)) for x,y in self.pairs]
|
||||
return '{}try {} {}'.format(self.getLabelPrefix(), tryb, '\n'.join(parts))
|
||||
|
||||
def tree(self, printer, tree):
|
||||
parts = [map(tree, t) for t in self.pairs]
|
||||
return [self.__class__.__name__, self.label, tree(self.tryb), parts]
|
||||
|
||||
class IfStatement(LazyLabelBase):
|
||||
def __init__(self, labelfunc, begink, endk, expr, scopes):
|
||||
super(IfStatement, self).__init__(labelfunc, begink, endk)
|
||||
self.expr = expr # don't rename without changing how var replacement works!
|
||||
self.scopes = scopes
|
||||
|
||||
def getScopes(self): return self.scopes
|
||||
|
||||
def print_(self, printer, print_):
|
||||
lbl = self.getLabelPrefix()
|
||||
parts = [self.expr] + list(self.scopes)
|
||||
|
||||
if len(self.scopes) == 1:
|
||||
parts = [print_(x) for x in parts]
|
||||
return '{}if ({}) {}'.format(lbl, *parts)
|
||||
|
||||
# Special case handling for 'else if'
|
||||
fblock = self.scopes[1]
|
||||
if len(fblock.statements) == 1:
|
||||
stmt = fblock.statements[-1]
|
||||
if isinstance(stmt, IfStatement) and stmt.label is None:
|
||||
parts[-1] = stmt
|
||||
parts = [print_(x) for x in parts]
|
||||
return '{}if ({}) {} else {}'.format(lbl, *parts)
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, self.label, tree(self.expr), map(tree, self.scopes)]
|
||||
|
||||
class SwitchStatement(LazyLabelBase):
|
||||
def __init__(self, labelfunc, begink, endk, expr, pairs):
|
||||
super(SwitchStatement, self).__init__(labelfunc, begink, endk)
|
||||
self.expr = expr # don't rename without changing how var replacement works!
|
||||
self.pairs = pairs
|
||||
|
||||
def getScopes(self): return zip(*self.pairs)[1]
|
||||
def hasDefault(self): return None in zip(*self.pairs)[0]
|
||||
|
||||
def print_(self, printer, print_):
|
||||
expr = print_(self.expr)
|
||||
|
||||
def printCase(keys):
|
||||
if keys is None:
|
||||
return 'default: '
|
||||
assert keys
|
||||
return ''.join(map('case {}: '.format, sorted(keys)))
|
||||
|
||||
bodies = [(printCase(keys) + print_(scope)) for keys, scope in self.pairs]
|
||||
if self.pairs[-1][0] is None and len(self.pairs[-1][1].statements) == 0:
|
||||
bodies.pop()
|
||||
|
||||
contents = '\n'.join(bodies)
|
||||
indented = [' '+line for line in contents.splitlines()]
|
||||
lines = ['{'] + indented + ['}']
|
||||
return '{}switch({}) {}'.format(self.getLabelPrefix(), expr, '\n'.join(lines))
|
||||
|
||||
def tree(self, printer, tree):
|
||||
parts = []
|
||||
for keys, scope in self.pairs:
|
||||
parts.append([[None] if keys is None else sorted(keys), tree(scope)])
|
||||
return [self.__class__.__name__, self.label, tree(self.expr), parts]
|
||||
|
||||
class WhileStatement(LazyLabelBase):
|
||||
def __init__(self, labelfunc, begink, endk, parts):
|
||||
super(WhileStatement, self).__init__(labelfunc, begink, endk)
|
||||
self.expr = Literal.TRUE
|
||||
self.parts = parts
|
||||
assert len(self.parts) == 1
|
||||
|
||||
def getScopes(self): return self.parts
|
||||
|
||||
def print_(self, printer, print_):
|
||||
parts = print_(self.expr), print_(self.parts[0])
|
||||
return '{}while({}) {}'.format(self.getLabelPrefix(), *parts)
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, self.label, tree(self.expr), tree(self.parts[0])]
|
||||
|
||||
class StatementBlock(LazyLabelBase):
|
||||
def __init__(self, labelfunc, begink, endk, statements, jumpk, labelable=True):
|
||||
super(StatementBlock, self).__init__(labelfunc, begink, endk)
|
||||
self.parent = None # should be assigned later
|
||||
self.statements = statements
|
||||
self.jumpKey = jumpk
|
||||
self.labelable = labelable
|
||||
|
||||
def doesFallthrough(self): return self.jumpKey is None or self.jumpKey == self.breakKey
|
||||
|
||||
def getScopes(self): return self,
|
||||
|
||||
def print_(self, printer, print_):
|
||||
assert self.labelable or self.label is None
|
||||
contents = '\n'.join(print_(x) for x in self.statements)
|
||||
indented = [' '+line for line in contents.splitlines()]
|
||||
# indented[:0] = [' //{} {}'.format(self,x) for x in (self.continueKey, self.breakKey, self.jumpKey)]
|
||||
lines = [self.getLabelPrefix() + '{'] + indented + ['}']
|
||||
return '\n'.join(lines)
|
||||
|
||||
@staticmethod
|
||||
def join(*scopes):
|
||||
blists = [s.bases for s in scopes if s is not None] # allow None to represent the universe (top element)
|
||||
if not blists:
|
||||
return None
|
||||
common = [x for x in zip(*blists) if len(set(x)) == 1]
|
||||
return common[-1][0]
|
||||
|
||||
def tree(self, printer, tree): return ['BlockStatement', self.label, map(tree, self.statements)]
|
||||
|
||||
#############################################################################################################################################
|
||||
# Careful, order is important here!
|
||||
_assignable_sprims = objtypes.ByteTT, objtypes.ShortTT, objtypes.CharTT
|
||||
_assignable_lprims = objtypes.IntTT, objtypes.LongTT, objtypes.FloatTT, objtypes.DoubleTT
|
||||
|
||||
# Also used in boolize.py
|
||||
def isPrimativeAssignable(x, y): # x = fromt, y = to
|
||||
assert objtypes.dim(x) == objtypes.dim(y) == 0
|
||||
|
||||
if x == y or (x in _assignable_sprims and y in _assignable_lprims):
|
||||
return True
|
||||
elif (x in _assignable_lprims and y in _assignable_lprims):
|
||||
return _assignable_lprims.index(x) <= _assignable_lprims.index(y)
|
||||
else:
|
||||
return (x, y) == (objtypes.ByteTT, objtypes.ShortTT)
|
||||
|
||||
def isReferenceType(tt):
|
||||
return tt == objtypes.NullTT or objtypes.dim(tt) or (objtypes.className(tt) is not None)
|
||||
|
||||
def isJavaAssignable(env, fromt, to):
|
||||
if fromt is None or to is None: # this should never happen, except during debugging
|
||||
return True
|
||||
|
||||
if isReferenceType(to):
|
||||
assert isReferenceType(fromt)
|
||||
# todo - make it check interfaces too
|
||||
return objtypes.isSubtype(env, fromt, to)
|
||||
else: # allowed if numeric conversion is widening
|
||||
return isPrimativeAssignable(fromt, to)
|
||||
|
||||
_int_tts = objtypes.LongTT, objtypes.IntTT, objtypes.ShortTT, objtypes.CharTT, objtypes.ByteTT
|
||||
def makeCastExpr(newtt, expr, fixEnv=None):
|
||||
if newtt == expr.dtype:
|
||||
return expr
|
||||
|
||||
# if casting a literal with compatible type, just create a literal of the new type
|
||||
if isinstance(expr, Literal):
|
||||
allowed_conversions = [
|
||||
(objtypes.FloatTT, objtypes.DoubleTT),
|
||||
(objtypes.IntTT, objtypes.LongTT),
|
||||
(objtypes.IntTT, objtypes.BoolTT),
|
||||
(objtypes.BoolTT, objtypes.IntTT),
|
||||
]
|
||||
if (expr.dtype, newtt) in allowed_conversions:
|
||||
return Literal(newtt, expr.val)
|
||||
|
||||
if newtt == objtypes.IntTT and expr.dtype == objtypes.BoolTT:
|
||||
return Ternary(expr, Literal.ONE, Literal.ZERO)
|
||||
elif newtt == objtypes.BoolTT and expr.dtype == objtypes.IntTT:
|
||||
return BinaryInfix('!=', [expr, Literal.ZERO], objtypes.BoolTT)
|
||||
|
||||
ret = Cast(TypeName(newtt), expr)
|
||||
if fixEnv is not None:
|
||||
ret = ret.fix(fixEnv)
|
||||
return ret
|
||||
#############################################################################################################################################
|
||||
# Precedence:
|
||||
# 0 - pseudoprimary
|
||||
# 5 - pseudounary
|
||||
# 10-19 binary infix
|
||||
# 20 - ternary
|
||||
# 21 - assignment
|
||||
# Associativity: L = Left, R = Right, A = Full
|
||||
|
||||
class JavaExpression(object):
|
||||
precedence = 0 # Default precedence
|
||||
params = [] # for subclasses that don't have params
|
||||
|
||||
def complexity(self): return 1 + max(e.complexity() for e in self.params) if self.params else 0
|
||||
|
||||
def postFlatIter(self):
|
||||
return itertools.chain([self], *[expr.postFlatIter() for expr in self.params])
|
||||
|
||||
def print_(self, printer, print_):
|
||||
return self.fmt.format(*[print_(expr) for expr in self.params])
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, map(tree, self.params)]
|
||||
|
||||
def replaceSubExprs(self, rdict):
|
||||
if self in rdict:
|
||||
return rdict[self]
|
||||
self.params = [param.replaceSubExprs(rdict) for param in self.params]
|
||||
return self
|
||||
|
||||
def fixLiterals(self):
|
||||
self.params = [param.fixLiterals() for param in self.params]
|
||||
return self
|
||||
|
||||
def addCasts(self, env):
|
||||
for param in self.params:
|
||||
param.addCasts(env)
|
||||
self.addCasts_sub(env)
|
||||
|
||||
def addCasts_sub(self, env): pass
|
||||
|
||||
def addParens(self):
|
||||
for param in self.params:
|
||||
param.addParens()
|
||||
self.params = list(self.params) # Copy before editing, just to be extra safe
|
||||
self.addParens_sub()
|
||||
|
||||
def addParens_sub(self): pass
|
||||
|
||||
def isLocalAssign(self): return isinstance(self, Assignment) and isinstance(self.params[0], Local)
|
||||
|
||||
def __repr__(self): # pragma: no cover
|
||||
return type(self).__name__.rpartition('.')[-1] + ' ' + visitor.DefaultVisitor().visit(self)
|
||||
__str__ = __repr__
|
||||
|
||||
class ArrayAccess(JavaExpression):
|
||||
def __init__(self, *params):
|
||||
if params[0].dtype == objtypes.NullTT:
|
||||
# Unfortunately, Java doesn't really support array access on null constants
|
||||
#So we'll just cast it to Object[] as a hack
|
||||
param = makeCastExpr(objtypes.withDimInc(objtypes.ObjectTT, 1), params[0])
|
||||
params = param, params[1]
|
||||
|
||||
self.params = list(params)
|
||||
self.fmt = '{}[{}]'
|
||||
|
||||
@property
|
||||
def dtype(self): return objtypes.withDimInc(self.params[0].dtype, -1)
|
||||
|
||||
def addParens_sub(self):
|
||||
p0 = self.params[0]
|
||||
if p0.precedence > 0 or isinstance(p0, ArrayCreation):
|
||||
self.params[0] = Parenthesis(p0)
|
||||
|
||||
class ArrayCreation(JavaExpression):
|
||||
def __init__(self, tt, *sizeargs):
|
||||
self.dim = objtypes.dim(tt)
|
||||
self.params = [TypeName(objtypes.withNoDim(tt))] + list(sizeargs)
|
||||
self.dtype = tt
|
||||
assert self.dim >= len(sizeargs) > 0
|
||||
self.fmt = 'new {}' + '[{}]'*len(sizeargs) + '[]'*(self.dim-len(sizeargs))
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, map(tree, self.params), self.dim]
|
||||
|
||||
class Assignment(JavaExpression):
|
||||
precedence = 21
|
||||
def __init__(self, *params):
|
||||
self.params = list(params)
|
||||
self.fmt = '{} = {}'
|
||||
|
||||
@property
|
||||
def dtype(self): return self.params[0].dtype
|
||||
|
||||
def addCasts_sub(self, env):
|
||||
left, right = self.params
|
||||
if not isJavaAssignable(env, right.dtype, left.dtype):
|
||||
expr = makeCastExpr(left.dtype, right, fixEnv=env)
|
||||
self.params = [left, expr]
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, map(tree, self.params), '']
|
||||
|
||||
_binary_ptable = ['* / %', '+ -', '<< >> >>>',
|
||||
'< > <= >= instanceof', '== !=',
|
||||
'&', '^', '|', '&&', '||']
|
||||
|
||||
binary_precedences = {}
|
||||
for _ops, _val in zip(_binary_ptable, range(10,20)):
|
||||
for _op in _ops.split():
|
||||
binary_precedences[_op] = _val
|
||||
|
||||
class BinaryInfix(JavaExpression):
|
||||
def __init__(self, opstr, params, dtype=None):
|
||||
assert len(params) == 2
|
||||
self.params = params
|
||||
self.opstr = opstr
|
||||
self.fmt = '{{}} {} {{}}'.format(opstr)
|
||||
self._dtype = dtype
|
||||
self.precedence = binary_precedences[opstr]
|
||||
|
||||
@property
|
||||
def dtype(self): return self.params[0].dtype if self._dtype is None else self._dtype
|
||||
|
||||
def addParens_sub(self):
|
||||
myprec = self.precedence
|
||||
associative = myprec >= 15 # for now we treat +, *, etc as nonassociative due to floats
|
||||
|
||||
for i, p in enumerate(self.params):
|
||||
if p.precedence > myprec:
|
||||
self.params[i] = Parenthesis(p)
|
||||
elif p.precedence == myprec and i > 0 and not associative:
|
||||
self.params[i] = Parenthesis(p)
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, map(tree, self.params), self.opstr]
|
||||
|
||||
class Cast(JavaExpression):
|
||||
precedence = 5
|
||||
def __init__(self, *params):
|
||||
self.dtype = params[0].tt
|
||||
self.params = list(params)
|
||||
self.fmt = '({}){}'
|
||||
|
||||
def fix(self, env):
|
||||
tt, expr = self.dtype, self.params[1]
|
||||
# "Impossible" casts are a compile error in Java.
|
||||
# This can be fixed with an intermediate cast to Object
|
||||
if isReferenceType(tt):
|
||||
if not isJavaAssignable(env, tt, expr.dtype):
|
||||
if not isJavaAssignable(env, expr.dtype, tt):
|
||||
expr = makeCastExpr(objtypes.ObjectTT, expr)
|
||||
self.params = [self.params[0], expr]
|
||||
return self
|
||||
|
||||
def addCasts_sub(self, env): self.fix(env)
|
||||
def addParens_sub(self):
|
||||
p1 = self.params[1]
|
||||
if p1.precedence > 5 or (isinstance(p1, UnaryPrefix) and p1.opstr[0] in '-+'):
|
||||
self.params[1] = Parenthesis(p1)
|
||||
|
||||
class ClassInstanceCreation(JavaExpression):
|
||||
def __init__(self, typename, tts, arguments):
|
||||
self.typename, self.tts, self.params = typename, tts, arguments
|
||||
self.dtype = typename.tt
|
||||
|
||||
def print_(self, printer, print_):
|
||||
return 'new {}({})'.format(print_(self.typename), ', '.join(print_(x) for x in self.params))
|
||||
|
||||
def tree(self, printer, tree):
|
||||
return [self.__class__.__name__, map(tree, self.params), tree(self.typename)]
|
||||
|
||||
def addCasts_sub(self, env):
|
||||
newparams = []
|
||||
for tt, expr in zip(self.tts, self.params):
|
||||
if expr.dtype != tt and (ALWAYS_CAST_PARAMS or not isJavaAssignable(env, expr.dtype, tt)):
|
||||
expr = makeCastExpr(tt, expr, fixEnv=env)
|
||||
newparams.append(expr)
|
||||
self.params = newparams
|
||||
|
||||
class FieldAccess(JavaExpression):
|
||||
def __init__(self, primary, name, dtype, op=None, printLeft=True):
|
||||
self.dtype = dtype
|
||||
self.params = [primary]
|
||||
self.op, self.name = op, name
|
||||
self.printLeft = printLeft
|
||||
# self.params, self.name = [primary], escapeString(name)
|
||||
# self.fmt = ('{}.' if printLeft else '') + self.name
|
||||
|
||||
def print_(self, printer, print_):
|
||||
if self.op is None:
|
||||
name = self.name
|
||||
assert name in ('length','class')
|
||||
else:
|
||||
cls, name, desc = self.op.target, self.op.name, self.op.desc
|
||||
name = escapeString(printer.fieldName(cls, name, desc))
|
||||
pre = print_(self.params[0])+'.' if self.printLeft else ''
|
||||
return pre+name
|
||||
|
||||
def tree(self, printer, tree):
|
||||
if self.op is None:
|
||||
trip = None, self.name, None
|
||||
else:
|
||||
trip = self.op.target, self.op.name, self.op.desc
|
||||
return [self.__class__.__name__, map(tree, self.params), trip, self.printLeft]
|
||||
|
||||
def addParens_sub(self):
|
||||
p0 = self.params[0]
|
||||
if p0.precedence > 0:
|
||||
self.params[0] = Parenthesis(p0)
|
||||
|
||||
def printFloat(x, isSingle):
|
||||
assert x >= 0.0 and not math.isinf(x)
|
||||
suffix = 'f' if isSingle else ''
|
||||
if isSingle and x > 0.0:
|
||||
# Try to find more compract representation for floats, since repr treats everything as doubles
|
||||
m, e = math.frexp(x)
|
||||
half_ulp2 = math.ldexp(1.0, max(e - 25, -150)) # don't bother doubling when near the upper range of a given e value
|
||||
half_ulp1 = (half_ulp2/2) if m == 0.5 and e >= -125 else half_ulp2
|
||||
lbound, ubound = x-half_ulp1, x+half_ulp2
|
||||
assert lbound < x < ubound
|
||||
s = '{:g}'.format(x).replace('+','')
|
||||
if lbound < float(s) < ubound: # strict ineq to avoid potential double rounding issues
|
||||
return s + suffix
|
||||
return repr(x) + suffix
|
||||
|
||||
class Literal(JavaExpression):
|
||||
def __init__(self, vartype, val):
|
||||
self.dtype = vartype
|
||||
self.val = val
|
||||
if self.dtype == objtypes.ClassTT:
|
||||
self.params = [TypeName(val)]
|
||||
|
||||
def getStr(self):
|
||||
if self.dtype == objtypes.StringTT:
|
||||
return '"' + escapeString(self.val) + '"'
|
||||
elif self.dtype == objtypes.IntTT:
|
||||
return str(self.val)
|
||||
elif self.dtype == objtypes.LongTT:
|
||||
return str(self.val) + 'L'
|
||||
elif self.dtype == objtypes.FloatTT or self.dtype == objtypes.DoubleTT:
|
||||
return printFloat(self.val, self.dtype == objtypes.FloatTT)
|
||||
elif self.dtype == objtypes.NullTT:
|
||||
return 'null'
|
||||
elif self.dtype == objtypes.BoolTT:
|
||||
return 'true' if self.val else 'false'
|
||||
|
||||
def fixLiterals(self):
|
||||
# From the point of view of the Java Language, there is no such thing as a negative literal.
|
||||
# This replaces invalid literal values with unary minus (and division for non-finite floats)
|
||||
if self.dtype == objtypes.IntTT or self.dtype == objtypes.LongTT:
|
||||
if self.val < 0:
|
||||
return UnaryPrefix('-', Literal(self.dtype, -self.val))
|
||||
elif self.dtype == objtypes.FloatTT or self.dtype == objtypes.DoubleTT:
|
||||
x = self.val
|
||||
zero = Literal.DZERO if self.dtype == objtypes.DoubleTT else Literal.FZERO
|
||||
if math.isnan(x):
|
||||
return BinaryInfix('/', [zero, zero])
|
||||
elif math.isinf(x): #+/- inf
|
||||
numerator = Literal(self.dtype, math.copysign(1.0, x)).fixLiterals()
|
||||
return BinaryInfix('/', [numerator, zero])
|
||||
# finite negative numbers
|
||||
if math.copysign(1.0, x) == -1.0:
|
||||
return UnaryPrefix('-', Literal(self.dtype, math.copysign(x, 1.0)))
|
||||
|
||||
return self
|
||||
|
||||
def print_(self, printer, print_):
|
||||
if self.dtype == objtypes.ClassTT:
|
||||
# for printing class literals
|
||||
return '{}.class'.format(print_(self.params[0]))
|
||||
return self.getStr()
|
||||
|
||||
def tree(self, printer, tree):
|
||||
result = tree(self.params[0]) if self.dtype == objtypes.ClassTT else self.getStr()
|
||||
return [self.__class__.__name__, result, self.dtype]
|
||||
|
||||
def _key(self): return self.dtype, self.val
|
||||
def __eq__(self, other): return type(self) == type(other) and self._key() == other._key()
|
||||
def __ne__(self, other): return type(self) != type(other) or self._key() != other._key()
|
||||
def __hash__(self): return hash(self._key())
|
||||
Literal.FALSE = Literal(objtypes.BoolTT, 0)
|
||||
Literal.TRUE = Literal(objtypes.BoolTT, 1)
|
||||
Literal.N_ONE = Literal(objtypes.IntTT, -1)
|
||||
Literal.ZERO = Literal(objtypes.IntTT, 0)
|
||||
Literal.ONE = Literal(objtypes.IntTT, 1)
|
||||
|
||||
Literal.LZERO = Literal(objtypes.LongTT, 0)
|
||||
Literal.FZERO = Literal(objtypes.FloatTT, 0.0)
|
||||
Literal.DZERO = Literal(objtypes.DoubleTT, 0.0)
|
||||
Literal.NULL = Literal(objtypes.NullTT, None)
|
||||
|
||||
_init_d = {objtypes.BoolTT: Literal.FALSE,
|
||||
objtypes.IntTT: Literal.ZERO,
|
||||
objtypes.LongTT: Literal.LZERO,
|
||||
objtypes.FloatTT: Literal.FZERO,
|
||||
objtypes.DoubleTT: Literal.DZERO}
|
||||
def dummyLiteral(tt):
|
||||
return _init_d.get(tt, Literal.NULL)
|
||||
|
||||
class Local(JavaExpression):
|
||||
def __init__(self, vartype, namefunc):
|
||||
self.dtype = vartype
|
||||
self.name = None
|
||||
self.func = namefunc
|
||||
|
||||
def print_(self, printer, print_):
|
||||
if self.name is None:
|
||||
self.name = self.func(self)
|
||||
return self.name
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, self.print_(None, None)]
|
||||
|
||||
class MethodInvocation(JavaExpression):
|
||||
def __init__(self, left, name, tts, arguments, op, dtype):
|
||||
if left is None:
|
||||
self.params = arguments
|
||||
else:
|
||||
self.params = [left] + arguments
|
||||
self.hasLeft = (left is not None)
|
||||
self.dtype = dtype
|
||||
self.name = name
|
||||
self.tts = tts
|
||||
self.op = op # keep around for future reference and new merging
|
||||
|
||||
def print_(self, printer, print_):
|
||||
cls, name, desc = self.op.target, self.op.name, self.op.desc
|
||||
if name != self.name:
|
||||
assert name == '<init>'
|
||||
name = self.name
|
||||
else:
|
||||
name = escapeString(printer.methodName(cls, name, desc))
|
||||
|
||||
if self.hasLeft:
|
||||
left, arguments = self.params[0], self.params[1:]
|
||||
return '{}.{}({})'.format(print_(left), name, ', '.join(print_(x) for x in arguments))
|
||||
else:
|
||||
arguments = self.params
|
||||
return '{}({})'.format(name, ', '.join(print_(x) for x in arguments))
|
||||
|
||||
def tree(self, printer, tree):
|
||||
trip = self.op.target, self.op.name, self.op.desc
|
||||
return [self.__class__.__name__, map(tree, self.params), trip, self.name, self.hasLeft]
|
||||
|
||||
def addCasts_sub(self, env):
|
||||
newparams = []
|
||||
for tt, expr in zip(self.tts, self.params):
|
||||
if expr.dtype != tt and (ALWAYS_CAST_PARAMS or not isJavaAssignable(env, expr.dtype, tt)):
|
||||
expr = makeCastExpr(tt, expr, fixEnv=env)
|
||||
newparams.append(expr)
|
||||
self.params = newparams
|
||||
|
||||
def addParens_sub(self):
|
||||
if self.hasLeft:
|
||||
p0 = self.params[0]
|
||||
if p0.precedence > 0:
|
||||
self.params[0] = Parenthesis(p0)
|
||||
|
||||
class Parenthesis(JavaExpression):
|
||||
def __init__(self, param):
|
||||
self.params = [param]
|
||||
self.fmt = '({})'
|
||||
|
||||
@property
|
||||
def dtype(self): return self.params[0].dtype
|
||||
|
||||
class Ternary(JavaExpression):
|
||||
precedence = 20
|
||||
def __init__(self, *params):
|
||||
self.params = list(params)
|
||||
self.fmt = '{} ? {} : {}'
|
||||
|
||||
@property
|
||||
def dtype(self): return self.params[1].dtype
|
||||
|
||||
def addParens_sub(self):
|
||||
# Add unecessary parenthesis to complex conditions for readability
|
||||
if self.params[0].precedence >= 20 or self.params[0].complexity() > 0:
|
||||
self.params[0] = Parenthesis(self.params[0])
|
||||
if self.params[2].precedence > 20:
|
||||
self.params[2] = Parenthesis(self.params[2])
|
||||
|
||||
class TypeName(JavaExpression):
|
||||
def __init__(self, tt):
|
||||
self.dtype = None
|
||||
self.tt = tt
|
||||
|
||||
def print_(self, printer, print_):
|
||||
name = objtypes.className(self.tt)
|
||||
if name is not None:
|
||||
name = printer.className(name)
|
||||
name = escapeString(name.replace('/','.'))
|
||||
if name.rpartition('.')[0] == 'java.lang':
|
||||
name = name.rpartition('.')[2]
|
||||
else:
|
||||
name = objtypes.primName(self.tt)
|
||||
s = name + '[]'*objtypes.dim(self.tt)
|
||||
return s
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, self.tt]
|
||||
|
||||
def complexity(self): return -1 # exprs which have this as a param won't be bumped up to 1 uncessarily
|
||||
|
||||
class CatchTypeNames(JavaExpression): # Used for caught exceptions, which can have multiple types specified
|
||||
def __init__(self, env, tts):
|
||||
assert(tts and not any(objtypes.dim(tt) for tt in tts)) # at least one type, no array types
|
||||
self.tnames = map(TypeName, tts)
|
||||
self.dtype = objtypes.commonSupertype(env, tts)
|
||||
|
||||
def print_(self, printer, print_):
|
||||
return ' | '.join(print_(tn) for tn in self.tnames)
|
||||
|
||||
def tree(self, printer, tree): return [self.__class__.__name__, map(tree, self.tnames)]
|
||||
|
||||
class UnaryPrefix(JavaExpression):
|
||||
precedence = 5
|
||||
def __init__(self, opstr, param, dtype=None):
|
||||
self.params = [param]
|
||||
self.opstr = opstr
|
||||
self.fmt = opstr + '{}'
|
||||
self._dtype = dtype
|
||||
|
||||
@property
|
||||
def dtype(self): return self.params[0].dtype if self._dtype is None else self._dtype
|
||||
|
||||
def addParens_sub(self):
|
||||
p0 = self.params[0]
|
||||
if p0.precedence > 5 or (isinstance(p0, UnaryPrefix) and p0.opstr[0] == self.opstr[0]):
|
||||
self.params[0] = Parenthesis(p0)
|
||||
|
||||
def tree(self, printer, tree): return ['Unary', map(tree, self.params), self.opstr, False]
|
||||
|
||||
class Dummy(JavaExpression):
|
||||
def __init__(self, fmt, params, isNew=False, dtype=None):
|
||||
self.params = params
|
||||
self.fmt = fmt
|
||||
self.isNew = isNew
|
||||
self.dtype = dtype
|
||||
|
|
@ -0,0 +1,138 @@
|
|||
from ..ssa import objtypes
|
||||
|
||||
from . import ast
|
||||
from .stringescape import escapeString as escape
|
||||
|
||||
class Comments(object):
|
||||
def __init__(self):
|
||||
self.lines = []
|
||||
|
||||
def add(self, s):
|
||||
self.lines.extend(s.strip('\n').split('\n'))
|
||||
|
||||
def print_(self, printer, print_):
|
||||
return ''.join(map('// {}\n'.format, self.lines))
|
||||
|
||||
class MethodDef(object):
|
||||
def __init__(self, class_, flags, name, desc, retType, paramDecls, body):
|
||||
self.flagstr = flags + ' ' if flags else ''
|
||||
self.retType, self.paramDecls = retType, paramDecls
|
||||
self.body = body
|
||||
self.comments = Comments()
|
||||
self.triple = class_.name, name, desc
|
||||
self.throws = None
|
||||
|
||||
if name == '<clinit>':
|
||||
self.isStaticInit, self.isConstructor = True, False
|
||||
elif name == '<init>':
|
||||
self.isStaticInit, self.isConstructor = False, True
|
||||
self.clsname = ast.TypeName(objtypes.TypeTT(class_.name, 0))
|
||||
else:
|
||||
self.isStaticInit, self.isConstructor = False, False
|
||||
|
||||
def print_(self, printer, print_):
|
||||
header = print_(self.comments)
|
||||
argstr = ', '.join(print_(decl) for decl in self.paramDecls)
|
||||
if self.isStaticInit:
|
||||
header += 'static'
|
||||
elif self.isConstructor:
|
||||
name = print_(self.clsname).rpartition('.')[-1]
|
||||
header += '{}{}({})'.format(self.flagstr, name, argstr)
|
||||
else:
|
||||
name = printer.methodName(*self.triple)
|
||||
header += '{}{} {}({})'.format(self.flagstr, print_(self.retType), escape(name), argstr)
|
||||
|
||||
if self.throws is not None:
|
||||
header += ' throws ' + print_(self.throws)
|
||||
|
||||
if self.body is None:
|
||||
if 'abstract' not in self.flagstr and 'native' not in self.flagstr:
|
||||
# Create dummy body for decompiler error
|
||||
return header + ' {/*error*/throw null;}\n'
|
||||
return header + ';\n'
|
||||
else:
|
||||
return header + ' ' + print_(self.body)
|
||||
|
||||
def tree(self, printer, tree):
|
||||
return {
|
||||
'triple': self.triple,
|
||||
'flags': self.flagstr.split(),
|
||||
'ret': tree(self.retType),
|
||||
'params': map(tree, self.paramDecls),
|
||||
'comments': self.comments.lines,
|
||||
'body': tree(self.body),
|
||||
'throws': tree(self.throws),
|
||||
}
|
||||
|
||||
class FieldDef(object):
|
||||
def __init__(self, flags, type_, class_, name, desc, expr=None):
|
||||
self.flagstr = flags + ' ' if flags else ''
|
||||
self.type_ = type_
|
||||
self.name = name
|
||||
self.expr = None if expr is None else ast.makeCastExpr(type_.tt, expr)
|
||||
self.triple = class_.name, name, desc
|
||||
|
||||
def print_(self, printer, print_):
|
||||
name = escape(printer.fieldName(*self.triple))
|
||||
if self.expr is not None:
|
||||
return '{}{} {} = {};'.format(self.flagstr, print_(self.type_), name, print_(self.expr))
|
||||
return '{}{} {};'.format(self.flagstr, print_(self.type_), name)
|
||||
|
||||
def tree(self, printer, tree):
|
||||
return {
|
||||
'triple': self.triple,
|
||||
'type': tree(self.type_),
|
||||
'flags': self.flagstr.split(),
|
||||
'expr': tree(self.expr),
|
||||
}
|
||||
|
||||
class ClassDef(object):
|
||||
def __init__(self, flags, isInterface, name, superc, interfaces, fields, methods):
|
||||
self.flagstr = flags + ' ' if flags else ''
|
||||
self.isInterface = isInterface
|
||||
self.name = ast.TypeName(objtypes.TypeTT(name,0))
|
||||
self.super = ast.TypeName(objtypes.TypeTT(superc,0)) if superc is not None else None
|
||||
self.interfaces = [ast.TypeName(objtypes.TypeTT(iname,0)) for iname in interfaces]
|
||||
self.fields = fields
|
||||
self.methods = methods
|
||||
if superc == 'java/lang/Object':
|
||||
self.super = None
|
||||
|
||||
def print_(self, printer, print_):
|
||||
contents = ''
|
||||
if self.fields:
|
||||
contents = '\n'.join(print_(x) for x in self.fields)
|
||||
if self.methods:
|
||||
if contents:
|
||||
contents += '\n\n' # extra line to divide fields and methods
|
||||
contents += '\n\n'.join(print_(x) for x in self.methods)
|
||||
|
||||
indented = [' '+line for line in contents.splitlines()]
|
||||
name = print_(self.name).rpartition('.')[-1]
|
||||
defname = 'interface' if self.isInterface else 'class'
|
||||
header = '{}{} {}'.format(self.flagstr, defname, name)
|
||||
|
||||
if self.super:
|
||||
header += ' extends ' + print_(self.super)
|
||||
if self.interfaces:
|
||||
if self.isInterface:
|
||||
assert self.super is None
|
||||
header += ' extends ' + ', '.join(print_(x) for x in self.interfaces)
|
||||
else:
|
||||
header += ' implements ' + ', '.join(print_(x) for x in self.interfaces)
|
||||
|
||||
lines = [header + ' {'] + indented + ['}']
|
||||
return '\n'.join(lines) + '\n'
|
||||
|
||||
# Experimental - don't use!
|
||||
def tree(self, printer, tree):
|
||||
return {
|
||||
'rawname': objtypes.className(self.name.tt),
|
||||
'name': tree(self.name),
|
||||
'super': tree(self.super),
|
||||
'flags': self.flagstr.split(),
|
||||
'isInterface': self.isInterface,
|
||||
'interfaces': map(tree, self.interfaces),
|
||||
'fields': map(tree, self.fields),
|
||||
'methods': map(tree, self.methods),
|
||||
}
|
||||
|
|
@ -0,0 +1,322 @@
|
|||
from .. import opnames
|
||||
from ..namegen import LabelGen
|
||||
from ..ssa import objtypes, ssa_jumps, ssa_ops, ssa_types
|
||||
from ..verifier.descriptors import parseFieldDescriptor, parseMethodDescriptor
|
||||
|
||||
from . import ast
|
||||
from .setree import SEBlockItem, SEIf, SEScope, SESwitch, SETry, SEWhile
|
||||
|
||||
# prefixes for name generation
|
||||
_prefix_map = {objtypes.IntTT:'i', objtypes.LongTT:'j',
|
||||
objtypes.FloatTT:'f', objtypes.DoubleTT:'d',
|
||||
objtypes.BoolTT:'b', objtypes.StringTT:'s'}
|
||||
|
||||
_ssaToTT = {ssa_types.SSA_INT:objtypes.IntTT, ssa_types.SSA_LONG:objtypes.LongTT,
|
||||
ssa_types.SSA_FLOAT:objtypes.FloatTT, ssa_types.SSA_DOUBLE:objtypes.DoubleTT}
|
||||
class VarInfo(object):
|
||||
def __init__(self, method, blocks, namegen):
|
||||
self.env = method.class_.env
|
||||
self.labelgen = LabelGen().next
|
||||
|
||||
returnTypes = parseMethodDescriptor(method.descriptor, unsynthesize=False)[-1]
|
||||
self.return_tt = objtypes.verifierToSynthetic(returnTypes[0]) if returnTypes else None
|
||||
self.clsname = method.class_.name
|
||||
self._namegen = namegen
|
||||
|
||||
self._uninit_vars = {}
|
||||
self._vars = {}
|
||||
self._tts = {}
|
||||
for block in blocks:
|
||||
for var, uc in block.unaryConstraints.items():
|
||||
if var.type == ssa_types.SSA_OBJECT:
|
||||
tt = uc.getSingleTType() # temp hack
|
||||
if uc.types.isBoolOrByteArray():
|
||||
tt = objtypes.TypeTT(objtypes.BExpr, objtypes.dim(tt)+1)
|
||||
# assert (objtypes.BoolTT[0], tt[1]) in uc.types.exact
|
||||
else:
|
||||
tt = _ssaToTT[var.type]
|
||||
self._tts[var] = tt
|
||||
|
||||
def _nameCallback(self, expr):
|
||||
prefix = _prefix_map.get(expr.dtype, 'a')
|
||||
return self._namegen.getPrefix(prefix)
|
||||
|
||||
def _newVar(self, var, num, isCast):
|
||||
tt = self._tts[var]
|
||||
if var.const is not None and not isCast:
|
||||
return ast.Literal(tt, var.const)
|
||||
|
||||
if var.name:
|
||||
# important to not add num when it is 0, since we currently
|
||||
# use var names to force 'this'
|
||||
temp = '{}_{}'.format(var.name, num) if num else var.name
|
||||
if isCast:
|
||||
temp += 'c'
|
||||
namefunc = lambda expr:temp
|
||||
else:
|
||||
namefunc = self._nameCallback
|
||||
|
||||
result = ast.Local(tt, namefunc)
|
||||
# merge all variables of uninitialized type to simplify fixObjectCreations in javamethod.py
|
||||
if var.uninit_orig_num is not None and not isCast:
|
||||
result = self._uninit_vars.setdefault(var.uninit_orig_num, result)
|
||||
return result
|
||||
|
||||
def var(self, node, var, isCast=False):
|
||||
key = node, var, isCast
|
||||
try:
|
||||
return self._vars[key]
|
||||
except KeyError:
|
||||
new = self._newVar(key[1], key[0].num, key[2])
|
||||
self._vars[key] = new
|
||||
return new
|
||||
|
||||
def customVar(self, tt, prefix): # for use with ignored exceptions
|
||||
namefunc = lambda expr: self._namegen.getPrefix(prefix)
|
||||
return ast.Local(tt, namefunc)
|
||||
|
||||
#########################################################################################
|
||||
_math_types = (ssa_ops.IAdd, ssa_ops.IDiv, ssa_ops.IMul, ssa_ops.IRem, ssa_ops.ISub)
|
||||
_math_types += (ssa_ops.IAnd, ssa_ops.IOr, ssa_ops.IShl, ssa_ops.IShr, ssa_ops.IUshr, ssa_ops.IXor)
|
||||
_math_types += (ssa_ops.FAdd, ssa_ops.FDiv, ssa_ops.FMul, ssa_ops.FRem, ssa_ops.FSub)
|
||||
_math_symbols = dict(zip(_math_types, '+ / * % - & | << >> >>> ^ + / * % -'.split()))
|
||||
def _convertJExpr(op, getExpr, clsname):
|
||||
params = [getExpr(var) for var in op.params]
|
||||
assert None not in params
|
||||
expr = None
|
||||
|
||||
# Have to do this one seperately since it isn't an expression statement
|
||||
if isinstance(op, ssa_ops.Throw):
|
||||
return ast.ThrowStatement(params[0])
|
||||
|
||||
if isinstance(op, _math_types):
|
||||
opdict = _math_symbols
|
||||
expr = ast.BinaryInfix(opdict[type(op)], params)
|
||||
elif isinstance(op, ssa_ops.ArrLength):
|
||||
expr = ast.FieldAccess(params[0], 'length', objtypes.IntTT)
|
||||
elif isinstance(op, ssa_ops.ArrLoad):
|
||||
expr = ast.ArrayAccess(*params)
|
||||
elif isinstance(op, ssa_ops.ArrStore):
|
||||
expr = ast.ArrayAccess(params[0], params[1])
|
||||
expr = ast.Assignment(expr, params[2])
|
||||
elif isinstance(op, ssa_ops.CheckCast):
|
||||
expr = ast.Cast(ast.TypeName(op.target_tt), params[0])
|
||||
elif isinstance(op, ssa_ops.Convert):
|
||||
expr = ast.makeCastExpr(_ssaToTT[op.target], params[0])
|
||||
elif isinstance(op, (ssa_ops.FCmp, ssa_ops.ICmp)):
|
||||
boolt = objtypes.BoolTT
|
||||
cn1, c0, c1 = ast.Literal.N_ONE, ast.Literal.ZERO, ast.Literal.ONE
|
||||
|
||||
ascend = isinstance(op, ssa_ops.ICmp) or op.NaN_val == 1
|
||||
if ascend:
|
||||
expr = ast.Ternary(ast.BinaryInfix('<',params,boolt), cn1, ast.Ternary(ast.BinaryInfix('==',params,boolt), c0, c1))
|
||||
else:
|
||||
assert op.NaN_val == -1
|
||||
expr = ast.Ternary(ast.BinaryInfix('>',params,boolt), c1, ast.Ternary(ast.BinaryInfix('==',params,boolt), c0, cn1))
|
||||
elif isinstance(op, ssa_ops.FieldAccess):
|
||||
dtype = objtypes.verifierToSynthetic(parseFieldDescriptor(op.desc, unsynthesize=False)[0])
|
||||
|
||||
if op.instruction[0] in (opnames.GETSTATIC, opnames.PUTSTATIC):
|
||||
printLeft = (op.target != clsname) # Don't print classname if it is a static field in current class
|
||||
tt = objtypes.TypeTT(op.target, 0) # Doesn't handle arrays, but they don't have any fields anyway
|
||||
expr = ast.FieldAccess(ast.TypeName(tt), op.name, dtype, op, printLeft=printLeft)
|
||||
else:
|
||||
expr = ast.FieldAccess(params[0], op.name, dtype, op)
|
||||
|
||||
if op.instruction[0] in (opnames.PUTFIELD, opnames.PUTSTATIC):
|
||||
expr = ast.Assignment(expr, params[-1])
|
||||
|
||||
elif isinstance(op, ssa_ops.FNeg):
|
||||
expr = ast.UnaryPrefix('-', params[0])
|
||||
elif isinstance(op, ssa_ops.InstanceOf):
|
||||
args = [params[0], ast.TypeName(op.target_tt)]
|
||||
expr = ast.BinaryInfix('instanceof', args, dtype=objtypes.BoolTT)
|
||||
elif isinstance(op, ssa_ops.Invoke):
|
||||
vtypes, rettypes = parseMethodDescriptor(op.desc, unsynthesize=False)
|
||||
tt_types = objtypes.verifierToSynthetic_seq(vtypes)
|
||||
ret_type = objtypes.verifierToSynthetic(rettypes[0]) if rettypes else None
|
||||
target_tt = op.target_tt
|
||||
|
||||
if objtypes.dim(target_tt) and op.name == "clone": # In Java, T[].clone returns T[] rather than Object
|
||||
ret_type = target_tt
|
||||
|
||||
if op.instruction[0] == opnames.INVOKEINIT and op.isThisCtor:
|
||||
name = 'this' if (op.target == clsname) else 'super'
|
||||
expr = ast.MethodInvocation(None, name, tt_types, params[1:], op, ret_type)
|
||||
elif op.instruction[0] == opnames.INVOKESTATIC: # TODO - fix this for special super calls
|
||||
expr = ast.MethodInvocation(ast.TypeName(target_tt), op.name, [None]+tt_types, params, op, ret_type)
|
||||
else:
|
||||
expr = ast.MethodInvocation(params[0], op.name, [target_tt]+tt_types, params[1:], op, ret_type)
|
||||
elif isinstance(op, ssa_ops.InvokeDynamic):
|
||||
vtypes, rettypes = parseMethodDescriptor(op.desc, unsynthesize=False)
|
||||
ret_type = objtypes.verifierToSynthetic(rettypes[0]) if rettypes else None
|
||||
fmt = '/*invokedynamic*/'
|
||||
if ret_type is not None:
|
||||
fmt += '{{{}}}'.format(len(params))
|
||||
params.append(ast.dummyLiteral(ret_type))
|
||||
expr = ast.Dummy(fmt, params, dtype=ret_type)
|
||||
elif isinstance(op, ssa_ops.Monitor):
|
||||
fmt = '/*monexit({})*/' if op.exit else '/*monenter({})*/'
|
||||
expr = ast.Dummy(fmt, params)
|
||||
elif isinstance(op, ssa_ops.MultiNewArray):
|
||||
expr = ast.ArrayCreation(op.tt, *params)
|
||||
elif isinstance(op, ssa_ops.New):
|
||||
expr = ast.Dummy('/*<unmerged new> {}*/', [ast.TypeName(op.tt)], isNew=True)
|
||||
elif isinstance(op, ssa_ops.NewArray):
|
||||
expr = ast.ArrayCreation(op.tt, params[0])
|
||||
elif isinstance(op, ssa_ops.Truncate):
|
||||
tt = {(True,16): objtypes.ShortTT, (False,16): objtypes.CharTT, (True,8): objtypes.ByteTT}[op.signed, op.width]
|
||||
expr = ast.Cast(ast.TypeName(tt), params[0])
|
||||
if op.rval is not None and expr:
|
||||
expr = ast.Assignment(getExpr(op.rval), expr)
|
||||
|
||||
if expr is None: # Temporary hack
|
||||
if isinstance(op, (ssa_ops.TryReturn, ssa_ops.ExceptionPhi, ssa_ops.MagicThrow)):
|
||||
return None # Don't print out anything
|
||||
return ast.ExpressionStatement(expr)
|
||||
|
||||
#########################################################################################
|
||||
def _createASTBlock(info, endk, node):
|
||||
getExpr = lambda var: info.var(node, var)
|
||||
op2expr = lambda op: _convertJExpr(op, getExpr, info.clsname)
|
||||
|
||||
block = node.block
|
||||
if block is not None:
|
||||
split_ind = 0
|
||||
if isinstance(block.jump, ssa_jumps.OnException):
|
||||
# find index of first throwing instruction, so we can insert eassigns before it later
|
||||
assert isinstance(block.lines[-1], ssa_ops.ExceptionPhi)
|
||||
split_ind = block.lines.index(block.lines[-1].params[0].origin)
|
||||
|
||||
lines_before = filter(None, map(op2expr, block.lines[:split_ind]))
|
||||
lines_after = filter(None, map(op2expr, block.lines[split_ind:]))
|
||||
else:
|
||||
lines_before, lines_after = [], []
|
||||
|
||||
# Kind of hackish: If the block ends in a cast and hence it is not known to always
|
||||
# succeed, assign the results of the cast rather than passing through the variable
|
||||
# unchanged. The cast will actually be second to last in block.lines due to the ephi
|
||||
outreplace = {}
|
||||
if block and len(block.lines) >= 2:
|
||||
temp_op = block.lines[-2]
|
||||
if lines_after and isinstance(temp_op, ssa_ops.CheckCast):
|
||||
assert isinstance(lines_after[-1].expr, ast.Cast)
|
||||
var = temp_op.params[0]
|
||||
cexpr = lines_after[-1].expr
|
||||
lhs = info.var(node, var, True)
|
||||
assert lhs != cexpr.params[1]
|
||||
lines_after[-1].expr = ast.Assignment(lhs, cexpr)
|
||||
nvar = outreplace[var] = lines_after[-1].expr.params[0]
|
||||
nvar.dtype = cexpr.dtype
|
||||
|
||||
eassigns = []
|
||||
nassigns = []
|
||||
for n2 in node.successors:
|
||||
assert (n2 in node.outvars) != (n2 in node.eassigns)
|
||||
if n2 in node.eassigns:
|
||||
for outv, inv in zip(node.eassigns[n2], n2.invars):
|
||||
if outv is None: # this is how we mark the thrown exception, which
|
||||
# obviously doesn't get an explicit assignment statement
|
||||
continue
|
||||
expr = ast.Assignment(info.var(n2, inv), info.var(node, outv))
|
||||
if expr.params[0] != expr.params[1]:
|
||||
eassigns.append(ast.ExpressionStatement(expr))
|
||||
else:
|
||||
for outv, inv in zip(node.outvars[n2], n2.invars):
|
||||
right = outreplace.get(outv, info.var(node, outv))
|
||||
expr = ast.Assignment(info.var(n2, inv), right)
|
||||
if expr.params[0] != expr.params[1]:
|
||||
nassigns.append(ast.ExpressionStatement(expr))
|
||||
|
||||
# Need to put exception assignments before first throwing statement
|
||||
# While normal assignments must come last as they may depend on it
|
||||
statements = lines_before + eassigns + lines_after + nassigns
|
||||
|
||||
norm_successors = node.normalSuccessors()
|
||||
jump = None if block is None else block.jump
|
||||
if isinstance(jump, (ssa_jumps.Rethrow, ssa_jumps.Return)):
|
||||
assert not norm_successors
|
||||
assert not node.eassigns and not node.outvars
|
||||
if isinstance(jump, ssa_jumps.Rethrow):
|
||||
param = info.var(node, jump.params[-1])
|
||||
statements.append(ast.ThrowStatement(param))
|
||||
else:
|
||||
if len(jump.params) > 0:
|
||||
param = info.var(node, jump.params[0])
|
||||
statements.append(ast.ReturnStatement(param, info.return_tt))
|
||||
else:
|
||||
statements.append(ast.ReturnStatement())
|
||||
breakKey, jumpKey = endk, None
|
||||
elif len(norm_successors) == 0:
|
||||
assert isinstance(jump, ssa_jumps.OnException)
|
||||
breakKey, jumpKey = endk, None
|
||||
elif len(norm_successors) == 1: # normal successors
|
||||
breakKey, jumpKey = endk, norm_successors[0]._key
|
||||
else: # case of if and switch jumps handled in parent scope
|
||||
assert len(norm_successors) > 1
|
||||
breakKey, jumpKey = endk, endk
|
||||
|
||||
new = ast.StatementBlock(info.labelgen, node._key, breakKey, statements, jumpKey)
|
||||
assert None not in statements
|
||||
return new
|
||||
|
||||
_cmp_strs = dict(zip(('eq','ne','lt','ge','gt','le'), "== != < >= > <=".split()))
|
||||
def _createASTSub(info, current, ftitem, forceUnlabled=False):
|
||||
begink = current.entryBlock._key
|
||||
endk = ftitem.entryBlock._key if ftitem is not None else None
|
||||
|
||||
if isinstance(current, SEBlockItem):
|
||||
return _createASTBlock(info, endk, current.node)
|
||||
elif isinstance(current, SEScope):
|
||||
ftitems = current.items[1:] + [ftitem]
|
||||
parts = [_createASTSub(info, item, newft) for item, newft in zip(current.items, ftitems)]
|
||||
return ast.StatementBlock(info.labelgen, begink, endk, parts, endk, labelable=(not forceUnlabled))
|
||||
elif isinstance(current, SEWhile):
|
||||
parts = [_createASTSub(info, scope, current, True) for scope in current.getScopes()]
|
||||
return ast.WhileStatement(info.labelgen, begink, endk, tuple(parts))
|
||||
elif isinstance(current, SETry):
|
||||
assert len(current.getScopes()) == 2
|
||||
parts = [_createASTSub(info, scope, ftitem, True) for scope in current.getScopes()]
|
||||
catchnode = current.getScopes()[-1].entryBlock
|
||||
declt = ast.CatchTypeNames(info.env, current.toptts)
|
||||
|
||||
if current.catchvar is None: # exception is ignored and hence not referred to by the graph, so we need to make our own
|
||||
catchvar = info.customVar(declt.dtype, 'ignoredException')
|
||||
else:
|
||||
catchvar = info.var(catchnode, current.catchvar)
|
||||
decl = ast.VariableDeclarator(declt, catchvar)
|
||||
pairs = [(decl, parts[1])]
|
||||
return ast.TryStatement(info.labelgen, begink, endk, parts[0], pairs)
|
||||
|
||||
# Create a fake key to represent the beginning of the conditional statement itself
|
||||
# doesn't matter what it is as long as it's unique
|
||||
midk = begink + (-1,)
|
||||
node = current.head.node
|
||||
jump = node.block.jump
|
||||
|
||||
if isinstance(current, SEIf):
|
||||
parts = [_createASTSub(info, scope, ftitem, True) for scope in current.getScopes()]
|
||||
cmp_str = _cmp_strs[jump.cmp]
|
||||
exprs = [info.var(node, var) for var in jump.params]
|
||||
ifexpr = ast.BinaryInfix(cmp_str, exprs, objtypes.BoolTT)
|
||||
new = ast.IfStatement(info.labelgen, midk, endk, ifexpr, tuple(parts))
|
||||
|
||||
elif isinstance(current, SESwitch):
|
||||
ftitems = current.ordered[1:] + [ftitem]
|
||||
parts = [_createASTSub(info, item, newft, True) for item, newft in zip(current.ordered, ftitems)]
|
||||
for part in parts:
|
||||
part.breakKey = endk # createSub will assume break should be ft, which isn't the case with switch statements
|
||||
|
||||
expr = info.var(node, jump.params[0])
|
||||
pairs = zip(current.ordered_keysets, parts)
|
||||
new = ast.SwitchStatement(info.labelgen, midk, endk, expr, pairs)
|
||||
|
||||
# bundle head and if together so we can return as single statement
|
||||
headscope = _createASTBlock(info, midk, node)
|
||||
assert headscope.jumpKey is midk
|
||||
return ast.StatementBlock(info.labelgen, begink, endk, [headscope, new], endk)
|
||||
|
||||
def createAST(method, ssagraph, seroot, namegen):
|
||||
info = VarInfo(method, ssagraph.blocks, namegen)
|
||||
astroot = _createASTSub(info, seroot, None)
|
||||
return astroot, info
|
||||
|
|
@ -0,0 +1,179 @@
|
|||
import collections
|
||||
|
||||
from .. import graph_util
|
||||
from ..ssa import objtypes
|
||||
from ..ssa.objtypes import BExpr, BoolTT, ByteTT, CharTT, IntTT, ShortTT
|
||||
|
||||
from . import ast
|
||||
|
||||
# Class union-find data structure except that we don't bother with weighting trees and singletons are implicit
|
||||
# Also, booleans are forced to be seperate roots
|
||||
FORCED_ROOTS = True, False
|
||||
class UnionFind(object):
|
||||
def __init__(self):
|
||||
self.d = {}
|
||||
|
||||
def find(self, x):
|
||||
if x not in self.d:
|
||||
return x
|
||||
path = [x]
|
||||
while path[-1] in self.d:
|
||||
path.append(self.d[path[-1]])
|
||||
root = path.pop()
|
||||
for y in path:
|
||||
self.d[y] = root
|
||||
return root
|
||||
|
||||
def union(self, x, x2):
|
||||
if x is None or x2 is None:
|
||||
return
|
||||
root1, root2 = self.find(x), self.find(x2)
|
||||
if root2 in FORCED_ROOTS:
|
||||
root1, root2 = root2, root1
|
||||
if root1 != root2 and root2 not in FORCED_ROOTS:
|
||||
self.d[root2] = root1
|
||||
|
||||
##############################################################
|
||||
def visitStatementTree(scope, callback, catchcb=None):
|
||||
for item in scope.statements:
|
||||
for sub in item.getScopes():
|
||||
visitStatementTree(sub, callback, catchcb)
|
||||
if item.expr is not None:
|
||||
callback(item, item.expr)
|
||||
if catchcb is not None and isinstance(item, ast.TryStatement):
|
||||
for pair in item.pairs:
|
||||
catchcb(pair[0])
|
||||
|
||||
int_tags = frozenset(map(objtypes.baset, [IntTT, ShortTT, CharTT, ByteTT, BoolTT]))
|
||||
array_tags = frozenset(map(objtypes.baset, [ByteTT, BoolTT]) + [objtypes.BExpr])
|
||||
|
||||
# Fix int/bool and byte[]/bool[] vars
|
||||
def boolizeVars(root, arg_vars):
|
||||
varlist = []
|
||||
sets = UnionFind()
|
||||
|
||||
def visitExpr(expr, forceExact=False):
|
||||
# see if we have to merge
|
||||
if isinstance(expr, ast.Assignment) or isinstance(expr, ast.BinaryInfix) and expr.opstr in ('==','!=','&','|','^'):
|
||||
subs = [visitExpr(param) for param in expr.params]
|
||||
sets.union(*subs) # these operators can work on either type but need the same type on each side
|
||||
elif isinstance(expr, ast.ArrayAccess):
|
||||
sets.union(False, visitExpr(expr.params[1])) # array index is int only
|
||||
elif isinstance(expr, ast.BinaryInfix) and expr.opstr in ('* / % + - << >> >>>'):
|
||||
sets.union(False, visitExpr(expr.params[0])) # these operators are int only
|
||||
sets.union(False, visitExpr(expr.params[1]))
|
||||
|
||||
if isinstance(expr, ast.Local):
|
||||
tag, dim = objtypes.baset(expr.dtype), objtypes.dim(expr.dtype)
|
||||
if (dim == 0 and tag in int_tags) or (dim > 0 and tag in array_tags):
|
||||
# the only "unknown" vars are bexpr[] and ints. All else have fixed types
|
||||
if forceExact or (tag != BExpr and tag != objtypes.baset(IntTT)):
|
||||
sets.union(tag == objtypes.baset(BoolTT), expr)
|
||||
varlist.append(expr)
|
||||
return sets.find(expr)
|
||||
elif isinstance(expr, ast.Literal):
|
||||
if expr.dtype == IntTT and expr.val not in (0,1):
|
||||
return False
|
||||
return None # if val is 0 or 1, or the literal is a null, it is freely convertable
|
||||
elif isinstance(expr, ast.Assignment) or (isinstance(expr, ast.BinaryInfix) and expr.opstr in ('&','|','^')):
|
||||
return subs[0]
|
||||
elif isinstance(expr, (ast.ArrayAccess, ast.Parenthesis, ast.UnaryPrefix)):
|
||||
return visitExpr(expr.params[0])
|
||||
elif expr.dtype is not None and objtypes.baset(expr.dtype) != BExpr:
|
||||
return expr.dtype[0] == objtypes.baset(BoolTT)
|
||||
return None
|
||||
|
||||
def visitStatement(item, expr):
|
||||
root = visitExpr(expr)
|
||||
if isinstance(item, ast.ReturnStatement):
|
||||
forced_val = (objtypes.baset(item.tt) == objtypes.baset(BoolTT))
|
||||
sets.union(forced_val, root)
|
||||
elif isinstance(item, ast.SwitchStatement):
|
||||
sets.union(False, root) # Switch must take an int, not a bool
|
||||
|
||||
for expr in arg_vars:
|
||||
visitExpr(expr, forceExact=True)
|
||||
visitStatementTree(root, callback=visitStatement)
|
||||
|
||||
# Fix the propagated types
|
||||
for var in set(varlist):
|
||||
tag, dim = objtypes.baset(var.dtype), objtypes.dim(var.dtype)
|
||||
assert tag in int_tags or (dim>0 and tag == BExpr)
|
||||
# make everything bool which is not forced to int
|
||||
if sets.find(var) != False:
|
||||
var.dtype = objtypes.withDimInc(BoolTT, dim)
|
||||
elif dim > 0:
|
||||
var.dtype = objtypes.withDimInc(ByteTT, dim)
|
||||
|
||||
# Fix everything else back up
|
||||
def fixExpr(item, expr):
|
||||
for param in expr.params:
|
||||
fixExpr(None, param)
|
||||
|
||||
if isinstance(expr, ast.Assignment):
|
||||
left, right = expr.params
|
||||
if objtypes.baset(left.dtype) in int_tags and objtypes.dim(left.dtype) == 0:
|
||||
if not ast.isPrimativeAssignable(right.dtype, left.dtype):
|
||||
expr.params = [left, ast.makeCastExpr(left.dtype, right)]
|
||||
elif isinstance(expr, ast.BinaryInfix):
|
||||
a, b = expr.params
|
||||
# shouldn't need to do anything here for arrays
|
||||
if expr.opstr in '== != & | ^' and a.dtype == BoolTT or b.dtype == BoolTT:
|
||||
expr.params = [ast.makeCastExpr(BoolTT, v) for v in expr.params]
|
||||
visitStatementTree(root, callback=fixExpr)
|
||||
|
||||
# Fix vars of interface/object type
|
||||
# TODO: do this properly
|
||||
def interfaceVars(env, root, arg_vars):
|
||||
varlist = []
|
||||
consts = {}
|
||||
assigns = collections.defaultdict(list)
|
||||
|
||||
def isInterfaceVar(expr):
|
||||
if not isinstance(expr, ast.Local) or not objtypes.isBaseTClass(expr.dtype):
|
||||
return False
|
||||
if objtypes.className(expr.dtype) == objtypes.className(objtypes.ObjectTT):
|
||||
return True
|
||||
return env.isInterface(objtypes.className(expr.dtype))
|
||||
|
||||
def updateConst(var, tt):
|
||||
varlist.append(var)
|
||||
if var not in consts:
|
||||
consts[var] = tt
|
||||
else:
|
||||
consts[var] = objtypes.commonSupertype(env, [consts[var], tt])
|
||||
|
||||
def visitStatement(item, expr):
|
||||
if isinstance(expr, ast.Assignment) and objtypes.isBaseTClass(expr.dtype):
|
||||
left, right = expr.params
|
||||
if isInterfaceVar(left):
|
||||
if isInterfaceVar(right):
|
||||
assigns[left].append(right)
|
||||
varlist.append(right)
|
||||
varlist.append(left)
|
||||
else:
|
||||
updateConst(left, right.dtype)
|
||||
|
||||
def visitCatchDecl(decl):
|
||||
updateConst(decl.local, decl.typename.dtype)
|
||||
|
||||
for expr in arg_vars:
|
||||
if objtypes.isBaseTClass(expr.dtype):
|
||||
updateConst(expr, expr.dtype)
|
||||
visitStatementTree(root, callback=visitStatement, catchcb=visitCatchDecl)
|
||||
|
||||
# Now calculate actual types and fix
|
||||
newtypes = {}
|
||||
|
||||
# visit variables in topological order. Doesn't handle case of loops, but this is a temporary hack anyway
|
||||
order = graph_util.topologicalSort(varlist, lambda v:assigns[v])
|
||||
for var in order:
|
||||
assert var not in newtypes
|
||||
|
||||
tts = [newtypes.get(right, objtypes.ObjectTT) for right in assigns[var]]
|
||||
if var in consts:
|
||||
tts.append(consts[var])
|
||||
newtypes[var] = newtype = objtypes.commonSupertype(env, tts)
|
||||
if newtype != objtypes.ObjectTT and newtype != var.dtype and newtype != objtypes.NullTT:
|
||||
# assert objtypes.baset(var.dtype) == objtypes.baset(objtypes.ObjectTT)
|
||||
var.dtype = newtype
|
||||
|
|
@ -0,0 +1,227 @@
|
|||
from collections import defaultdict as ddict
|
||||
|
||||
from .. import graph_util
|
||||
from ..ssa import objtypes
|
||||
|
||||
from . import ast
|
||||
|
||||
def flattenDict(replace):
|
||||
for k in list(replace):
|
||||
while replace[k] in replace:
|
||||
replace[k] = replace[replace[k]]
|
||||
|
||||
# The basic block in our temporary CFG
|
||||
# instead of code, it merely contains a list of defs and uses
|
||||
# This is an extended basic block, i.e. it only terminates in a normal jump(s).
|
||||
# exceptions can be thrown from various points within the block
|
||||
class DUBlock(object):
|
||||
def __init__(self, key):
|
||||
self.key = key
|
||||
self.caught_excepts = ()
|
||||
self.lines = [] # 3 types of lines: ('use', var), ('def', (var, var2_opt)), or ('canthrow', None)
|
||||
self.e_successors = []
|
||||
self.n_successors = []
|
||||
self.vars = None # vars used or defined within the block. Does NOT include caught exceptions
|
||||
|
||||
def canThrow(self): return ('canthrow', None) in self.lines
|
||||
|
||||
def recalcVars(self):
|
||||
self.vars = set()
|
||||
for line_t, data in self.lines:
|
||||
if line_t == 'use':
|
||||
self.vars.add(data)
|
||||
elif line_t == 'def':
|
||||
self.vars.add(data[0])
|
||||
if data[1] is not None:
|
||||
self.vars.add(data[1])
|
||||
|
||||
def replace(self, replace):
|
||||
if not self.vars.isdisjoint(replace):
|
||||
newlines = []
|
||||
for line_t, data in self.lines:
|
||||
if line_t == 'use':
|
||||
data = replace.get(data, data)
|
||||
elif line_t == 'def':
|
||||
data = replace.get(data[0], data[0]), replace.get(data[1], data[1])
|
||||
newlines.append((line_t, data))
|
||||
self.lines = newlines
|
||||
for k, v in replace.items():
|
||||
if k in self.vars:
|
||||
self.vars.remove(k)
|
||||
self.vars.add(v)
|
||||
|
||||
def simplify(self):
|
||||
# try to prune redundant instructions
|
||||
last = None
|
||||
newlines = []
|
||||
for line in self.lines:
|
||||
if line[0] == 'def':
|
||||
if line[1][0] == line[1][1]:
|
||||
continue
|
||||
elif line == last:
|
||||
continue
|
||||
newlines.append(line)
|
||||
last = line
|
||||
self.lines = newlines
|
||||
self.recalcVars()
|
||||
|
||||
def varOrNone(expr):
|
||||
return expr if isinstance(expr, ast.Local) else None
|
||||
|
||||
def canThrow(expr):
|
||||
if isinstance(expr, (ast.ArrayAccess, ast.ArrayCreation, ast.Cast, ast.ClassInstanceCreation, ast.FieldAccess, ast.MethodInvocation)):
|
||||
return True
|
||||
if isinstance(expr, ast.BinaryInfix) and expr.opstr in ('/','%'): # check for possible division by 0
|
||||
return expr.dtype not in (objtypes.FloatTT, objtypes.DoubleTT)
|
||||
return False
|
||||
|
||||
def visitExpr(expr, lines):
|
||||
if expr is None:
|
||||
return
|
||||
if isinstance(expr, ast.Local):
|
||||
lines.append(('use', expr))
|
||||
|
||||
if isinstance(expr, ast.Assignment):
|
||||
lhs, rhs = map(varOrNone, expr.params)
|
||||
|
||||
# with assignment we need to only visit LHS if it isn't a local in order to avoid spurious uses
|
||||
# also, we need to visit RHS before generating the def
|
||||
if lhs is None:
|
||||
visitExpr(expr.params[0], lines)
|
||||
visitExpr(expr.params[1], lines)
|
||||
if lhs is not None:
|
||||
lines.append(('def', (lhs, rhs)))
|
||||
else:
|
||||
for param in expr.params:
|
||||
visitExpr(param, lines)
|
||||
|
||||
if canThrow(expr):
|
||||
lines.append(('canthrow', None))
|
||||
|
||||
class DUGraph(object):
|
||||
def __init__(self):
|
||||
self.blocks = []
|
||||
self.entry = None
|
||||
|
||||
def makeBlock(self, key, break_dict, caught_except, myexcept_parents):
|
||||
block = DUBlock(key)
|
||||
self.blocks.append(block)
|
||||
|
||||
for parent in break_dict[block.key]:
|
||||
parent.n_successors.append(block)
|
||||
del break_dict[block.key]
|
||||
|
||||
assert (myexcept_parents is None) == (caught_except is None)
|
||||
if caught_except is not None: # this is the head of a catch block:
|
||||
block.caught_excepts = (caught_except,)
|
||||
for parent in myexcept_parents:
|
||||
parent.e_successors.append(block)
|
||||
return block
|
||||
|
||||
def finishBlock(self, block, catch_stack):
|
||||
# register exception handlers for completed old block and calculate var set
|
||||
assert(block.vars is None) # make sure it wasn't finished twice
|
||||
if block.canThrow():
|
||||
for clist in catch_stack:
|
||||
clist.append(block)
|
||||
block.recalcVars()
|
||||
|
||||
def visitScope(self, scope, break_dict, catch_stack, caught_except=None, myexcept_parents=None, head_block=None):
|
||||
# catch_stack is copy on modify
|
||||
if head_block is None:
|
||||
head_block = block = self.makeBlock(scope.continueKey, break_dict, caught_except, myexcept_parents)
|
||||
else:
|
||||
block = head_block
|
||||
|
||||
for stmt in scope.statements:
|
||||
if isinstance(stmt, (ast.ExpressionStatement, ast.ThrowStatement, ast.ReturnStatement)):
|
||||
visitExpr(stmt.expr, block.lines)
|
||||
if isinstance(stmt, ast.ThrowStatement):
|
||||
block.lines.append(('canthrow', None))
|
||||
continue
|
||||
|
||||
# compound statements
|
||||
assert stmt.continueKey is not None
|
||||
if isinstance(stmt, (ast.IfStatement, ast.SwitchStatement)):
|
||||
visitExpr(stmt.expr, block.lines)
|
||||
|
||||
if isinstance(stmt, ast.SwitchStatement):
|
||||
ft = not stmt.hasDefault()
|
||||
else:
|
||||
ft = len(stmt.getScopes()) == 1
|
||||
|
||||
for sub in stmt.getScopes():
|
||||
break_dict[sub.continueKey].append(block)
|
||||
self.visitScope(sub, break_dict, catch_stack)
|
||||
if ft:
|
||||
break_dict[stmt.breakKey].append(block)
|
||||
|
||||
elif isinstance(stmt, ast.WhileStatement):
|
||||
if stmt.expr != ast.Literal.TRUE: # while(cond)
|
||||
assert stmt.breakKey is not None
|
||||
self.finishBlock(block, catch_stack)
|
||||
block = self.makeBlock(stmt.continueKey, break_dict, None, None)
|
||||
visitExpr(stmt.expr, block.lines)
|
||||
break_dict[stmt.breakKey].append(block)
|
||||
|
||||
break_dict[stmt.continueKey].append(block)
|
||||
body_block = self.visitScope(stmt.getScopes()[0], break_dict, catch_stack)
|
||||
continue_target = body_block if stmt.expr == ast.Literal.TRUE else block
|
||||
|
||||
for parent in break_dict[stmt.continueKey]:
|
||||
parent.n_successors.append(continue_target)
|
||||
del break_dict[stmt.continueKey]
|
||||
|
||||
elif isinstance(stmt, ast.TryStatement):
|
||||
new_stack = catch_stack + [[] for _ in stmt.pairs]
|
||||
|
||||
break_dict[stmt.tryb.continueKey].append(block)
|
||||
self.visitScope(stmt.tryb, break_dict, new_stack)
|
||||
|
||||
for cdecl, catchb in stmt.pairs:
|
||||
parents = new_stack.pop()
|
||||
self.visitScope(catchb, break_dict, catch_stack, cdecl.local, parents)
|
||||
assert new_stack == catch_stack
|
||||
else:
|
||||
assert isinstance(stmt, ast.StatementBlock)
|
||||
break_dict[stmt.continueKey].append(block)
|
||||
self.visitScope(stmt, break_dict, catch_stack, head_block=block)
|
||||
|
||||
if not isinstance(stmt, ast.StatementBlock): # if we passed it to subscope, it will be finished in the subcall
|
||||
self.finishBlock(block, catch_stack)
|
||||
|
||||
if stmt.breakKey is not None: # start new block after return from compound statement
|
||||
block = self.makeBlock(stmt.breakKey, break_dict, None, None)
|
||||
else:
|
||||
block = None # should never be accessed anyway if we're exiting abruptly
|
||||
|
||||
if scope.jumpKey is not None:
|
||||
break_dict[scope.jumpKey].append(block)
|
||||
|
||||
if block is not None:
|
||||
self.finishBlock(block, catch_stack)
|
||||
return head_block # head needs to be returned in case of loops so we can fix up backedges
|
||||
|
||||
def makeCFG(self, root):
|
||||
break_dict = ddict(list)
|
||||
self.visitScope(root, break_dict, [])
|
||||
self.entry = self.blocks[0] # entry point should always be first block generated
|
||||
|
||||
reached = graph_util.topologicalSort([self.entry], lambda block:(block.n_successors + block.e_successors))
|
||||
# if len(reached) != len(self.blocks):
|
||||
# print 'warning, {} blocks unreachable!'.format(len(self.blocks) - len(reached))
|
||||
self.blocks = reached
|
||||
|
||||
def replace(self, replace):
|
||||
flattenDict(replace)
|
||||
for block in self.blocks:
|
||||
block.replace(replace)
|
||||
|
||||
def simplify(self):
|
||||
for block in self.blocks:
|
||||
block.simplify()
|
||||
|
||||
def makeGraph(root):
|
||||
g = DUGraph()
|
||||
g.makeCFG(root)
|
||||
return g
|
||||
|
|
@ -0,0 +1,127 @@
|
|||
from collections import defaultdict as ddict
|
||||
import itertools
|
||||
|
||||
def unique(seq): return len(set(seq)) == len(seq)
|
||||
|
||||
# This module provides a view of the ssa graph that can be modified without
|
||||
# touching the underlying graph. This proxy is tailored towards the need of
|
||||
# cfg structuring, so it allows easy duplication and indirection of nodes,
|
||||
# but assumes that the underlying variables and statements are immutable
|
||||
|
||||
class BlockProxy(object):
|
||||
def __init__(self, key, counter, block=None):
|
||||
self.bkey = key
|
||||
self.num = next(counter)
|
||||
self.counter = counter
|
||||
self.block = block
|
||||
|
||||
self.predecessors = []
|
||||
self.successors = []
|
||||
self.outvars = {}
|
||||
self.eassigns = {} # exception edge assignments, used after try constraint creation
|
||||
self._key = self.bkey, self.num
|
||||
|
||||
# to be assigned later
|
||||
self.invars = self.blockdict = None
|
||||
# assigned by structuring.py calcNoLoopNeighbors
|
||||
self.successors_nl = self.predecessors_nl = self.norm_suc_nl = None
|
||||
|
||||
def replaceSuccessors(self, rmap):
|
||||
update = lambda k:rmap.get(k,k)
|
||||
|
||||
self.successors = map(update, self.successors)
|
||||
self.outvars = {update(k):v for k,v in self.outvars.items()}
|
||||
if self.block is not None:
|
||||
d1 = self.blockdict
|
||||
self.blockdict = {(b.key,t):update(d1[b.key,t]) for (b,t) in self.block.jump.getSuccessorPairs()}
|
||||
|
||||
def newIndirect(self): # for use during graph creation
|
||||
new = BlockProxy(self.bkey, self.counter)
|
||||
new.invars = self.invars
|
||||
new.outvars = {self:new.invars}
|
||||
new.blockdict = None
|
||||
new.successors = [self]
|
||||
self.predecessors.append(new)
|
||||
return new
|
||||
|
||||
def newDuplicate(self): # for use by structuring.structure return inlining
|
||||
new = BlockProxy(self.bkey, self.counter, self.block)
|
||||
new.invars = self.invars
|
||||
new.outvars = self.outvars.copy()
|
||||
new.blockdict = self.blockdict
|
||||
new.successors = self.successors[:]
|
||||
return new
|
||||
|
||||
def indirectEdges(self, edges):
|
||||
# Should only be called once graph is completely set up. newIndirect is used during graph creation
|
||||
new = self.newIndirect()
|
||||
for parent in edges:
|
||||
self.predecessors.remove(parent)
|
||||
new.predecessors.append(parent)
|
||||
parent.replaceSuccessors({self:new})
|
||||
return new
|
||||
|
||||
def normalSuccessors(self): # only works once try constraints have been created
|
||||
return [x for x in self.successors if x in self.outvars]
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
fmt = 'PB {}x{}' if self.num else 'PB {0}'
|
||||
return fmt.format(self.bkey, self.num)
|
||||
__repr__ = __str__
|
||||
|
||||
|
||||
def createGraphProxy(ssagraph):
|
||||
assert(not ssagraph.procs) # should have already been inlined
|
||||
|
||||
nodes = [BlockProxy(b.key, itertools.count(), block=b) for b in ssagraph.blocks]
|
||||
allnodes = nodes[:] # will also contain indirected nodes
|
||||
|
||||
entryNode = None
|
||||
intypes = ddict(set)
|
||||
for n in nodes:
|
||||
invars = [phi.rval for phi in n.block.phis]
|
||||
for b, t in n.block.jump.getSuccessorPairs():
|
||||
intypes[b.key].add(t)
|
||||
|
||||
if n.bkey == ssagraph.entryKey:
|
||||
assert(not entryNode and not invars) # shouldn't have more than one entryBlock and entryBlock shouldn't have phis
|
||||
entryNode = n
|
||||
invars = ssagraph.inputArgs # store them in the node so we don't have to keep track seperately
|
||||
invars = [x for x in invars if x is not None] # will have None placeholders for Long and Double arguments
|
||||
n.invars = invars
|
||||
|
||||
lookup = {}
|
||||
for n in nodes:
|
||||
assert len(intypes[n.bkey]) != 2 # should have been handled by graph.splitDualInedges()
|
||||
|
||||
if False in intypes[n.bkey]:
|
||||
lookup[n.bkey, False] = n
|
||||
if True in intypes[n.bkey]:
|
||||
lookup[n.bkey, True] = n
|
||||
assert unique(lookup.values())
|
||||
|
||||
for n in nodes:
|
||||
n.blockdict = lookup
|
||||
block = n.block
|
||||
for (block2, t) in block.jump.getSuccessorPairs():
|
||||
out = [phi.get((block, t)) for phi in block2.phis]
|
||||
|
||||
n2 = lookup[block2.key, t]
|
||||
n.outvars[n2] = out
|
||||
n.successors.append(n2)
|
||||
n2.predecessors.append(n)
|
||||
|
||||
# sanity check
|
||||
for n in allnodes:
|
||||
assert (n.block is not None) == (n.num == 0)
|
||||
assert (n is entryNode) == (len(n.predecessors) == 0)
|
||||
assert unique(n.predecessors)
|
||||
assert unique(n.successors)
|
||||
for pn in n.predecessors:
|
||||
assert n in pn.successors
|
||||
assert set(n.outvars) == set(n.successors)
|
||||
for sn in n.successors:
|
||||
assert n in sn.predecessors
|
||||
assert len(n.outvars[sn]) == len(sn.invars)
|
||||
|
||||
return entryNode, allnodes
|
||||
|
|
@ -0,0 +1,70 @@
|
|||
import struct
|
||||
|
||||
from ..ssa import objtypes
|
||||
from ..verifier.descriptors import parseFieldDescriptor
|
||||
|
||||
from . import ast, ast2, javamethod, throws
|
||||
from .reserved import reserved_identifiers
|
||||
|
||||
def loadConstValue(cpool, index):
|
||||
entry_type = cpool.pool[index][0]
|
||||
args = cpool.getArgs(index)
|
||||
|
||||
# Note: field constant values cannot be class literals
|
||||
tt = {'Int':objtypes.IntTT, 'Long':objtypes.LongTT,
|
||||
'Float':objtypes.FloatTT, 'Double':objtypes.DoubleTT,
|
||||
'String':objtypes.StringTT}[entry_type]
|
||||
return ast.Literal(tt, args[0]).fixLiterals()
|
||||
|
||||
def _getField(field):
|
||||
flags = [x.lower() for x in sorted(field.flags) if x not in ('SYNTHETIC','ENUM')]
|
||||
desc = field.descriptor
|
||||
dtype = objtypes.verifierToSynthetic(parseFieldDescriptor(desc, unsynthesize=False)[0])
|
||||
|
||||
initexpr = None
|
||||
if field.static:
|
||||
cpool = field.class_.cpool
|
||||
const_attrs = [data for name,data in field.attributes if name == 'ConstantValue']
|
||||
if const_attrs:
|
||||
assert len(const_attrs) == 1
|
||||
data = const_attrs[0]
|
||||
index = struct.unpack('>h', data)[0]
|
||||
initexpr = loadConstValue(cpool, index)
|
||||
return ast2.FieldDef(' '.join(flags), ast.TypeName(dtype), field.class_, field.name, desc, initexpr)
|
||||
|
||||
def _getMethod(method, cb, forbidden_identifiers, skip_errors):
|
||||
try:
|
||||
graph = cb(method) if method.code is not None else None
|
||||
print 'Decompiling method', method.name.encode('utf8'), method.descriptor.encode('utf8')
|
||||
code_ast = javamethod.generateAST(method, graph, forbidden_identifiers)
|
||||
return code_ast
|
||||
except Exception as e:
|
||||
if not skip_errors:
|
||||
raise
|
||||
|
||||
import traceback
|
||||
message = traceback.format_exc()
|
||||
code_ast = javamethod.generateAST(method, None, forbidden_identifiers)
|
||||
code_ast.comments.add(message)
|
||||
print message
|
||||
return code_ast
|
||||
|
||||
# Method argument allows decompilng only a single method, primarily useful for debugging
|
||||
def generateAST(cls, cb, skip_errors, method=None, add_throws=False):
|
||||
methods = cls.methods if method is None else [cls.methods[method]]
|
||||
fi = set(reserved_identifiers)
|
||||
for field in cls.fields:
|
||||
fi.add(field.name)
|
||||
forbidden_identifiers = frozenset(fi)
|
||||
|
||||
myflags = [x.lower() for x in sorted(cls.flags) if x not in ('INTERFACE','SUPER','SYNTHETIC','ANNOTATION','ENUM')]
|
||||
isInterface = 'INTERFACE' in cls.flags
|
||||
|
||||
superc = cls.supername
|
||||
interfaces = [cls.cpool.getArgsCheck('Class', index) for index in cls.interfaces_raw] # todo - change when class actually loads interfaces
|
||||
|
||||
field_defs = [_getField(f) for f in cls.fields]
|
||||
method_defs = [_getMethod(m, cb, forbidden_identifiers, skip_errors) for m in methods]
|
||||
if add_throws:
|
||||
throws.addSingle(cls.env, method_defs)
|
||||
return ast2.ClassDef(' '.join(myflags), isInterface, cls.name, superc, interfaces, field_defs, method_defs)
|
||||
|
|
@ -0,0 +1,886 @@
|
|||
import collections
|
||||
from functools import partial
|
||||
import operator
|
||||
|
||||
from .. import graph_util
|
||||
from ..namegen import NameGen
|
||||
from ..ssa import objtypes
|
||||
from ..verifier.descriptors import parseMethodDescriptor
|
||||
|
||||
from . import ast, ast2, astgen, boolize, graphproxy, mergevariables, structuring
|
||||
|
||||
class DeclInfo(object):
|
||||
__slots__ = "declScope scope defs".split()
|
||||
def __init__(self):
|
||||
self.declScope = self.scope = None
|
||||
self.defs = []
|
||||
|
||||
def findVarDeclInfo(root, predeclared):
|
||||
info = collections.OrderedDict()
|
||||
def visit(scope, expr):
|
||||
for param in expr.params:
|
||||
visit(scope, param)
|
||||
|
||||
if expr.isLocalAssign():
|
||||
left, right = expr.params
|
||||
info[left].defs.append(right)
|
||||
elif isinstance(expr, (ast.Local, ast.Literal)):
|
||||
# this would be so much nicer if we had Ordered defaultdicts
|
||||
info.setdefault(expr, DeclInfo())
|
||||
info[expr].scope = ast.StatementBlock.join(info[expr].scope, scope)
|
||||
|
||||
def visitDeclExpr(scope, expr):
|
||||
info.setdefault(expr, DeclInfo())
|
||||
assert scope is not None and info[expr].declScope is None
|
||||
info[expr].declScope = scope
|
||||
|
||||
for expr in predeclared:
|
||||
visitDeclExpr(root, expr)
|
||||
|
||||
stack = [(root,root)]
|
||||
while stack:
|
||||
scope, stmt = stack.pop()
|
||||
if isinstance(stmt, ast.StatementBlock):
|
||||
stack.extend((stmt,sub) for sub in stmt.statements)
|
||||
else:
|
||||
stack.extend((subscope,subscope) for subscope in stmt.getScopes())
|
||||
# temp hack
|
||||
if stmt.expr is not None:
|
||||
visit(scope, stmt.expr)
|
||||
if isinstance(stmt, ast.TryStatement):
|
||||
for catchdecl, body in stmt.pairs:
|
||||
visitDeclExpr(body, catchdecl.local)
|
||||
return info
|
||||
|
||||
def reverseBoolExpr(expr):
|
||||
assert expr.dtype == objtypes.BoolTT
|
||||
if isinstance(expr, ast.BinaryInfix):
|
||||
symbols = "== != < >= > <=".split()
|
||||
floatts = (objtypes.FloatTT, objtypes.DoubleTT)
|
||||
if expr.opstr in symbols:
|
||||
sym2 = symbols[symbols.index(expr.opstr) ^ 1]
|
||||
left, right = expr.params
|
||||
# be sure not to reverse floating point comparisons since it's not equivalent for NaN
|
||||
if expr.opstr in symbols[:2] or (left.dtype not in floatts and right.dtype not in floatts):
|
||||
return ast.BinaryInfix(sym2, [left, right], objtypes.BoolTT)
|
||||
elif isinstance(expr, ast.UnaryPrefix) and expr.opstr == '!':
|
||||
return expr.params[0]
|
||||
return ast.UnaryPrefix('!', expr)
|
||||
|
||||
def getSubscopeIter(root):
|
||||
stack = [root]
|
||||
while stack:
|
||||
scope = stack.pop()
|
||||
if isinstance(scope, ast.StatementBlock):
|
||||
stack.extend(scope.statements)
|
||||
yield scope
|
||||
else:
|
||||
stack.extend(scope.getScopes())
|
||||
|
||||
def mayBreakTo(root, forbidden):
|
||||
assert None not in forbidden
|
||||
for scope in getSubscopeIter(root):
|
||||
if scope.jumpKey in forbidden:
|
||||
# We return true if scope has forbidden jump and is reachable
|
||||
# We assume there is no unreachable code, so in order for a scope
|
||||
# jump to be unreachable, it must end in a return, throw, or a
|
||||
# compound statement, all of which are not reachable or do not
|
||||
# break out of the statement. We omit adding last.breakKey to
|
||||
# forbidden since it should always match scope.jumpKey anyway
|
||||
if not scope.statements:
|
||||
return True
|
||||
last = scope.statements[-1]
|
||||
if not last.getScopes():
|
||||
if not isinstance(last, (ast.ReturnStatement, ast.ThrowStatement)):
|
||||
return True
|
||||
else:
|
||||
# If and switch statements may allow fallthrough
|
||||
# A while statement with condition may break implicitly
|
||||
if isinstance(last, ast.IfStatement) and len(last.getScopes()) == 1:
|
||||
return True
|
||||
if isinstance(last, ast.SwitchStatement) and not last.hasDefault():
|
||||
return True
|
||||
if isinstance(last, ast.WhileStatement) and last.expr != ast.Literal.TRUE:
|
||||
return True
|
||||
|
||||
if not isinstance(last, ast.WhileStatement):
|
||||
for sub in last.getScopes():
|
||||
assert sub.breakKey == last.breakKey == scope.jumpKey
|
||||
return False
|
||||
|
||||
def replaceKeys(top, replace):
|
||||
assert None not in replace
|
||||
get = lambda k:replace.get(k,k)
|
||||
|
||||
if top.getScopes():
|
||||
if isinstance(top, ast.StatementBlock) and get(top.breakKey) is None:
|
||||
# breakKey can be None with non-None jumpKey when we're a scope in a switch statement that falls through
|
||||
# and the end of the switch statement is unreachable
|
||||
assert get(top.jumpKey) is None or not top.labelable
|
||||
|
||||
top.breakKey = get(top.breakKey)
|
||||
if isinstance(top, ast.StatementBlock):
|
||||
top.jumpKey = get(top.jumpKey)
|
||||
for item in top.statements:
|
||||
replaceKeys(item, replace)
|
||||
else:
|
||||
for scope in top.getScopes():
|
||||
replaceKeys(scope, replace)
|
||||
|
||||
NONE_SET = frozenset([None])
|
||||
def _preorder(scope, func):
|
||||
newitems = []
|
||||
for i, item in enumerate(scope.statements):
|
||||
for sub in item.getScopes():
|
||||
_preorder(sub, func)
|
||||
|
||||
val = func(scope, item)
|
||||
vals = [item] if val is None else val
|
||||
newitems.extend(vals)
|
||||
scope.statements = newitems
|
||||
|
||||
def _fixObjectCreations(scope, item):
|
||||
'''Combines new/invokeinit pairs into Java constructor calls'''
|
||||
# Thanks to the uninitialized variable merging prior to AST generation,
|
||||
# we can safely assume there are no copies to worry about
|
||||
expr = item.expr
|
||||
if isinstance(expr, ast.Assignment):
|
||||
left, right = expr.params
|
||||
if isinstance(right, ast.Dummy) and right.isNew:
|
||||
return [] # remove item
|
||||
elif isinstance(expr, ast.MethodInvocation) and expr.name == '<init>':
|
||||
left = expr.params[0]
|
||||
newexpr = ast.ClassInstanceCreation(ast.TypeName(left.dtype), expr.tts[1:], expr.params[1:])
|
||||
item.expr = ast.Assignment(left, newexpr)
|
||||
|
||||
def _pruneRethrow_cb(item):
|
||||
'''Convert try{A} catch(T t) {throw t;} to {A}'''
|
||||
while item.pairs:
|
||||
decl, body = item.pairs[-1]
|
||||
caught, lines = decl.local, body.statements
|
||||
|
||||
if len(lines) == 1:
|
||||
line = lines[0]
|
||||
if isinstance(line, ast.ThrowStatement) and line.expr == caught:
|
||||
item.pairs = item.pairs[:-1]
|
||||
continue
|
||||
break
|
||||
if not item.pairs:
|
||||
new = item.tryb
|
||||
assert new.breakKey == item.breakKey
|
||||
assert new.continueKey == item.continueKey
|
||||
assert not new.labelable
|
||||
new.labelable = True
|
||||
return new
|
||||
return item
|
||||
|
||||
def _pruneIfElse_cb(item):
|
||||
'''Convert if(A) {B} else {} to if(A) {B}'''
|
||||
if len(item.scopes) > 1:
|
||||
tblock, fblock = item.scopes
|
||||
|
||||
# if true block is empty, swap it with false so we can remove it
|
||||
if not tblock.statements and tblock.doesFallthrough():
|
||||
item.expr = reverseBoolExpr(item.expr)
|
||||
tblock, fblock = item.scopes = fblock, tblock
|
||||
|
||||
if not fblock.statements and fblock.doesFallthrough():
|
||||
item.scopes = tblock,
|
||||
|
||||
# if(A) {B throw/return ... } else {C} -> if(A) {B throw/return ...} {C}
|
||||
if len(item.scopes) > 1:
|
||||
# How much we want the block to be first, or (3, 0) if it can't be simplified
|
||||
def priority(block):
|
||||
# If an empty block survives to this point, it must end in a break so we can simplify in this case too
|
||||
if not block.statements:
|
||||
return 2, 0
|
||||
# If any complex statements, there might be a labeled break, so it's not safe
|
||||
if any(stmt.getScopes() for stmt in block.statements):
|
||||
return 3, 0
|
||||
# prefer if(...) {throw...} return... to if(...) {return...} throw...
|
||||
if isinstance(block.statements[-1], ast.ThrowStatement):
|
||||
return 0, len(block.statements)
|
||||
elif isinstance(block.statements[-1], ast.ReturnStatement):
|
||||
return 1, len(block.statements)
|
||||
return 3, 0
|
||||
|
||||
if priority(fblock) < priority(tblock):
|
||||
item.expr = reverseBoolExpr(item.expr)
|
||||
tblock, fblock = item.scopes = fblock, tblock
|
||||
|
||||
if priority(tblock) < (3, 0):
|
||||
assert tblock.statements or not tblock.doesFallthrough()
|
||||
item.scopes = tblock,
|
||||
item.breakKey = fblock.continueKey
|
||||
fblock.labelable = True
|
||||
return [item], fblock
|
||||
|
||||
# If cond is !(x), reverse it back to simplify cond
|
||||
if isinstance(item.expr, ast.UnaryPrefix) and item.expr.opstr == '!':
|
||||
item.expr = reverseBoolExpr(item.expr)
|
||||
tblock, fblock = item.scopes = fblock, tblock
|
||||
|
||||
# if(A) {if(B) {C}} -> if(A && B) {C}
|
||||
tblock = item.scopes[0]
|
||||
if len(item.scopes) == 1 and len(tblock.statements) == 1 and tblock.doesFallthrough():
|
||||
first = tblock.statements[0]
|
||||
if isinstance(first, ast.IfStatement) and len(first.scopes) == 1:
|
||||
item.expr = ast.BinaryInfix('&&',[item.expr, first.expr], objtypes.BoolTT)
|
||||
item.scopes = first.scopes
|
||||
return [], item
|
||||
|
||||
def _whileCondition_cb(item):
|
||||
'''Convert while(true) {if(A) {B break;} else {C} D} to while(!A) {{C} D} {B}
|
||||
and while(A) {if(B) {break;} else {C} D} to while(A && !B) {{C} D}'''
|
||||
failure = [], item # what to return if we didn't inline
|
||||
body = item.getScopes()[0]
|
||||
if not body.statements or not isinstance(body.statements[0], ast.IfStatement):
|
||||
return failure
|
||||
|
||||
head = body.statements[0]
|
||||
cond = head.expr
|
||||
trueb, falseb = (head.getScopes() + (None,))[:2]
|
||||
|
||||
# Make sure it doesn't continue the loop or break out of the if statement
|
||||
badjumps1 = frozenset([head.breakKey, item.continueKey]) - NONE_SET
|
||||
if mayBreakTo(trueb, badjumps1):
|
||||
if falseb is not None and not mayBreakTo(falseb, badjumps1):
|
||||
cond = reverseBoolExpr(cond)
|
||||
trueb, falseb = falseb, trueb
|
||||
else:
|
||||
return failure
|
||||
assert not mayBreakTo(trueb, badjumps1)
|
||||
|
||||
|
||||
trivial = not trueb.statements and trueb.jumpKey == item.breakKey
|
||||
# If we already have a condition, only a simple break is allowed
|
||||
if not trivial and item.expr != ast.Literal.TRUE:
|
||||
return failure
|
||||
|
||||
# If break body is nontrival, we can't insert this after the end of the loop unless
|
||||
# We're sure that nothing else in the loop breaks out
|
||||
badjumps2 = frozenset([item.breakKey]) - NONE_SET
|
||||
if not trivial:
|
||||
restloop = [falseb] if falseb is not None else []
|
||||
restloop += body.statements[1:]
|
||||
if body.jumpKey == item.breakKey or any(mayBreakTo(s, badjumps2) for s in restloop):
|
||||
return failure
|
||||
|
||||
# Now inline everything
|
||||
item.expr = _simplifyExpressions(ast.BinaryInfix('&&', [item.expr, reverseBoolExpr(cond)]))
|
||||
if falseb is None:
|
||||
body.continueKey = body.statements.pop(0).breakKey
|
||||
else:
|
||||
body.continueKey = falseb.continueKey
|
||||
body.statements[0] = falseb
|
||||
falseb.labelable = True
|
||||
trueb.labelable = True
|
||||
|
||||
if item.breakKey is None: # Make sure to maintain invariant that bkey=None -> jkey=None
|
||||
assert trueb.doesFallthrough()
|
||||
trueb.jumpKey = trueb.breakKey = None
|
||||
trueb.breakKey = item.breakKey
|
||||
assert trueb.continueKey is not None
|
||||
if not trivial:
|
||||
item.breakKey = trueb.continueKey
|
||||
|
||||
# Trueb doesn't break to head.bkey but there might be unreacahble jumps, so we replace
|
||||
# it too. We don't replace item.ckey because it should never appear, even as an
|
||||
# unreachable jump
|
||||
replaceKeys(trueb, {head.breakKey:trueb.breakKey, item.breakKey:trueb.breakKey})
|
||||
return [item], trueb
|
||||
|
||||
def _simplifyBlocksSub(scope, item, isLast):
|
||||
rest = []
|
||||
if isinstance(item, ast.TryStatement):
|
||||
item = _pruneRethrow_cb(item)
|
||||
elif isinstance(item, ast.IfStatement):
|
||||
rest, item = _pruneIfElse_cb(item)
|
||||
elif isinstance(item, ast.WhileStatement):
|
||||
rest, item = _whileCondition_cb(item)
|
||||
|
||||
if isinstance(item, ast.StatementBlock):
|
||||
assert item.breakKey is not None or item.jumpKey is None
|
||||
# If bkey is None, it can't be broken to
|
||||
# If contents can also break to enclosing scope, it's always safe to inline
|
||||
bkey = item.breakKey
|
||||
if bkey is None or (bkey == scope.breakKey and scope.labelable):
|
||||
rest, item.statements = rest + item.statements, []
|
||||
# Now inline statements at the beginning of the scope that don't need to break to it
|
||||
for sub in item.statements[:]:
|
||||
if sub.getScopes() and sub.breakKey != bkey and mayBreakTo(sub, frozenset([bkey])):
|
||||
break
|
||||
rest.append(item.statements.pop(0))
|
||||
|
||||
if not item.statements:
|
||||
if item.jumpKey != bkey:
|
||||
assert isLast
|
||||
scope.jumpKey = item.jumpKey
|
||||
assert scope.breakKey is not None or scope.jumpKey is None
|
||||
return rest
|
||||
return rest + [item]
|
||||
|
||||
def _simplifyBlocks(scope):
|
||||
newitems = []
|
||||
for item in reversed(scope.statements):
|
||||
isLast = not newitems # may be true if all subsequent items pruned
|
||||
if isLast and item.getScopes():
|
||||
if item.breakKey != scope.jumpKey:# and item.breakKey is not None:
|
||||
replaceKeys(item, {item.breakKey: scope.jumpKey})
|
||||
|
||||
for sub in reversed(item.getScopes()):
|
||||
_simplifyBlocks(sub)
|
||||
vals = _simplifyBlocksSub(scope, item, isLast)
|
||||
newitems += reversed(vals)
|
||||
scope.statements = newitems[::-1]
|
||||
|
||||
|
||||
_op2bits = {'==':2, '!=':13, '<':1, '<=':3, '>':4, '>=':6}
|
||||
_bit2ops_float = {v:k for k,v in _op2bits.items()}
|
||||
_bit2ops = {(v & 7):k for k,v in _op2bits.items()}
|
||||
|
||||
def _getBitfield(expr):
|
||||
if isinstance(expr, ast.BinaryInfix):
|
||||
if expr.opstr in ('==','!=','<','<=','>','>='):
|
||||
# We don't want to merge expressions if they could have side effects
|
||||
# so only allow literals and locals
|
||||
if all(isinstance(p, (ast.Literal, ast.Local)) for p in expr.params):
|
||||
return _op2bits[expr.opstr], list(expr.params)
|
||||
elif expr.opstr in ('&','&&','|','||'):
|
||||
bits1, args1 = _getBitfield(expr.params[0])
|
||||
bits2, args2 = _getBitfield(expr.params[1])
|
||||
if args1 == args2:
|
||||
bits = (bits1 & bits2) if '&' in expr.opstr else (bits1 | bits2)
|
||||
return bits, args1
|
||||
elif isinstance(expr, ast.UnaryPrefix) and expr.opstr == '!':
|
||||
bits, args = _getBitfield(expr.params[0])
|
||||
return ~bits, args
|
||||
return 0, None
|
||||
|
||||
def _mergeComparisons(expr):
|
||||
# a <= b && a != b -> a < b, etc.
|
||||
bits, args = _getBitfield(expr)
|
||||
if args is None:
|
||||
return expr
|
||||
|
||||
assert not hasSideEffects(args[0]) and not hasSideEffects(args[1])
|
||||
if args[0].dtype in (objtypes.FloatTT, objtypes.DoubleTT):
|
||||
mask, d = 15, _bit2ops_float
|
||||
else:
|
||||
mask, d = 7, _bit2ops
|
||||
|
||||
bits &= mask
|
||||
notbits = (~bits) & mask
|
||||
|
||||
if bits == 0:
|
||||
return ast.Literal.TRUE
|
||||
elif notbits == 0:
|
||||
return ast.Literal.FALSE
|
||||
elif bits in d:
|
||||
return ast.BinaryInfix(d[bits], args, objtypes.BoolTT)
|
||||
elif notbits in d:
|
||||
return ast.UnaryPrefix('!', ast.BinaryInfix(d[notbits], args, objtypes.BoolTT))
|
||||
return expr
|
||||
|
||||
def _simplifyExpressions(expr):
|
||||
TRUE, FALSE = ast.Literal.TRUE, ast.Literal.FALSE
|
||||
bools = {True:TRUE, False:FALSE}
|
||||
opfuncs = {'<': operator.lt, '<=': operator.le, '>': operator.gt, '>=': operator.ge}
|
||||
|
||||
simplify = _simplifyExpressions
|
||||
expr.params = map(simplify, expr.params)
|
||||
|
||||
if isinstance(expr, ast.BinaryInfix):
|
||||
left, right = expr.params
|
||||
op = expr.opstr
|
||||
if op in ('==','!=','<','<=','>','>=') and isinstance(right, ast.Literal):
|
||||
# la cmp lb -> result (i.e. constant propagation on literal comparisons)
|
||||
if isinstance(left, ast.Literal):
|
||||
if op in ('==','!='):
|
||||
# these could be string or class literals, but those are always nonnull so it still works
|
||||
res = (left == right) == (op == '==')
|
||||
else:
|
||||
assert left.dtype == right.dtype
|
||||
res = opfuncs[op](left.val, right.val)
|
||||
expr = bools[res]
|
||||
# (a ? lb : c) cmp ld -> a ? (lb cmp ld) : (c cmp ld)
|
||||
elif isinstance(left, ast.Ternary) and isinstance(left.params[1], ast.Literal):
|
||||
left.params[1] = simplify(ast.BinaryInfix(op, [left.params[1], right], expr._dtype))
|
||||
left.params[2] = simplify(ast.BinaryInfix(op, [left.params[2], right], expr._dtype))
|
||||
expr = left
|
||||
|
||||
# a ? true : b -> a || b
|
||||
# a ? false : b -> !a && b
|
||||
if isinstance(expr, ast.Ternary) and expr.dtype == objtypes.BoolTT:
|
||||
cond, val1, val2 = expr.params
|
||||
if not isinstance(val1, ast.Literal): # try to get bool literal to the front
|
||||
cond, val1, val2 = reverseBoolExpr(cond), val2, val1
|
||||
|
||||
if val1 == TRUE:
|
||||
expr = ast.BinaryInfix('||', [cond, val2], objtypes.BoolTT)
|
||||
elif val1 == FALSE:
|
||||
expr = ast.BinaryInfix('&&', [reverseBoolExpr(cond), val2], objtypes.BoolTT)
|
||||
|
||||
# true && a -> a, etc.
|
||||
if isinstance(expr, ast.BinaryInfix) and expr.opstr in ('&&','||'):
|
||||
left, right = expr.params
|
||||
if expr.opstr == '&&':
|
||||
if left == TRUE or (right == FALSE and not hasSideEffects(left)):
|
||||
expr = right
|
||||
elif left == FALSE or right == TRUE:
|
||||
expr = left
|
||||
else:
|
||||
if left == TRUE or right == FALSE:
|
||||
expr = left
|
||||
elif left == FALSE or (right == TRUE and not hasSideEffects(left)):
|
||||
expr = right
|
||||
# a > b || a == b -> a >= b, etc.
|
||||
expr = _mergeComparisons(expr)
|
||||
|
||||
# a == true -> a
|
||||
# a == false -> !a
|
||||
if isinstance(expr, ast.BinaryInfix) and expr.opstr in ('==, !=') and expr.params[0].dtype == objtypes.BoolTT:
|
||||
left, right = expr.params
|
||||
if not isinstance(left, ast.Literal): # try to get bool literal to the front
|
||||
left, right = right, left
|
||||
if isinstance(left, ast.Literal):
|
||||
flip = (left == TRUE) != (expr.opstr == '==')
|
||||
expr = reverseBoolExpr(right) if flip else right
|
||||
|
||||
# !a ? b : c -> a ? c : b
|
||||
if isinstance(expr, ast.Ternary) and isinstance(expr.params[0], ast.UnaryPrefix):
|
||||
cond, val1, val2 = expr.params
|
||||
if cond.opstr == '!':
|
||||
expr.params = [reverseBoolExpr(cond), val2, val1]
|
||||
|
||||
# 0 - a -> -a
|
||||
if isinstance(expr, ast.BinaryInfix) and expr.opstr == '-':
|
||||
if expr.params[0] == ast.Literal.ZERO or expr.params[0] == ast.Literal.LZERO:
|
||||
expr = ast.UnaryPrefix('-', expr.params[1])
|
||||
|
||||
# (double)4.2f -> 4.2, etc.
|
||||
if isinstance(expr, ast.Cast) and isinstance(expr.params[1], ast.Literal):
|
||||
expr = ast.makeCastExpr(expr.dtype, expr.params[1])
|
||||
|
||||
return expr
|
||||
|
||||
def _setScopeParents(scope):
|
||||
for item in scope.statements:
|
||||
for sub in item.getScopes():
|
||||
sub.bases = scope.bases + (sub,)
|
||||
_setScopeParents(sub)
|
||||
|
||||
def _replaceExpressions(scope, item, rdict):
|
||||
# Must be done before local declarations are created since it doesn't touch/remove them
|
||||
if item.expr is not None:
|
||||
item.expr = item.expr.replaceSubExprs(rdict)
|
||||
# remove redundant assignments i.e. x=x;
|
||||
if isinstance(item.expr, ast.Assignment):
|
||||
assert isinstance(item, ast.ExpressionStatement)
|
||||
left, right = item.expr.params
|
||||
if left == right:
|
||||
return []
|
||||
return [item]
|
||||
|
||||
def _oldMergeVariables(root, predeclared):
|
||||
_setScopeParents(root)
|
||||
info = findVarDeclInfo(root, predeclared)
|
||||
|
||||
lvars = [expr for expr in info if isinstance(expr, ast.Local)]
|
||||
forbidden = set()
|
||||
# If var has any defs which aren't a literal or local, mark it as a leaf node (it can't be merged into something)
|
||||
for var in lvars:
|
||||
if not all(isinstance(expr, (ast.Local, ast.Literal)) for expr in info[var].defs):
|
||||
forbidden.add(var)
|
||||
elif info[var].declScope is not None:
|
||||
forbidden.add(var)
|
||||
|
||||
sccs = graph_util.tarjanSCC(lvars, lambda var:([] if var in forbidden else info[var].defs))
|
||||
# the sccs will be in topolgical order
|
||||
varmap = {}
|
||||
for scc in sccs:
|
||||
if forbidden.isdisjoint(scc):
|
||||
alldefs = []
|
||||
for expr in scc:
|
||||
for def_ in info[expr].defs:
|
||||
if def_ not in scc:
|
||||
alldefs.append(varmap[def_])
|
||||
if len(set(alldefs)) == 1:
|
||||
target = alldefs[0]
|
||||
if all(var.dtype == target.dtype for var in scc):
|
||||
scope = ast.StatementBlock.join(*(info[var].scope for var in scc))
|
||||
scope = ast.StatementBlock.join(scope, info[target].declScope) # scope is unchanged if declScope is none like usual
|
||||
if info[target].declScope is None or info[target].declScope == scope:
|
||||
for var in scc:
|
||||
varmap[var] = target
|
||||
info[target].scope = ast.StatementBlock.join(scope, info[target].scope)
|
||||
continue
|
||||
# fallthrough if merging is impossible
|
||||
for var in scc:
|
||||
varmap[var] = var
|
||||
if len(info[var].defs) > 1:
|
||||
forbidden.add(var)
|
||||
_preorder(root, partial(_replaceExpressions, rdict=varmap))
|
||||
|
||||
def _mergeVariables(root, predeclared, isstatic):
|
||||
_oldMergeVariables(root, predeclared)
|
||||
|
||||
rdict = mergevariables.mergeVariables(root, isstatic, predeclared)
|
||||
_preorder(root, partial(_replaceExpressions, rdict=rdict))
|
||||
|
||||
_oktypes = ast.BinaryInfix, ast.Local, ast.Literal, ast.Parenthesis, ast.Ternary, ast.TypeName, ast.UnaryPrefix
|
||||
def hasSideEffects(expr):
|
||||
if not isinstance(expr, _oktypes):
|
||||
return True
|
||||
# check for division by 0. If it's a float or dividing by nonzero literal, it's ok
|
||||
elif isinstance(expr, ast.BinaryInfix) and expr.opstr in ('/','%'):
|
||||
if expr.dtype not in (objtypes.FloatTT, objtypes.DoubleTT):
|
||||
divisor = expr.params[-1]
|
||||
if not isinstance(divisor, ast.Literal) or divisor.val == 0:
|
||||
return True
|
||||
return False
|
||||
|
||||
def _inlineVariables(root):
|
||||
# first find all variables with a single def and use
|
||||
defs = collections.defaultdict(list)
|
||||
uses = collections.defaultdict(int)
|
||||
|
||||
def visitExprFindDefs(expr):
|
||||
if expr.isLocalAssign():
|
||||
defs[expr.params[0]].append(expr)
|
||||
elif isinstance(expr, ast.Local):
|
||||
uses[expr] += 1
|
||||
|
||||
def visitFindDefs(scope, item):
|
||||
if item.expr is not None:
|
||||
stack = [item.expr]
|
||||
while stack:
|
||||
expr = stack.pop()
|
||||
visitExprFindDefs(expr)
|
||||
stack.extend(expr.params)
|
||||
|
||||
_preorder(root, visitFindDefs)
|
||||
# These should have 2 uses since the initial assignment also counts
|
||||
replacevars = {k for k,v in defs.items() if len(v)==1 and uses[k]==2 and k.dtype == v[0].params[1].dtype}
|
||||
def doReplacement(item, pairs):
|
||||
old, new = item.expr.params
|
||||
assert isinstance(old, ast.Local) and old.dtype == new.dtype
|
||||
stack = [(True, (True, item2, expr)) for item2, expr in reversed(pairs) if expr is not None]
|
||||
while stack:
|
||||
recurse, args = stack.pop()
|
||||
|
||||
if recurse:
|
||||
canReplace, parent, expr = args
|
||||
stack.append((False, expr))
|
||||
|
||||
# TODO - fix this for real
|
||||
if expr.complexity() > 30:
|
||||
canReplace = False
|
||||
|
||||
# For ternaries, we don't want to replace into the conditionally
|
||||
# evaluated part, but we still need to check those parts for
|
||||
# barriers. For both ternaries and short circuit operators, the
|
||||
# first param is always evaluated, so it is safe
|
||||
if isinstance(expr, ast.Ternary) or isinstance(expr, ast.BinaryInfix) and expr.opstr in ('&&','||'):
|
||||
for param in reversed(expr.params[1:]):
|
||||
stack.append((True, (False, expr, param)))
|
||||
stack.append((True, (canReplace, expr, expr.params[0])))
|
||||
# For assignments, we unroll the LHS arguments, because if assigning
|
||||
# to an array or field, we don't want that to serve as a barrier
|
||||
elif isinstance(expr, ast.Assignment):
|
||||
left, right = expr.params
|
||||
stack.append((True, (canReplace, expr, right)))
|
||||
if isinstance(left, (ast.ArrayAccess, ast.FieldAccess)):
|
||||
for param in reversed(left.params):
|
||||
stack.append((True, (canReplace, left, param)))
|
||||
else:
|
||||
assert isinstance(left, ast.Local)
|
||||
else:
|
||||
for param in reversed(expr.params):
|
||||
stack.append((True, (canReplace, expr, param)))
|
||||
|
||||
if expr == old:
|
||||
if canReplace:
|
||||
if isinstance(parent, ast.JavaExpression):
|
||||
params = parent.params = list(parent.params)
|
||||
params[params.index(old)] = new
|
||||
else: # replacing in a top level statement
|
||||
assert parent.expr == old
|
||||
parent.expr = new
|
||||
return canReplace
|
||||
else:
|
||||
expr = args
|
||||
if hasSideEffects(expr):
|
||||
return False
|
||||
return False
|
||||
|
||||
def visitReplace(scope):
|
||||
newstatements = []
|
||||
for item in reversed(scope.statements):
|
||||
for sub in item.getScopes():
|
||||
visitReplace(sub)
|
||||
|
||||
if isinstance(item.expr, ast.Assignment) and item.expr.params[0] in replacevars:
|
||||
expr_roots = []
|
||||
for item2 in newstatements:
|
||||
# Don't inline into a while condition as it may be evaluated more than once
|
||||
if not isinstance(item2, ast.WhileStatement):
|
||||
expr_roots.append((item2, item2.expr))
|
||||
if item2.getScopes():
|
||||
break
|
||||
success = doReplacement(item, expr_roots)
|
||||
if success:
|
||||
continue
|
||||
newstatements.insert(0, item)
|
||||
scope.statements = newstatements
|
||||
visitReplace(root)
|
||||
|
||||
def _createDeclarations(root, predeclared):
|
||||
_setScopeParents(root)
|
||||
info = findVarDeclInfo(root, predeclared)
|
||||
localdefs = collections.defaultdict(list)
|
||||
newvars = [var for var in info if isinstance(var, ast.Local) and info[var].declScope is None]
|
||||
remaining = set(newvars)
|
||||
|
||||
def mdVisitVarUse(var):
|
||||
decl = ast.VariableDeclarator(ast.TypeName(var.dtype), var)
|
||||
# The compiler treats statements as if they can throw any exception at any time, so
|
||||
# it may think variables are not definitely assigned even when they really are.
|
||||
# Therefore, we give an unused initial value to every variable declaration
|
||||
# TODO - find a better way to handle this
|
||||
right = ast.dummyLiteral(var.dtype)
|
||||
localdefs[info[var].scope].append(ast.LocalDeclarationStatement(decl, right))
|
||||
remaining.remove(var)
|
||||
|
||||
def mdVisitScope(scope):
|
||||
if isinstance(scope, ast.StatementBlock):
|
||||
for i,stmt in enumerate(scope.statements):
|
||||
if isinstance(stmt, ast.ExpressionStatement):
|
||||
if isinstance(stmt.expr, ast.Assignment):
|
||||
var, right = stmt.expr.params
|
||||
if var in remaining and scope == info[var].scope:
|
||||
decl = ast.VariableDeclarator(ast.TypeName(var.dtype), var)
|
||||
new = ast.LocalDeclarationStatement(decl, right)
|
||||
scope.statements[i] = new
|
||||
remaining.remove(var)
|
||||
if stmt.expr is not None:
|
||||
top = stmt.expr
|
||||
for expr in top.postFlatIter():
|
||||
if expr in remaining:
|
||||
mdVisitVarUse(expr)
|
||||
for sub in stmt.getScopes():
|
||||
mdVisitScope(sub)
|
||||
|
||||
mdVisitScope(root)
|
||||
# print remaining
|
||||
assert not remaining
|
||||
assert None not in localdefs
|
||||
for scope, ldefs in localdefs.items():
|
||||
scope.statements = ldefs + scope.statements
|
||||
|
||||
def _createTernaries(scope, item):
|
||||
if isinstance(item, ast.IfStatement) and len(item.getScopes()) == 2:
|
||||
block1, block2 = item.getScopes()
|
||||
|
||||
if (len(block1.statements) == len(block2.statements) == 1) and block1.jumpKey == block2.jumpKey:
|
||||
s1, s2 = block1.statements[0], block2.statements[0]
|
||||
e1, e2 = s1.expr, s2.expr
|
||||
|
||||
if isinstance(s1, ast.ReturnStatement) and isinstance(s2, ast.ReturnStatement):
|
||||
expr = None if e1 is None else ast.Ternary(item.expr, e1, e2)
|
||||
item = ast.ReturnStatement(expr, s1.tt)
|
||||
elif isinstance(s1, ast.ExpressionStatement) and isinstance(s2, ast.ExpressionStatement):
|
||||
if isinstance(e1, ast.Assignment) and isinstance(e2, ast.Assignment):
|
||||
# if e1.params[0] == e2.params[0] and max(e1.params[1].complexity(), e2.params[1].complexity()) <= 1:
|
||||
if e1.params[0] == e2.params[0]:
|
||||
expr = ast.Ternary(item.expr, e1.params[1], e2.params[1])
|
||||
temp = ast.ExpressionStatement(ast.Assignment(e1.params[0], expr))
|
||||
|
||||
if not block1.doesFallthrough():
|
||||
assert not block2.doesFallthrough()
|
||||
item = ast.StatementBlock(item.func, item.continueKey, item.breakKey, [temp], block1.jumpKey)
|
||||
else:
|
||||
item = temp
|
||||
if item.expr is not None:
|
||||
item.expr = _simplifyExpressions(item.expr)
|
||||
return [item]
|
||||
|
||||
def _fixExprStatements(scope, item, namegen):
|
||||
if isinstance(item, ast.ExpressionStatement):
|
||||
right = item.expr
|
||||
if not isinstance(right, (ast.Assignment, ast.ClassInstanceCreation, ast.MethodInvocation)) and right.dtype is not None:
|
||||
left = ast.Local(right.dtype, lambda expr:namegen.getPrefix('dummy'))
|
||||
decl = ast.VariableDeclarator(ast.TypeName(left.dtype), left)
|
||||
item = ast.LocalDeclarationStatement(decl, right)
|
||||
return [item]
|
||||
|
||||
def _fixLiterals(scope, item):
|
||||
item.fixLiterals()
|
||||
|
||||
def _addCastsAndParens(scope, item, env):
|
||||
item.addCastsAndParens(env)
|
||||
|
||||
def _fallsThrough(scope, usedBreakTargets):
|
||||
# Check if control reaches end of scope and there is no break statement
|
||||
# We don't have to check keys since breaks should have already been generated for child scopes
|
||||
# for main scope there won't be one yet, but we don't care since we're just looking for
|
||||
# whether end of scope is reached on the main scope
|
||||
if not scope.statements:
|
||||
return True
|
||||
last = scope.statements[-1]
|
||||
if isinstance(last, (ast.JumpStatement, ast.ReturnStatement, ast.ThrowStatement)):
|
||||
return False
|
||||
elif not last.getScopes():
|
||||
return True
|
||||
# Scope ends with a complex statement. Determine whether it can fallthrough
|
||||
if last in usedBreakTargets:
|
||||
return True
|
||||
# Less strict than Java reachability rules, but we aren't bothering to follow them exactly
|
||||
if isinstance(last, ast.WhileStatement):
|
||||
return last.expr != ast.Literal.TRUE
|
||||
elif isinstance(last, ast.SwitchStatement):
|
||||
return not last.hasDefault() or _fallsThrough(last.getScopes()[-1], usedBreakTargets)
|
||||
else:
|
||||
if isinstance(last, ast.IfStatement) and len(last.getScopes()) < 2:
|
||||
return True
|
||||
return any(_fallsThrough(sub, usedBreakTargets) for sub in last.getScopes())
|
||||
|
||||
def _chooseJump(choices, breakPair, continuePair):
|
||||
assert None not in choices
|
||||
if breakPair in choices:
|
||||
return breakPair
|
||||
if continuePair in choices:
|
||||
return continuePair
|
||||
# Try to find an already labeled target
|
||||
for b, t in choices:
|
||||
if b.label is not None:
|
||||
return b, t
|
||||
return choices[0]
|
||||
|
||||
def _generateJumps(scope, usedBreakTargets, targets=collections.defaultdict(tuple), breakPair=None, continuePair=None, fallthroughs=NONE_SET, dryRun=False):
|
||||
assert None in fallthroughs
|
||||
newfallthroughs = fallthroughs
|
||||
newcontinuePair = continuePair
|
||||
newbreakPair = breakPair
|
||||
|
||||
if scope.jumpKey not in fallthroughs:
|
||||
newfallthroughs = frozenset([None, scope.jumpKey])
|
||||
|
||||
for item in reversed(scope.statements):
|
||||
if not item.getScopes():
|
||||
newfallthroughs = NONE_SET
|
||||
continue
|
||||
|
||||
if isinstance(item, ast.WhileStatement):
|
||||
newfallthroughs = frozenset([None, item.continueKey])
|
||||
else:
|
||||
newfallthroughs |= frozenset([item.breakKey])
|
||||
|
||||
newtargets = targets.copy()
|
||||
if isinstance(item, ast.WhileStatement):
|
||||
newtargets[item.continueKey] += ((item, True),)
|
||||
newcontinuePair = item, True
|
||||
if isinstance(item, (ast.WhileStatement, ast.SwitchStatement)):
|
||||
newbreakPair = item, False
|
||||
newtargets[item.breakKey] += ((item, False),)
|
||||
|
||||
for subscope in reversed(item.getScopes()):
|
||||
_generateJumps(subscope, usedBreakTargets, newtargets, newbreakPair, newcontinuePair, newfallthroughs, dryRun=dryRun)
|
||||
if isinstance(item, ast.SwitchStatement):
|
||||
newfallthroughs = frozenset([None, subscope.continueKey])
|
||||
newfallthroughs = frozenset([None, item.continueKey])
|
||||
|
||||
for item in scope.statements:
|
||||
if isinstance(item, ast.StatementBlock) and item.statements:
|
||||
if isinstance(item.statements[-1], ast.JumpStatement):
|
||||
assert item is scope.statements[-1] or item in usedBreakTargets
|
||||
|
||||
# Now that we've visited children, decide if we need a jump for this scope, and if so, generate it
|
||||
if scope.jumpKey not in fallthroughs:
|
||||
# Figure out if this jump is actually reachable
|
||||
if _fallsThrough(scope, usedBreakTargets):
|
||||
target, isContinue = pair = _chooseJump(targets[scope.jumpKey], breakPair, continuePair)
|
||||
if not isContinue:
|
||||
usedBreakTargets.add(target)
|
||||
if pair == breakPair or pair == continuePair:
|
||||
target = None
|
||||
# Now actually add the jump statement
|
||||
if not dryRun:
|
||||
scope.statements.append(ast.JumpStatement(target, isContinue))
|
||||
|
||||
def _pruneVoidReturn(scope):
|
||||
if scope.statements:
|
||||
last = scope.statements[-1]
|
||||
if isinstance(last, ast.ReturnStatement) and last.expr is None:
|
||||
scope.statements.pop()
|
||||
|
||||
def _pruneCtorSuper(scope):
|
||||
if scope.statements:
|
||||
stmt = scope.statements[0]
|
||||
if isinstance(stmt, ast.ExpressionStatement):
|
||||
expr = stmt.expr
|
||||
if isinstance(expr, ast.MethodInvocation) and len(expr.params) == 0 and expr.name == 'super':
|
||||
scope.statements = scope.statements[1:]
|
||||
|
||||
def generateAST(method, graph, forbidden_identifiers):
|
||||
env = method.class_.env
|
||||
namegen = NameGen(forbidden_identifiers)
|
||||
class_ = method.class_
|
||||
inputTypes = parseMethodDescriptor(method.descriptor, unsynthesize=False)[0]
|
||||
tts = objtypes.verifierToSynthetic_seq(inputTypes)
|
||||
|
||||
if graph is not None:
|
||||
graph.splitDualInedges()
|
||||
graph.fixLoops()
|
||||
entryNode, nodes = graphproxy.createGraphProxy(graph)
|
||||
if not method.static:
|
||||
entryNode.invars[0].name = 'this'
|
||||
|
||||
setree = structuring.structure(entryNode, nodes, (method.name == '<clinit>'))
|
||||
ast_root, varinfo = astgen.createAST(method, graph, setree, namegen)
|
||||
|
||||
argsources = [varinfo.var(entryNode, var) for var in entryNode.invars]
|
||||
disp_args = argsources if method.static else argsources[1:]
|
||||
for expr, tt in zip(disp_args, tts):
|
||||
expr.dtype = tt
|
||||
|
||||
decls = [ast.VariableDeclarator(ast.TypeName(expr.dtype), expr) for expr in disp_args]
|
||||
################################################################################################
|
||||
ast_root.bases = (ast_root,) # needed for our setScopeParents later
|
||||
|
||||
assert _generateJumps(ast_root, set(), dryRun=True) is None
|
||||
_preorder(ast_root, _fixObjectCreations)
|
||||
boolize.boolizeVars(ast_root, argsources)
|
||||
boolize.interfaceVars(env, ast_root, argsources)
|
||||
_simplifyBlocks(ast_root)
|
||||
assert _generateJumps(ast_root, set(), dryRun=True) is None
|
||||
|
||||
_mergeVariables(ast_root, argsources, method.static)
|
||||
|
||||
_preorder(ast_root, _createTernaries)
|
||||
_inlineVariables(ast_root)
|
||||
_simplifyBlocks(ast_root)
|
||||
_preorder(ast_root, _createTernaries)
|
||||
_inlineVariables(ast_root)
|
||||
_simplifyBlocks(ast_root)
|
||||
|
||||
_createDeclarations(ast_root, argsources)
|
||||
_preorder(ast_root, partial(_fixExprStatements, namegen=namegen))
|
||||
_preorder(ast_root, _fixLiterals)
|
||||
_preorder(ast_root, partial(_addCastsAndParens, env=env))
|
||||
_generateJumps(ast_root, set())
|
||||
_pruneVoidReturn(ast_root)
|
||||
_pruneCtorSuper(ast_root)
|
||||
else: # abstract or native method
|
||||
ast_root = None
|
||||
argsources = [ast.Local(tt, lambda expr:namegen.getPrefix('arg')) for tt in tts]
|
||||
decls = [ast.VariableDeclarator(ast.TypeName(expr.dtype), expr) for expr in argsources]
|
||||
|
||||
flags = method.flags - set(['BRIDGE','SYNTHETIC','VARARGS'])
|
||||
if method.name == '<init>': # More arbtirary restrictions. Yay!
|
||||
flags = flags - set(['ABSTRACT','STATIC','FINAL','NATIVE','STRICTFP','SYNCHRONIZED'])
|
||||
|
||||
flagstr = ' '.join(map(str.lower, sorted(flags)))
|
||||
inputTypes, returnTypes = parseMethodDescriptor(method.descriptor, unsynthesize=False)
|
||||
ret_tt = objtypes.verifierToSynthetic(returnTypes[0]) if returnTypes else objtypes.VoidTT
|
||||
return ast2.MethodDef(class_, flagstr, method.name, method.descriptor, ast.TypeName(ret_tt), decls, ast_root)
|
||||
|
|
@ -0,0 +1,285 @@
|
|||
import heapq
|
||||
|
||||
from .cfg import flattenDict, makeGraph
|
||||
|
||||
# Variables x and y can safely be merged when it is true that for any use of y (respectively x)
|
||||
# that sees a definition of y, either there are no intervening definitions of x, or x was known
|
||||
# to be equal to y *at the point of its most recent definition*
|
||||
# Given this info, we greedily merge related variables, that is, those where one is assigned to the other
|
||||
# to calculate which variables can be merged, we first have to build a CFG from the Java AST again
|
||||
|
||||
class VarInfo(object):
|
||||
__slots__ = "key", "defs", "rdefs", "extracount"
|
||||
def __init__(self, key):
|
||||
self.key = key
|
||||
self.defs = set()
|
||||
self.rdefs = set()
|
||||
self.extracount = 0
|
||||
|
||||
def priority(self):
|
||||
return (len(self.defs) + self.extracount), self.key
|
||||
|
||||
class EqualityData(object):
|
||||
def __init__(self, d=None):
|
||||
# Equal values point to a representative object instance. Singletons are not represented at all for efficiency
|
||||
# None represents the top value (i.e. this point has not been visited yet)
|
||||
self.d = d.copy() if d is not None else None
|
||||
|
||||
def _newval(self): return object()
|
||||
|
||||
def initialize(self): # initialize to bottom value (all variables unequal)
|
||||
assert self.d is None
|
||||
self.d = {}
|
||||
|
||||
def handleAssign(self, var1, var2=None):
|
||||
if var1 == var2:
|
||||
return
|
||||
if var2 is None:
|
||||
if var1 in self.d:
|
||||
del self.d[var1]
|
||||
else:
|
||||
self.d[var1] = self.d.setdefault(var2, self._newval())
|
||||
assert self.iseq(var1, var2)
|
||||
|
||||
def iseq(self, var1, var2):
|
||||
assert var1 != var2
|
||||
return var1 in self.d and var2 in self.d and self.d[var1] is self.d[var2]
|
||||
|
||||
def merge_update(self, other):
|
||||
if other.d is None:
|
||||
return
|
||||
elif self.d is None:
|
||||
self.d = other.d.copy()
|
||||
else:
|
||||
d1, d2 = self.d, other.d
|
||||
new = {}
|
||||
todo = list(set(d1) & set(d2))
|
||||
while todo:
|
||||
cur = todo.pop()
|
||||
matches = [k for k in todo if d1[k] is d1[cur] and d2[k] is d2[cur]]
|
||||
if not matches:
|
||||
continue
|
||||
new[cur] = self._newval()
|
||||
for k in matches:
|
||||
new[k] = new[cur]
|
||||
todo = [k for k in todo if k not in new]
|
||||
self.d = new
|
||||
|
||||
def copy(self): return EqualityData(self.d)
|
||||
|
||||
def __eq__(self, other):
|
||||
if self.d is None or other.d is None:
|
||||
return self.d is other.d
|
||||
if self.d == other.d:
|
||||
return True
|
||||
if set(self.d) != set(other.d):
|
||||
return False
|
||||
match = {}
|
||||
for k in self.d:
|
||||
if match.setdefault(self.d[k], other.d[k]) != other.d[k]:
|
||||
return False
|
||||
return True
|
||||
|
||||
def __ne__(self, other): return not self == other
|
||||
def __hash__(self): raise TypeError('unhashable type')
|
||||
|
||||
def calcEqualityData(graph):
|
||||
graph.simplify()
|
||||
blocks = graph.blocks
|
||||
d = {b:[EqualityData()] for b in blocks}
|
||||
|
||||
d[graph.entry][0].initialize()
|
||||
stack = [graph.entry]
|
||||
dirty = set(blocks)
|
||||
|
||||
while stack:
|
||||
block = stack.pop()
|
||||
if block not in dirty:
|
||||
continue
|
||||
dirty.remove(block)
|
||||
|
||||
cur = d[block][0].copy()
|
||||
e_out = EqualityData()
|
||||
del d[block][1:]
|
||||
|
||||
for line_t, data in block.lines:
|
||||
if line_t == 'def':
|
||||
cur.handleAssign(*data)
|
||||
d[block].append(cur.copy())
|
||||
elif line_t == 'canthrow':
|
||||
e_out.merge_update(cur)
|
||||
|
||||
for out, successors in [(e_out, block.e_successors), (cur, block.n_successors)]:
|
||||
stack += successors
|
||||
for suc in successors:
|
||||
old = d[suc][0].copy()
|
||||
d[suc][0].merge_update(out)
|
||||
if old != d[suc][0]:
|
||||
dirty.add(suc)
|
||||
|
||||
for block in blocks:
|
||||
assert d[block][0].d is not None
|
||||
assert not dirty
|
||||
return d
|
||||
|
||||
class VarMergeInfo(object):
|
||||
def __init__(self, graph, methodparams, isstatic):
|
||||
self.info = {}
|
||||
self.final, self.unmergeable, self.external = set(), set(), set()
|
||||
self.equality = None # to be calculated later
|
||||
self.graph = graph
|
||||
|
||||
self.pending_graph_replaces = {}
|
||||
self.touched_vars = set()
|
||||
|
||||
# initialize variables and assignment data
|
||||
for var in methodparams:
|
||||
self._addvar(var)
|
||||
self.external.update(methodparams)
|
||||
if not isstatic:
|
||||
self.final.add(methodparams[0])
|
||||
|
||||
for block in graph.blocks:
|
||||
for line_t, data in block.lines:
|
||||
if line_t == 'def':
|
||||
self._addassign(data[0], data[1])
|
||||
for caught in block.caught_excepts:
|
||||
self._addvar(caught)
|
||||
self.external.add(caught)
|
||||
self.unmergeable.add(caught)
|
||||
|
||||
# initialization helper funcs
|
||||
def _addvar(self, v):
|
||||
return self.info.setdefault(v, VarInfo(len(self.info)))
|
||||
|
||||
def _addassign(self, v1, v2):
|
||||
info = self._addvar(v1)
|
||||
if v2 is not None:
|
||||
info.defs.add(v2)
|
||||
self._addvar(v2).rdefs.add(v1)
|
||||
else:
|
||||
info.extracount += 1
|
||||
|
||||
# process helper funcs
|
||||
def iseq(self, block, index, v1, v2):
|
||||
return self.equality[block][index].iseq(v1, v2)
|
||||
|
||||
def _doGraphReplacements(self):
|
||||
self.graph.replace(self.pending_graph_replaces)
|
||||
self.pending_graph_replaces = {}
|
||||
self.touched_vars = set()
|
||||
|
||||
def compat(self, v1, v2, doeq):
|
||||
if v1 in self.touched_vars or v2 in self.touched_vars:
|
||||
self._doGraphReplacements()
|
||||
|
||||
blocks = self.graph.blocks
|
||||
vok = {b:3 for b in blocks} # use bitmask v1ok = 1<<0, v2ok = 1<<1
|
||||
|
||||
stack = [b for b in blocks if v1 in b.vars or v2 in b.vars]
|
||||
while stack:
|
||||
block = stack.pop()
|
||||
cur = vok[block]
|
||||
e_out = 3
|
||||
|
||||
if v1 in block.vars or v2 in block.vars:
|
||||
defcount = 0
|
||||
for line_t, data in block.lines:
|
||||
if line_t == 'use':
|
||||
if (data == v1 and not cur & 1) or (data == v2 and not cur & 2):
|
||||
return False
|
||||
elif line_t == 'def':
|
||||
defcount += 1
|
||||
|
||||
if data[0] == v1 and data[1] != v1:
|
||||
cur = 1
|
||||
elif data[0] == v2 and data[1] != v2:
|
||||
cur = 2
|
||||
if doeq and self.iseq(block, defcount, v1, v2):
|
||||
cur = 3
|
||||
elif line_t == 'canthrow':
|
||||
e_out &= cur
|
||||
else:
|
||||
# v1 and v2 not touched in this block, so there is nothing to do
|
||||
e_out = cur
|
||||
|
||||
for out, successors in [(e_out, block.e_successors), (cur, block.n_successors)]:
|
||||
for suc in successors:
|
||||
if vok[suc] & out != vok[suc]:
|
||||
stack.append(suc)
|
||||
vok[suc] &= out
|
||||
return True
|
||||
|
||||
def process(self, replace, doeq):
|
||||
final, unmergeable, external = self.final, self.unmergeable, self.external
|
||||
d = self.info
|
||||
work_q = [(info.priority(), var) for var, info in d.items()]
|
||||
heapq.heapify(work_q)
|
||||
dirty = set(d) - external
|
||||
|
||||
while work_q:
|
||||
_, cur = heapq.heappop(work_q)
|
||||
if (cur in external) or cur not in dirty:
|
||||
continue
|
||||
dirty.remove(cur)
|
||||
|
||||
candidate_set = d[cur].defs - unmergeable
|
||||
if len(d[cur].defs) > 1 or d[cur].extracount > 0:
|
||||
candidate_set = candidate_set - final
|
||||
candidates = [v for v in candidate_set if v.dtype == cur.dtype]
|
||||
candidates = sorted(candidates, key=lambda v:d[v].key)
|
||||
assert cur not in candidates
|
||||
|
||||
# find first candidate that is actually compatible
|
||||
for parent in candidates:
|
||||
if self.compat(cur, parent, doeq):
|
||||
break
|
||||
else:
|
||||
continue # no candidates found
|
||||
|
||||
replace[cur] = parent
|
||||
self.pending_graph_replaces[cur] = parent
|
||||
self.touched_vars.add(cur)
|
||||
self.touched_vars.add(parent)
|
||||
|
||||
infc, infp = d[cur], d[parent]
|
||||
# Be careful, there could be a loop with cur in parent.defs
|
||||
infc.defs.remove(parent)
|
||||
infc.rdefs.discard(parent)
|
||||
infp.rdefs.remove(cur)
|
||||
infp.defs.discard(cur)
|
||||
|
||||
for var in d[cur].rdefs:
|
||||
d[var].defs.remove(cur)
|
||||
d[var].defs.add(parent)
|
||||
heapq.heappush(work_q, (d[var].priority(), var))
|
||||
|
||||
for var in d[cur].defs:
|
||||
d[var].rdefs.remove(cur)
|
||||
d[var].rdefs.add(parent)
|
||||
|
||||
d[parent].defs |= d[cur].defs
|
||||
d[parent].rdefs |= d[cur].rdefs
|
||||
d[parent].extracount += d[cur].extracount
|
||||
|
||||
del d[cur]
|
||||
heapq.heappush(work_q, (d[parent].priority(), parent))
|
||||
dirty.add(parent)
|
||||
|
||||
def processMain(self, replace):
|
||||
self.process(replace, False)
|
||||
self._doGraphReplacements()
|
||||
self.equality = calcEqualityData(self.graph)
|
||||
self.process(replace, True)
|
||||
|
||||
###############################################################################
|
||||
def mergeVariables(root, isstatic, parameters):
|
||||
# first, create CFG from the Java AST
|
||||
graph = makeGraph(root)
|
||||
mergeinfo = VarMergeInfo(graph, parameters, isstatic)
|
||||
|
||||
replace = {}
|
||||
mergeinfo.processMain(replace)
|
||||
|
||||
flattenDict(replace)
|
||||
return replace
|
||||
|
|
@ -0,0 +1,55 @@
|
|||
reserved_identifiers = '''
|
||||
abstract
|
||||
assert
|
||||
boolean
|
||||
break
|
||||
byte
|
||||
case
|
||||
catch
|
||||
char
|
||||
class
|
||||
const
|
||||
continue
|
||||
default
|
||||
do
|
||||
double
|
||||
else
|
||||
enum
|
||||
extends
|
||||
false
|
||||
final
|
||||
finally
|
||||
float
|
||||
for
|
||||
goto
|
||||
if
|
||||
implements
|
||||
import
|
||||
instanceof
|
||||
int
|
||||
interface
|
||||
long
|
||||
native
|
||||
new
|
||||
null
|
||||
package
|
||||
private
|
||||
protected
|
||||
public
|
||||
return
|
||||
short
|
||||
static
|
||||
strictfp
|
||||
super
|
||||
switch
|
||||
synchronized
|
||||
this
|
||||
throw
|
||||
throws
|
||||
transient
|
||||
true
|
||||
try
|
||||
void
|
||||
volatile
|
||||
while
|
||||
'''.split()
|
||||
|
|
@ -0,0 +1,63 @@
|
|||
import itertools
|
||||
|
||||
def update(self, items):
|
||||
self.entryBlock = items[0].entryBlock
|
||||
self.nodes = frozenset.union(*(i.nodes for i in items))
|
||||
temp = set(self.nodes)
|
||||
siter = itertools.chain.from_iterable(i.successors for i in items)
|
||||
self.successors = [n for n in siter if not n in temp and not temp.add(n)]
|
||||
|
||||
class SEBlockItem(object):
|
||||
def __init__(self, node):
|
||||
self.successors = node.norm_suc_nl # don't include backedges or exceptional edges
|
||||
self.node = node
|
||||
self.nodes = frozenset([node])
|
||||
self.entryBlock = node
|
||||
|
||||
def getScopes(self): return ()
|
||||
|
||||
class SEScope(object):
|
||||
def __init__(self, items):
|
||||
self.items = items
|
||||
update(self, items)
|
||||
|
||||
def getScopes(self): return ()
|
||||
|
||||
class SEWhile(object):
|
||||
def __init__(self, scope):
|
||||
self.body = scope
|
||||
update(self, [scope])
|
||||
|
||||
def getScopes(self): return self.body,
|
||||
|
||||
class SETry(object):
|
||||
def __init__(self, tryscope, catchscope, toptts, catchvar):
|
||||
self.scopes = tryscope, catchscope
|
||||
self.toptts = toptts
|
||||
self.catchvar = catchvar # none if ignored
|
||||
update(self, self.scopes)
|
||||
|
||||
def getScopes(self): return self.scopes
|
||||
|
||||
class SEIf(object):
|
||||
def __init__(self, head, newscopes):
|
||||
assert len(newscopes) == 2
|
||||
self.scopes = newscopes
|
||||
self.head = head
|
||||
update(self, [head] + newscopes)
|
||||
|
||||
def getScopes(self): return self.scopes
|
||||
|
||||
class SESwitch(object):
|
||||
def __init__(self, head, newscopes):
|
||||
self.scopes = newscopes
|
||||
self.head = head
|
||||
self.ordered = newscopes
|
||||
update(self, [head] + newscopes)
|
||||
|
||||
jump = head.node.block.jump
|
||||
keysets = {head.node.blockdict[b.key,False]:jump.reverse.get(b) for b in jump.getNormalSuccessors()}
|
||||
assert keysets.values().count(None) == 1
|
||||
self.ordered_keysets = [keysets[item.entryBlock] for item in newscopes]
|
||||
|
||||
def getScopes(self): return self.scopes
|
||||
|
|
@ -0,0 +1,27 @@
|
|||
# double quote, backslash, and newlines are forbidden
|
||||
ok_chars = " !#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[]^_`abcdefghijklmnopqrstuvwxyz{|}~"
|
||||
ok_chars = frozenset(ok_chars)
|
||||
|
||||
# these characters cannot use unicode escape codes due to the way Java escaping works
|
||||
late_escape = {u'\u0009':r'\t', u'\u000a':r'\n', u'\u000d':r'\r', u'\u0022':r'\"', u'\u005c':r'\\'}
|
||||
|
||||
def escapeString(u):
|
||||
if set(u) <= ok_chars:
|
||||
return u
|
||||
|
||||
escaped = []
|
||||
for c in u:
|
||||
if c in ok_chars:
|
||||
escaped.append(c)
|
||||
elif c in late_escape:
|
||||
escaped.append(late_escape[c])
|
||||
else:
|
||||
i = ord(c)
|
||||
if i <= 0xFFFF:
|
||||
escaped.append(r'\u{0:04x}'.format(i))
|
||||
else:
|
||||
i -= 0x10000
|
||||
high = 0xD800 + (i>>10)
|
||||
low = 0xDC00 + (i & 0x3FF)
|
||||
escaped.append(r'\u{0:04x}\u{1:04x}'.format(high,low))
|
||||
return ''.join(escaped)
|
||||
File diff suppressed because it is too large
Load Diff
|
|
@ -0,0 +1,48 @@
|
|||
from ..ssa import objtypes
|
||||
|
||||
from . import ast
|
||||
|
||||
# A simple throws declaration inferrer that only considers throw statements within the method
|
||||
# this is mostly just useful to make sure the ExceptionHandlers test compiles
|
||||
|
||||
def _visit_statement(env, stmt):
|
||||
if isinstance(stmt, ast.ThrowStatement):
|
||||
return stmt.expr.dtype
|
||||
|
||||
result = objtypes.NullTT
|
||||
if isinstance(stmt, ast.TryStatement):
|
||||
caught_types = []
|
||||
for catch, b in stmt.pairs:
|
||||
caught_types.extend(objtypes.className(tn.tt) for tn in catch.typename.tnames)
|
||||
if objtypes.ThrowableTT not in caught_types:
|
||||
temp = _visit_statement(env, stmt.tryb)
|
||||
if temp != objtypes.NullTT:
|
||||
assert objtypes.dim(temp) == 0
|
||||
name = objtypes.className(temp)
|
||||
if not any(env.isSubclass(name, caught) for caught in caught_types):
|
||||
result = temp
|
||||
|
||||
statements = zip(*stmt.pairs)[1]
|
||||
elif isinstance(stmt, ast.StatementBlock):
|
||||
statements = stmt.statements
|
||||
else:
|
||||
statements = stmt.getScopes()
|
||||
|
||||
for sub in statements:
|
||||
if result == objtypes.ThrowableTT:
|
||||
break
|
||||
result = objtypes.commonSupertype(env, [result, _visit_statement(env, sub)])
|
||||
|
||||
if result != objtypes.NullTT:
|
||||
if env.isSubclass(objtypes.className(result), 'java/lang/RuntimeException'):
|
||||
return objtypes.NullTT
|
||||
return result
|
||||
|
||||
def addSingle(env, meth_asts):
|
||||
for meth in meth_asts:
|
||||
if not meth.body:
|
||||
continue
|
||||
tt = _visit_statement(env, meth.body)
|
||||
assert objtypes.commonSupertype(env, [tt, objtypes.ThrowableTT]) == objtypes.ThrowableTT
|
||||
if tt != objtypes.NullTT:
|
||||
meth.throws = ast.TypeName(tt)
|
||||
|
|
@ -0,0 +1,14 @@
|
|||
# Override this to rename classes
|
||||
class DefaultVisitor(object):
|
||||
def visit(self, obj):
|
||||
return obj.print_(self, self.visit)
|
||||
|
||||
# Experimental - don't use!
|
||||
def toTree(self, obj):
|
||||
if obj is None:
|
||||
return None
|
||||
return obj.tree(self, self.toTree)
|
||||
|
||||
def className(self, name): return name
|
||||
def methodName(self, cls, name, desc): return name
|
||||
def fieldName(self, cls, name, desc): return name
|
||||
|
|
@ -0,0 +1,115 @@
|
|||
import collections
|
||||
|
||||
from . import bytecode
|
||||
from .attributes_raw import fixAttributeNames, get_attributes_raw
|
||||
from .classfileformat.reader import Reader
|
||||
|
||||
exceptionHandlerRaw = collections.namedtuple("exceptionHandlerRaw",
|
||||
["start","end","handler","type_ind"])
|
||||
|
||||
class Code(object):
|
||||
def __init__(self, method, bytestream, keepRaw):
|
||||
self.method = method
|
||||
self.class_ = method.class_
|
||||
|
||||
# Old versions use shorter fields for stack, locals, and code length
|
||||
field_fmt = ">HHL" if self.class_.version > (45,2) else ">BBH"
|
||||
self.stack, self.locals, codelen = bytestream.get(field_fmt)
|
||||
# assert codelen > 0 and codelen < 65536
|
||||
self.bytecode_raw = bytestream.getRaw(codelen)
|
||||
self.codelen = codelen
|
||||
|
||||
except_cnt = bytestream.get('>H')
|
||||
self.except_raw = [bytestream.get('>HHHH') for _ in range(except_cnt)]
|
||||
self.except_raw = [exceptionHandlerRaw(*t) for t in self.except_raw]
|
||||
attributes_raw = get_attributes_raw(bytestream)
|
||||
assert bytestream.size() == 0
|
||||
|
||||
if self.except_raw:
|
||||
assert self.stack >= 1
|
||||
|
||||
# print 'Parsing code for', method.name, method.descriptor, method.flags
|
||||
codestream = Reader(data=self.bytecode_raw)
|
||||
self.bytecode = bytecode.parseInstructions(codestream, self.isIdConstructor)
|
||||
self.attributes = fixAttributeNames(attributes_raw, self.class_.cpool)
|
||||
|
||||
for e in self.except_raw:
|
||||
assert e.start in self.bytecode
|
||||
assert e.end == codelen or e.end in self.bytecode
|
||||
assert e.handler in self.bytecode
|
||||
if keepRaw:
|
||||
self.attributes_raw = attributes_raw
|
||||
|
||||
# This is a callback passed to the bytecode parser to determine if a given method id represents a constructor
|
||||
def isIdConstructor(self, methId):
|
||||
args = self.class_.cpool.getArgsCheck('Method', methId)
|
||||
return args[1] == '<init>'
|
||||
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
lines = ['Stack: {}, Locals {}'.format(self.stack, self.locals)]
|
||||
|
||||
instructions = self.bytecode
|
||||
lines += ['{}: {}'.format(i, bytecode.printInstruction(instructions[i])) for i in sorted(instructions)]
|
||||
if self.except_raw:
|
||||
lines += ['Exception Handlers:']
|
||||
lines += map(str, self.except_raw)
|
||||
return '\n'.join(lines)
|
||||
|
||||
class Method(object):
|
||||
flagVals = {'PUBLIC':0x0001,
|
||||
'PRIVATE':0x0002,
|
||||
'PROTECTED':0x0004,
|
||||
'STATIC':0x0008,
|
||||
'FINAL':0x0010,
|
||||
'SYNCHRONIZED':0x0020,
|
||||
'BRIDGE':0x0040,
|
||||
'VARARGS':0x0080,
|
||||
'NATIVE':0x0100,
|
||||
'ABSTRACT':0x0400,
|
||||
'STRICTFP':0x0800,
|
||||
'SYNTHETIC':0x1000,
|
||||
}
|
||||
|
||||
def __init__(self, data, classFile, keepRaw):
|
||||
self.class_ = classFile
|
||||
cpool = self.class_.cpool
|
||||
|
||||
flags, name_id, desc_id, attributes_raw = data
|
||||
|
||||
self.name = cpool.getArgsCheck('Utf8', name_id)
|
||||
self.descriptor = cpool.getArgsCheck('Utf8', desc_id)
|
||||
# print 'Loading method ', self.name, self.descriptor
|
||||
self.attributes = fixAttributeNames(attributes_raw, cpool)
|
||||
|
||||
self.flags = set(name for name, mask in Method.flagVals.items() if (mask & flags))
|
||||
# Flags are ignored for <clinit>?
|
||||
if self.name == '<clinit>':
|
||||
self.flags = set(['STATIC'])
|
||||
|
||||
|
||||
self._checkFlags()
|
||||
self.static = 'STATIC' in self.flags
|
||||
self.native = 'NATIVE' in self.flags
|
||||
self.abstract = 'ABSTRACT' in self.flags
|
||||
self.isConstructor = (self.name == '<init>')
|
||||
|
||||
self.code = self._loadCode(keepRaw)
|
||||
if keepRaw:
|
||||
self.attributes_raw = attributes_raw
|
||||
self.name_id, self.desc_id = name_id, desc_id
|
||||
|
||||
def _checkFlags(self):
|
||||
assert len(self.flags & set(('PRIVATE','PROTECTED','PUBLIC'))) <= 1
|
||||
if 'ABSTRACT' in self.flags:
|
||||
assert not self.flags & set(['SYNCHRONIZED', 'PRIVATE', 'FINAL', 'STRICT', 'STATIC', 'NATIVE'])
|
||||
|
||||
def _loadCode(self, keepRaw):
|
||||
code_attrs = [a for a in self.attributes if a[0] == 'Code']
|
||||
if not (self.native or self.abstract):
|
||||
assert len(code_attrs) == 1
|
||||
code_raw = code_attrs[0][1]
|
||||
bytestream = Reader(code_raw)
|
||||
return Code(self, bytestream, keepRaw)
|
||||
assert not code_attrs
|
||||
return None
|
||||
|
|
@ -0,0 +1,18 @@
|
|||
import collections
|
||||
import itertools
|
||||
|
||||
class NameGen(object):
|
||||
def __init__(self, reserved=frozenset()):
|
||||
self.counters = collections.defaultdict(itertools.count)
|
||||
self.names = set(reserved)
|
||||
|
||||
def getPrefix(self, prefix, sep=''):
|
||||
newname = prefix
|
||||
while newname in self.names:
|
||||
newname = prefix + sep + str(next(self.counters[prefix]))
|
||||
self.names.add(newname)
|
||||
return newname
|
||||
|
||||
def LabelGen(prefix='label'):
|
||||
for i in itertools.count():
|
||||
yield prefix + str(i)
|
||||
|
|
@ -0,0 +1,65 @@
|
|||
ADD = 'add'
|
||||
AND = 'and'
|
||||
ANEWARRAY = 'anewarray'
|
||||
ARRLEN = 'arrlen'
|
||||
ARRLOAD = 'arrload'
|
||||
ARRLOAD_OBJ = 'arrload_obj'
|
||||
ARRSTORE = 'arrstore'
|
||||
ARRSTORE_OBJ = 'arrstore_obj'
|
||||
CHECKCAST = 'checkcast'
|
||||
CONST = 'const'
|
||||
CONSTNULL = 'constnull'
|
||||
CONVERT = 'convert'
|
||||
DIV = 'div'
|
||||
DUP = 'dup'
|
||||
DUP2 = 'dup2'
|
||||
DUP2X1 = 'dup2x1'
|
||||
DUP2X2 = 'dup2x2'
|
||||
DUPX1 = 'dupx1'
|
||||
DUPX2 = 'dupx2'
|
||||
FCMP = 'fcmp'
|
||||
GETFIELD = 'getfield'
|
||||
GETSTATIC = 'getstatic'
|
||||
GOTO = 'goto'
|
||||
IF_A = 'if_a'
|
||||
IF_ACMP = 'if_acmp'
|
||||
IF_I = 'if_i'
|
||||
IF_ICMP = 'if_icmp'
|
||||
IINC = 'iinc'
|
||||
INSTANCEOF = 'instanceof'
|
||||
INVOKEDYNAMIC = 'invokedynamic'
|
||||
INVOKEINIT = 'invokeinit'
|
||||
INVOKEINTERFACE = 'invokeinterface'
|
||||
INVOKESPECIAL = 'invokespecial'
|
||||
INVOKESTATIC = 'invokestatic'
|
||||
INVOKEVIRTUAL = 'invokevirtual'
|
||||
JSR = 'jsr'
|
||||
LCMP = 'lcmp'
|
||||
LDC = 'ldc'
|
||||
LOAD = 'load'
|
||||
MONENTER = 'monenter'
|
||||
MONEXIT = 'monexit'
|
||||
MUL = 'mul'
|
||||
MULTINEWARRAY = 'multinewarray'
|
||||
NEG = 'neg'
|
||||
NEW = 'new'
|
||||
NEWARRAY = 'newarray'
|
||||
NOP = 'nop'
|
||||
OR = 'or'
|
||||
POP = 'pop'
|
||||
POP2 = 'pop2'
|
||||
PUTFIELD = 'putfield'
|
||||
PUTSTATIC = 'putstatic'
|
||||
REM = 'rem'
|
||||
RET = 'ret'
|
||||
RETURN = 'return'
|
||||
SHL = 'shl'
|
||||
SHR = 'shr'
|
||||
STORE = 'store'
|
||||
SUB = 'sub'
|
||||
SWAP = 'swap'
|
||||
SWITCH = 'switch'
|
||||
THROW = 'throw'
|
||||
TRUNCATE = 'truncate'
|
||||
USHR = 'ushr'
|
||||
XOR = 'xor'
|
||||
|
|
@ -0,0 +1,191 @@
|
|||
from __future__ import print_function
|
||||
|
||||
import collections
|
||||
import errno
|
||||
from functools import partial
|
||||
import hashlib
|
||||
import os
|
||||
import os.path
|
||||
import platform
|
||||
import zipfile
|
||||
|
||||
# Various utility functions for the top level scripts (decompile.py, assemble.py, disassemble.py)
|
||||
|
||||
copyright = '''Krakatau Copyright (C) 2012-18 Robert Grosse
|
||||
This program is provided as open source under the GNU General Public License.
|
||||
See LICENSE.TXT for more details.
|
||||
'''
|
||||
|
||||
_osname = platform.system().lower()
|
||||
IS_WINDOWS = 'win' in _osname and 'darwin' not in _osname and 'cygwin' not in _osname
|
||||
|
||||
def findFiles(target, recursive, prefix):
|
||||
if target.endswith('.jar'):
|
||||
with zipfile.ZipFile(target, 'r') as archive:
|
||||
return [name.encode('utf8') for name in archive.namelist() if name.endswith(prefix)]
|
||||
else:
|
||||
if recursive:
|
||||
assert os.path.isdir(target)
|
||||
targets = []
|
||||
|
||||
for root, dirs, files in os.walk(target):
|
||||
targets += [os.path.join(root, fname) for fname in files if fname.endswith(prefix)]
|
||||
return targets
|
||||
else:
|
||||
return [target]
|
||||
|
||||
def normalizeClassname(name):
|
||||
if name.endswith('.class'):
|
||||
name = name[:-6]
|
||||
# Replacing backslashes is ugly since they can be in valid classnames too, but this seems the best option
|
||||
return name.replace('\\','/').replace('.','/')
|
||||
|
||||
# Windows stuff
|
||||
illegal_win_chars = frozenset('<>;:|?*\\/"%')
|
||||
pref_disp_chars = frozenset('abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_$0123456789')
|
||||
|
||||
# Prevent creating filename parts matching the legacy device filenames. While Krakatau can create these files
|
||||
# just fine thanks to using \\?\ paths, the resulting files are impossible to open or delete in Windows Explorer
|
||||
# or with similar tools, so they are a huge pain to deal with. Therefore, we don't generate them at all.
|
||||
illegal_parts = frozenset(['CON', 'PRN', 'AUX', 'NUL', 'COM1', 'COM2', 'COM3', 'COM4', 'COM5', 'COM6', 'COM7', 'COM8',
|
||||
'COM9', 'LPT1', 'LPT2', 'LPT3', 'LPT4', 'LPT5', 'LPT6', 'LPT7', 'LPT8', 'LPT9'])
|
||||
|
||||
class PathSanitizer(object):
|
||||
def __init__(self, base, suffix):
|
||||
self.base = base
|
||||
self.suffix = suffix
|
||||
|
||||
def is_part_ok(self, s, parents):
|
||||
if not 1 <= len(s) <= self.MAX_PART_LEN:
|
||||
return False
|
||||
# avoid potential collision with hashed parts
|
||||
if len(s) >= 66 and '__' in s:
|
||||
return False
|
||||
# . cannot appear in a valid class name, but might as well specifically exclude these, just in case
|
||||
if s.startswith('.') or '..' in s:
|
||||
return False
|
||||
return '\x1f' < min(s) <= max(s) < '\x7f'
|
||||
|
||||
def hash(self, s, suffix):
|
||||
left = ''.join(c for c in s if c in pref_disp_chars)
|
||||
right = '__' + hashlib.sha256(s.encode('utf8')).hexdigest() + suffix
|
||||
return left[:self.MAX_PART_LEN - len(right)] + right
|
||||
|
||||
def sanitize(self, path):
|
||||
if isinstance(path, bytes):
|
||||
path = path.decode()
|
||||
|
||||
oldparts = path.split('/')
|
||||
newparts = []
|
||||
for i, part in enumerate(oldparts):
|
||||
suffix = self.suffix if i + 1 == len(oldparts) else ''
|
||||
if self.is_part_ok(part + suffix, newparts):
|
||||
newparts.append(part + suffix)
|
||||
else:
|
||||
newparts.append(self.hash(part, suffix))
|
||||
|
||||
result = self.format_path([self.base] + newparts)
|
||||
if len(result) > self.MAX_PATH_LEN:
|
||||
result = self.format_path([self.base, self.hash(path, self.suffix)])
|
||||
assert result.endswith(self.suffix)
|
||||
return result
|
||||
|
||||
class LinuxPathSanitizer(PathSanitizer):
|
||||
MAX_PART_LEN = 255
|
||||
MAX_PATH_LEN = 4095
|
||||
|
||||
def __init__(self, *args):
|
||||
PathSanitizer.__init__(self, *args)
|
||||
|
||||
def format_path(self, parts):
|
||||
return os.path.join(*parts)
|
||||
|
||||
class WindowsPathSanitizer(PathSanitizer):
|
||||
MAX_PART_LEN = 255
|
||||
MAX_PATH_LEN = 32000 # close enough
|
||||
|
||||
def __init__(self, *args):
|
||||
PathSanitizer.__init__(self, *args)
|
||||
# keep track of previous paths to detect case-insensitive collisions
|
||||
self.prevs = collections.defaultdict(dict)
|
||||
|
||||
def is_part_ok(self, s, parents):
|
||||
if not PathSanitizer.is_part_ok(self, s, parents):
|
||||
return False
|
||||
if s.upper() in illegal_parts:
|
||||
return False
|
||||
# make sure nothing in the current directory is a case insensitive collision
|
||||
if self.prevs[tuple(parents)].setdefault(s.lower(), s) != s:
|
||||
return False
|
||||
return illegal_win_chars.isdisjoint(s)
|
||||
|
||||
def format_path(self, parts):
|
||||
return '\\\\?\\' + '\\'.join(parts)
|
||||
|
||||
class DirectoryWriter(object):
|
||||
def __init__(self, base_path, suffix):
|
||||
if base_path is None:
|
||||
base_path = os.getcwd()
|
||||
else:
|
||||
if not isinstance(base_path, str):
|
||||
base_path = base_path.decode('utf8')
|
||||
base_path = os.path.abspath(base_path)
|
||||
|
||||
if IS_WINDOWS:
|
||||
self.makepath = WindowsPathSanitizer(base_path, suffix).sanitize
|
||||
else:
|
||||
self.makepath = LinuxPathSanitizer(base_path, suffix).sanitize
|
||||
|
||||
def write(self, cname, data):
|
||||
out = self.makepath(cname)
|
||||
dirpath = os.path.dirname(out)
|
||||
|
||||
try:
|
||||
if dirpath:
|
||||
os.makedirs(dirpath)
|
||||
except OSError as exc:
|
||||
if exc.errno != errno.EEXIST:
|
||||
raise
|
||||
|
||||
mode = 'wb' if isinstance(data, bytes) else 'w'
|
||||
with open(out, mode) as f:
|
||||
f.write(data)
|
||||
return out
|
||||
|
||||
def __enter__(self): return self
|
||||
def __exit__(self, *args): pass
|
||||
|
||||
class JarWriter(object):
|
||||
def __init__(self, base_path, suffix):
|
||||
self.zip = zipfile.ZipFile(base_path, mode='w')
|
||||
self.suffix = suffix
|
||||
|
||||
def write(self, cname, data):
|
||||
info = zipfile.ZipInfo(cname + self.suffix, (1980, 1, 1, 0, 0, 0))
|
||||
self.zip.writestr(info, data)
|
||||
return 'zipfile'
|
||||
|
||||
def __enter__(self): self.zip.__enter__(); return self
|
||||
def __exit__(self, *args): self.zip.__exit__(*args)
|
||||
|
||||
class MockWriter(object):
|
||||
def __init__(self): self.results = []
|
||||
def write(self, cname, data): self.results.append((cname, data))
|
||||
def __enter__(self): return self
|
||||
def __exit__(self, *args): pass
|
||||
|
||||
def makeWriter(base_path, suffix):
|
||||
if base_path is not None:
|
||||
if base_path.endswith('.zip') or base_path.endswith('.jar'):
|
||||
return JarWriter(base_path, suffix)
|
||||
return DirectoryWriter(base_path, suffix)
|
||||
|
||||
###############################################################################
|
||||
def ignore(*args, **kwargs):
|
||||
pass
|
||||
|
||||
class Logger(object):
|
||||
def __init__(self, level):
|
||||
lvl = ['info', 'warning'].index(level)
|
||||
self.info = print if lvl <= 0 else ignore
|
||||
self.warn = print if lvl <= 1 else ignore
|
||||
|
|
@ -0,0 +1 @@
|
|||
from .graph import ssaFromVerified
|
||||
|
|
@ -0,0 +1,360 @@
|
|||
import collections
|
||||
import operator
|
||||
|
||||
from .. import opnames as vops
|
||||
from ..verifier import verifier_types
|
||||
|
||||
from . import ssa_jumps, ssa_ops, subproc
|
||||
from .blockmakerfuncs import ResultDict, instructionHandlers
|
||||
from .ssa_types import BasicBlock, SSA_OBJECT, slots_t
|
||||
|
||||
def toBits(x): return [i for i in range(x.bit_length()) if x & (1 << i)]
|
||||
|
||||
# keys for special blocks created at the cfg entry and exit. Negative keys ensures they don't collide
|
||||
ENTRY_KEY, RETURN_KEY, RETHROW_KEY = -1, -2, -3
|
||||
|
||||
def getUsedLocals(iNodes, iNodeD, exceptions):
|
||||
# For every instruction, find which locals at that point may be used in the future
|
||||
except_ranges = [(h, [node.key for node in iNodes if s <= node.key < e]) for s, e, h, i in exceptions]
|
||||
|
||||
old = collections.defaultdict(int)
|
||||
while 1:
|
||||
data = old.copy()
|
||||
# Do one iteration
|
||||
for node in reversed(iNodes):
|
||||
used = reduce(operator.__or__, (data[key] for key in node.successors), 0)
|
||||
|
||||
if node.instruction[0] == vops.LOAD:
|
||||
used |= 1 << node.instruction[2]
|
||||
elif node.instruction[0] == vops.IINC:
|
||||
used |= 1 << node.instruction[1]
|
||||
elif node.instruction[0] == vops.STORE:
|
||||
bits = 3 if node.instruction[1] in 'JD' else 1
|
||||
used &= ~(bits << node.instruction[2])
|
||||
elif node.instruction[0] == vops.RET:
|
||||
# If local is not in mask, it will use the value from the jsr instead of the ret
|
||||
mask = sum(1<<i for i in node.out_state.copy().maskFor(node.jsrTarget))
|
||||
used &= mask
|
||||
elif node.instruction[0] == vops.JSR and node.returnedFrom is not None:
|
||||
retnode = iNodeD[node.returnedFrom]
|
||||
assert node.successors == (retnode.jsrTarget,)
|
||||
mask = sum(1<<i for i in iNodeD[node.returnedFrom].out_state.copy().maskFor(retnode.jsrTarget))
|
||||
|
||||
assert node.next_instruction is not None
|
||||
used |= (data[node.next_instruction] & ~mask)
|
||||
data[node.key] |= used
|
||||
|
||||
for hkey, region in except_ranges:
|
||||
if data[hkey] != old[hkey]:
|
||||
for key in region:
|
||||
data[key] |= data[hkey]
|
||||
|
||||
if data == old:
|
||||
break
|
||||
old = data
|
||||
# for entry point, every program argument is marked used so we can preserve input arguments for later
|
||||
old[ENTRY_KEY] = (1 << len(iNodeD[0].state.locals)) - 1
|
||||
return old
|
||||
|
||||
def slotsRvals(inslots):
|
||||
stack = [(None if phi is None else phi.rval) for phi in inslots.stack]
|
||||
newlocals = {i: phi.rval for i, phi in inslots.locals.items() if phi is not None}
|
||||
return slots_t(stack=stack, locals=newlocals)
|
||||
|
||||
_jump_instrs = frozenset([vops.GOTO, vops.IF_A, vops.IF_ACMP, vops.IF_I, vops.IF_ICMP, vops.JSR, vops.SWITCH])
|
||||
class BlockMaker(object):
|
||||
def __init__(self, parent, iNodes, inputTypes, returnTypes, except_raw, opts):
|
||||
self.parent = parent
|
||||
self.blocks = []
|
||||
self.blockd = {}
|
||||
|
||||
self.iNodes = [n for n in iNodes if n.visited]
|
||||
self.iNodeD = {n.key: n for n in self.iNodes}
|
||||
exceptions = [eh for eh in except_raw if eh.handler in self.iNodeD]
|
||||
|
||||
# Calculate which locals are actually live at any point
|
||||
self.used_locals = getUsedLocals(self.iNodes, self.iNodeD, exceptions)
|
||||
|
||||
# create map of uninitialized -> initialized types so we can convert them
|
||||
self.initMap = {}
|
||||
for node in self.iNodes:
|
||||
if node.op == vops.NEW:
|
||||
self.initMap[node.stack_push[0]] = node.target_type
|
||||
self.initMap[verifier_types.T_UNINIT_THIS] = verifier_types.T_OBJECT(parent.class_.name)
|
||||
self.hasmonenter = any(node.instruction[0] == vops.MONENTER for node in self.iNodes)
|
||||
|
||||
self.entryBlock = self.makeBlockWithInslots(ENTRY_KEY, newlocals=inputTypes, stack=[])
|
||||
self.returnBlock = self.makeBlockWithInslots(RETURN_KEY, newlocals=[], stack=returnTypes)
|
||||
self.returnBlock.jump = ssa_jumps.Return(self, [phi.rval for phi in self.returnBlock.phis])
|
||||
self.rethrowBlock = self.makeBlockWithInslots(RETHROW_KEY, newlocals=[], stack=[verifier_types.THROWABLE_INFO])
|
||||
self.rethrowBlock.jump = ssa_jumps.Rethrow(self, [phi.rval for phi in self.rethrowBlock.phis])
|
||||
|
||||
# for ssagraph to copy
|
||||
self.inputArgs = slotsRvals(self.entryBlock.inslots).localsAsList
|
||||
self.entryBlock.phis = []
|
||||
|
||||
# We need to create stub blocks for every jump target so we can add them as successors during creation
|
||||
jump_targets = [eh.handler for eh in exceptions]
|
||||
for node in self.iNodes:
|
||||
if node.instruction[0] in _jump_instrs:
|
||||
jump_targets += node.successors
|
||||
# add jsr fallthroughs too
|
||||
if node.instruction[0] == vops.JSR and node.returnedFrom is not None:
|
||||
jump_targets.append(node.next_instruction)
|
||||
|
||||
# for simplicity, keep jsr stuff in individual instruction blocks.
|
||||
# Note that subproc.py will need to be modified if this is changed
|
||||
for node in self.iNodes:
|
||||
if node.instruction[0] in (vops.JSR, vops.RET):
|
||||
jump_targets.append(node.key)
|
||||
for key in jump_targets:
|
||||
if key not in self.blockd: # jump_targets may have duplicates
|
||||
self.makeBlock(key)
|
||||
|
||||
self.exceptionhandlers = []
|
||||
for (start, end, handler, index) in exceptions:
|
||||
catchtype = parent.getConstPoolArgs(index)[0] if index else 'java/lang/Throwable'
|
||||
self.exceptionhandlers.append((start, end, self.blockd[handler], catchtype))
|
||||
self.exceptionhandlers.append((0, 65536, self.rethrowBlock, 'java/lang/Throwable'))
|
||||
|
||||
# State variables for the append/builder loop
|
||||
self.current_block = self.entryBlock
|
||||
self.current_slots = slotsRvals(self.current_block.inslots)
|
||||
for node in self.iNodes:
|
||||
# First do a quick check if we have to start a new block
|
||||
if not self._canContinueBlock(node):
|
||||
self._startNewBlock(node.key)
|
||||
vals, outslot_norm = self._getInstrLine(node)
|
||||
|
||||
# Disable exception pruning
|
||||
if opts and not vals.jump:
|
||||
dummyvals = ResultDict(line=ssa_ops.MagicThrow(self.parent))
|
||||
if not self._canAppendInstrToCurrent(node.key, dummyvals):
|
||||
self._startNewBlock(node.key)
|
||||
assert self._canAppendInstrToCurrent(node.key, dummyvals)
|
||||
self._appendInstr(node, dummyvals, self.current_slots, check_terminate=False)
|
||||
vals, outslot_norm = self._getInstrLine(node)
|
||||
|
||||
if not self._canAppendInstrToCurrent(node.key, vals):
|
||||
self._startNewBlock(node.key)
|
||||
vals, outslot_norm = self._getInstrLine(node)
|
||||
|
||||
assert self._canAppendInstrToCurrent(node.key, vals)
|
||||
self._appendInstr(node, vals, outslot_norm)
|
||||
|
||||
# do sanity checks
|
||||
assert len(self.blocks) == len(self.blockd)
|
||||
for block in self.blocks:
|
||||
assert block.jump is not None and block.phis is not None
|
||||
assert len(block.predecessors) == len(set(block.predecessors))
|
||||
# cleanup temp vars
|
||||
block.inslots = None
|
||||
block.throwvars = None
|
||||
block.chpairs = None
|
||||
block.except_used = None
|
||||
block.locals_at_except = None
|
||||
|
||||
def _canContinueBlock(self, node):
|
||||
return (node.key not in self.blockd) and self.current_block.jump is None # fallthrough goto left as None
|
||||
|
||||
def _chPairsAt(self, address):
|
||||
chpairs = []
|
||||
for (start, end, handler, catchtype) in self.exceptionhandlers:
|
||||
if start <= address < end:
|
||||
chpairs.append((catchtype, handler))
|
||||
return chpairs
|
||||
|
||||
def _canAppendInstrToCurrent(self, address, vals):
|
||||
# If appending exception line to block with existing exceptions, make sure the handlers are the same
|
||||
# Also make sure that locals are compatible with all other exceptions in the block
|
||||
# If appending a jump, make sure there is no existing exceptions
|
||||
block = self.current_block
|
||||
if block.chpairs is not None:
|
||||
if vals.jump:
|
||||
return False
|
||||
if vals.line is not None and vals.line.outException is not None:
|
||||
chpairs = self._chPairsAt(address)
|
||||
if chpairs != block.chpairs:
|
||||
return False
|
||||
|
||||
newlocals = {i: self.current_slots.locals[i] for i in block.except_used}
|
||||
return newlocals == block.locals_at_except
|
||||
assert block.jump is None
|
||||
return True
|
||||
|
||||
def pruneUnused(self, key, newlocals):
|
||||
used = toBits(self.used_locals[key])
|
||||
return {i: newlocals[i] for i in used}
|
||||
|
||||
def _startNewBlock(self, key):
|
||||
''' We can't continue appending to the current block, so start a new one (or use existing one at location) '''
|
||||
# Make new block
|
||||
if key not in self.blockd:
|
||||
self.makeBlock(key)
|
||||
|
||||
# Finish current block
|
||||
block = self.current_block
|
||||
curslots = self.current_slots
|
||||
assert block.key != key
|
||||
if block.jump is None:
|
||||
if block.chpairs is not None:
|
||||
assert block.throwvars
|
||||
self._addOnException(block, self.blockd[key], curslots)
|
||||
else:
|
||||
assert not block.throwvars
|
||||
block.jump = ssa_jumps.Goto(self.parent, self.blockd[key])
|
||||
|
||||
if curslots is not None:
|
||||
self.mergeIn((block, False), key, curslots)
|
||||
|
||||
# Update state
|
||||
self.current_block = self.blockd[key]
|
||||
self.current_slots = slotsRvals(self.current_block.inslots)
|
||||
|
||||
def _getInstrLine(self, iNode):
|
||||
parent, initMap = self.parent, self.initMap
|
||||
inslots = self.current_slots
|
||||
instr = iNode.instruction
|
||||
|
||||
# internal variables won't have any preset type info associated, so we should add in the info from the verifier
|
||||
assert len(inslots.stack) == len(iNode.state.stack)
|
||||
for i, ivar in enumerate(inslots.stack):
|
||||
if ivar and ivar.type == SSA_OBJECT and ivar.decltype is None:
|
||||
parent.setObjVarData(ivar, iNode.state.stack[i], initMap)
|
||||
|
||||
for i, ivar in inslots.locals.items():
|
||||
if ivar and ivar.type == SSA_OBJECT and ivar.decltype is None:
|
||||
parent.setObjVarData(ivar, iNode.state.locals[i], initMap)
|
||||
|
||||
vals = instructionHandlers[instr[0]](self, inslots, iNode)
|
||||
newstack = vals.newstack if vals.newstack is not None else inslots.stack
|
||||
newlocals = vals.newlocals if vals.newlocals is not None else inslots.locals
|
||||
outslot_norm = slots_t(locals=newlocals, stack=newstack)
|
||||
return vals, outslot_norm
|
||||
|
||||
def _addOnException(self, block, fallthrough, outslot_norm):
|
||||
parent = self.parent
|
||||
assert block.throwvars and block.chpairs is not None
|
||||
ephi = ssa_ops.ExceptionPhi(parent, block.throwvars)
|
||||
block.lines.append(ephi)
|
||||
|
||||
assert block.jump is None
|
||||
block.jump = ssa_jumps.OnException(parent, ephi.outException, block.chpairs, fallthrough)
|
||||
outslot_except = slots_t(locals=block.locals_at_except, stack=[ephi.outException])
|
||||
for suc in block.jump.getExceptSuccessors():
|
||||
self.mergeIn((block, True), suc.key, outslot_except)
|
||||
|
||||
def _appendInstr(self, iNode, vals, outslot_norm, check_terminate=True):
|
||||
parent = self.parent
|
||||
block = self.current_block
|
||||
line, jump = vals.line, vals.jump
|
||||
if line is not None:
|
||||
block.lines.append(line)
|
||||
assert block.jump is None
|
||||
block.jump = jump
|
||||
|
||||
if line is not None and line.outException is not None:
|
||||
block.throwvars.append(line.outException)
|
||||
inslots = self.current_slots
|
||||
|
||||
if block.chpairs is None:
|
||||
block.chpairs = self._chPairsAt(iNode.key)
|
||||
temp = (self.used_locals[h.key] for t, h in block.chpairs)
|
||||
block.except_used = toBits(reduce(operator.__or__, temp, 0))
|
||||
block.locals_at_except = {i: inslots.locals[i] for i in block.except_used}
|
||||
|
||||
if check_terminate:
|
||||
# Return and Throw must be immediately ended because they don't have normal fallthrough
|
||||
# CheckCast must terminate block because cast type hack later on requires casts to be at end of block
|
||||
if iNode.instruction[0] in (vops.RETURN, vops.THROW) or isinstance(line, ssa_ops.CheckCast):
|
||||
fallthrough = self.getExceptFallthrough(iNode)
|
||||
self._addOnException(block, fallthrough, outslot_norm)
|
||||
|
||||
if block.jump is None:
|
||||
unmerged_slots = outslot_norm
|
||||
else:
|
||||
assert isinstance(block.jump, ssa_jumps.OnException) or not block.throwvars
|
||||
unmerged_slots = None
|
||||
# Make sure that branch targets are distinct, since this is assumed everywhere
|
||||
# Only necessary for if statements as the other jumps merge targets automatically
|
||||
# If statements with both branches jumping to same target are replaced with gotos
|
||||
block.jump = block.jump.reduceSuccessors([])
|
||||
|
||||
if isinstance(block.jump, subproc.ProcCallOp):
|
||||
self.mergeJSROut(iNode, block, outslot_norm)
|
||||
else:
|
||||
for suc in block.jump.getNormalSuccessors():
|
||||
self.mergeIn((block, False), suc.key, outslot_norm)
|
||||
self.current_slots = unmerged_slots
|
||||
assert (block.chpairs is None) == (block.except_used is None) == (block.locals_at_except is None)
|
||||
|
||||
def mergeIn(self, from_key, target_key, outslots):
|
||||
inslots = self.blockd[target_key].inslots
|
||||
|
||||
assert len(inslots.stack) == len(outslots.stack)
|
||||
for i, phi in enumerate(inslots.stack):
|
||||
if phi is not None:
|
||||
phi.add(from_key, outslots.stack[i])
|
||||
|
||||
for i, phi in inslots.locals.items():
|
||||
if phi is not None:
|
||||
phi.add(from_key, outslots.locals[i])
|
||||
|
||||
self.blockd[target_key].predecessors.append(from_key)
|
||||
|
||||
## Block Creation #########################################
|
||||
def _makePhiFromVType(self, block, vt):
|
||||
var = self.parent.makeVarFromVtype(vt, self.initMap)
|
||||
return None if var is None else ssa_ops.Phi(block, var)
|
||||
|
||||
def makeBlockWithInslots(self, key, newlocals, stack):
|
||||
assert key not in self.blockd
|
||||
block = BasicBlock(key)
|
||||
self.blocks.append(block)
|
||||
self.blockd[key] = block
|
||||
|
||||
# create inslot phis
|
||||
stack = [self._makePhiFromVType(block, vt) for vt in stack]
|
||||
newlocals = dict(enumerate(self._makePhiFromVType(block, vt) for vt in newlocals))
|
||||
newlocals = self.pruneUnused(key, newlocals)
|
||||
|
||||
block.inslots = slots_t(locals=newlocals, stack=stack)
|
||||
block.phis = [phi for phi in stack + block.inslots.localsAsList if phi is not None]
|
||||
return block
|
||||
|
||||
def makeBlock(self, key):
|
||||
node = self.iNodeD[key]
|
||||
return self.makeBlockWithInslots(key, node.state.locals, node.state.stack)
|
||||
|
||||
###########################################################
|
||||
def getExceptFallthrough(self, iNode):
|
||||
vop = iNode.instruction[0]
|
||||
if vop == vops.RETURN:
|
||||
return self.blockd[RETURN_KEY]
|
||||
elif vop == vops.THROW:
|
||||
return None
|
||||
key = iNode.successors[0]
|
||||
if key not in self.blockd:
|
||||
self.makeBlock(key)
|
||||
return self.blockd[key]
|
||||
|
||||
def mergeJSROut(self, jsrnode, block, outslot_norm):
|
||||
retnode = self.iNodeD[jsrnode.returnedFrom]
|
||||
jump = block.jump
|
||||
target_key, ft_key = jump.target.key, jump.fallthrough.key
|
||||
assert ft_key == jsrnode.next_instruction
|
||||
|
||||
# first merge regular jump to target
|
||||
self.mergeIn((block, False), target_key, outslot_norm)
|
||||
# create merged outslots for fallthrough
|
||||
fromcall = jump.output
|
||||
mask = [mask for key, mask in retnode.state.masks if key == target_key][0]
|
||||
|
||||
skiplocs = fromcall.locals
|
||||
retlocs = outslot_norm.locals
|
||||
merged = {i: (skiplocs.get(i) if i in mask else retlocs.get(i)) for i in (mask | frozenset(retlocs))}
|
||||
# jump.debug_skipvars = set(merged) - set(locals)
|
||||
|
||||
outslot_merged = slots_t(locals=merged, stack=fromcall.stack)
|
||||
# merge merged outputs with fallthrough
|
||||
self.mergeIn((block, False), ft_key, outslot_merged)
|
||||
|
|
@ -0,0 +1,461 @@
|
|||
from .. import opnames as vops
|
||||
from ..verifier.descriptors import parseFieldDescriptor, parseMethodDescriptor
|
||||
|
||||
from . import objtypes, ssa_jumps, ssa_ops, subproc
|
||||
from .ssa_types import SSA_DOUBLE, SSA_FLOAT, SSA_INT, SSA_LONG, SSA_OBJECT, slots_t
|
||||
|
||||
|
||||
_charToSSAType = {'D':SSA_DOUBLE, 'F':SSA_FLOAT, 'I':SSA_INT, 'J':SSA_LONG,
|
||||
'B':SSA_INT, 'C':SSA_INT, 'S':SSA_INT}
|
||||
def getCategory(c): return 2 if c in 'JD' else 1
|
||||
|
||||
class ResultDict(object):
|
||||
def __init__(self, line=None, jump=None, newstack=None, newlocals=None):
|
||||
self.line = line
|
||||
self.jump = jump
|
||||
self.newstack = newstack
|
||||
self.newlocals = newlocals
|
||||
|
||||
##############################################################################
|
||||
def makeConstVar(parent, type_, val):
|
||||
var = parent.makeVariable(type_)
|
||||
var.const = val
|
||||
return var
|
||||
|
||||
def parseArrOrClassName(desc):
|
||||
# Accept either a class or array descriptor or a raw class name.
|
||||
if desc.startswith('[') or desc.endswith(';'):
|
||||
vtypes = parseFieldDescriptor(desc, unsynthesize=False)
|
||||
tt = objtypes.verifierToSynthetic(vtypes[0])
|
||||
else:
|
||||
tt = objtypes.TypeTT(desc, 0)
|
||||
return tt
|
||||
|
||||
def _floatOrIntMath(fop, iop):
|
||||
def math1(maker, input_, iNode):
|
||||
cat = getCategory(iNode.instruction[1])
|
||||
isfloat = (iNode.instruction[1] in 'DF')
|
||||
op = fop if isfloat else iop
|
||||
|
||||
args = input_.stack[-cat*2::cat]
|
||||
line = op(maker.parent, args)
|
||||
|
||||
newstack = input_.stack[:-2*cat] + [line.rval] + [None]*(cat-1)
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
return math1
|
||||
|
||||
def _intMath(op, isShift):
|
||||
def math2(maker, input_, iNode):
|
||||
cat = getCategory(iNode.instruction[1])
|
||||
# some ops (i.e. shifts) always take int as second argument
|
||||
size = cat+1 if isShift else cat+cat
|
||||
args = input_.stack[-size::cat]
|
||||
line = op(maker.parent, args)
|
||||
newstack = input_.stack[:-size] + [line.rval] + [None]*(cat-1)
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
return math2
|
||||
##############################################################################
|
||||
|
||||
def _anewarray(maker, input_, iNode):
|
||||
name = maker.parent.getConstPoolArgs(iNode.instruction[1])[0]
|
||||
tt = parseArrOrClassName(name)
|
||||
line = ssa_ops.NewArray(maker.parent, input_.stack[-1], tt)
|
||||
newstack = input_.stack[:-1] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _arrlen(maker, input_, iNode):
|
||||
line = ssa_ops.ArrLength(maker.parent, input_.stack[-1:])
|
||||
newstack = input_.stack[:-1] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _arrload(maker, input_, iNode):
|
||||
type_ = _charToSSAType[iNode.instruction[1]]
|
||||
cat = getCategory(iNode.instruction[1])
|
||||
|
||||
line = ssa_ops.ArrLoad(maker.parent, input_.stack[-2:], type_)
|
||||
newstack = input_.stack[:-2] + [line.rval] + [None]*(cat-1)
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _arrload_obj(maker, input_, iNode):
|
||||
line = ssa_ops.ArrLoad(maker.parent, input_.stack[-2:], SSA_OBJECT)
|
||||
newstack = input_.stack[:-2] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _arrstore(maker, input_, iNode):
|
||||
if getCategory(iNode.instruction[1]) > 1:
|
||||
newstack, args = input_.stack[:-4], input_.stack[-4:-1]
|
||||
arr_vt, ind_vt = iNode.state.stack[-4:-2]
|
||||
else:
|
||||
newstack, args = input_.stack[:-3], input_.stack[-3:]
|
||||
arr_vt, ind_vt = iNode.state.stack[-3:-1]
|
||||
line = ssa_ops.ArrStore(maker.parent, args)
|
||||
|
||||
# Check if we can prune the exception early because the
|
||||
# array size and index are known constants
|
||||
if arr_vt.const is not None and ind_vt.const is not None:
|
||||
if 0 <= ind_vt.const < arr_vt.const:
|
||||
line.outException = None
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _arrstore_obj(maker, input_, iNode):
|
||||
line = ssa_ops.ArrStore(maker.parent, input_.stack[-3:])
|
||||
newstack = input_.stack[:-3]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _checkcast(maker, input_, iNode):
|
||||
index = iNode.instruction[1]
|
||||
desc = maker.parent.getConstPoolArgs(index)[0]
|
||||
tt = parseArrOrClassName(desc)
|
||||
line = ssa_ops.CheckCast(maker.parent, tt, input_.stack[-1:])
|
||||
return ResultDict(line=line)
|
||||
|
||||
def _const(maker, input_, iNode):
|
||||
ctype, val = iNode.instruction[1:]
|
||||
cat = getCategory(ctype)
|
||||
type_ = _charToSSAType[ctype]
|
||||
var = makeConstVar(maker.parent, type_, val)
|
||||
newstack = input_.stack + [var] + [None]*(cat-1)
|
||||
return ResultDict(newstack=newstack)
|
||||
|
||||
def _constnull(maker, input_, iNode):
|
||||
var = makeConstVar(maker.parent, SSA_OBJECT, 'null')
|
||||
var.decltype = objtypes.NullTT
|
||||
newstack = input_.stack + [var]
|
||||
return ResultDict(newstack=newstack)
|
||||
|
||||
def _convert(maker, input_, iNode):
|
||||
src_c, dest_c = iNode.instruction[1:]
|
||||
src_cat, dest_cat = getCategory(src_c), getCategory(dest_c)
|
||||
|
||||
stack, arg = input_.stack[:-src_cat], input_.stack[-src_cat]
|
||||
line = ssa_ops.Convert(maker.parent, arg, _charToSSAType[src_c], _charToSSAType[dest_c])
|
||||
|
||||
newstack = stack + [line.rval] + [None]*(dest_cat-1)
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _fcmp(maker, input_, iNode):
|
||||
op, c, NaN_val = iNode.instruction
|
||||
cat = getCategory(c)
|
||||
|
||||
args = input_.stack[-cat*2::cat]
|
||||
line = ssa_ops.FCmp(maker.parent, args, NaN_val)
|
||||
newstack = input_.stack[:-cat*2] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _field_access(maker, input_, iNode):
|
||||
index = iNode.instruction[1]
|
||||
target, name, desc = maker.parent.getConstPoolArgs(index)
|
||||
cat = len(parseFieldDescriptor(desc))
|
||||
|
||||
argcnt = cat if 'put' in iNode.instruction[0] else 0
|
||||
if not 'static' in iNode.instruction[0]:
|
||||
argcnt += 1
|
||||
splitInd = len(input_.stack) - argcnt
|
||||
|
||||
args = [x for x in input_.stack[splitInd:] if x is not None]
|
||||
line = ssa_ops.FieldAccess(maker.parent, iNode.instruction, (target, name, desc), args=args)
|
||||
newstack = input_.stack[:splitInd] + line.returned
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _goto(maker, input_, iNode):
|
||||
jump = ssa_jumps.Goto(maker.parent, maker.blockd[iNode.successors[0]])
|
||||
return ResultDict(jump=jump)
|
||||
|
||||
def _if_a(maker, input_, iNode):
|
||||
null = makeConstVar(maker.parent, SSA_OBJECT, 'null')
|
||||
null.decltype = objtypes.NullTT
|
||||
jump = ssa_jumps.If(maker.parent, iNode.instruction[1], map(maker.blockd.get, iNode.successors), (input_.stack[-1], null))
|
||||
newstack = input_.stack[:-1]
|
||||
return ResultDict(jump=jump, newstack=newstack)
|
||||
|
||||
def _if_i(maker, input_, iNode):
|
||||
zero = makeConstVar(maker.parent, SSA_INT, 0)
|
||||
jump = ssa_jumps.If(maker.parent, iNode.instruction[1], map(maker.blockd.get, iNode.successors), (input_.stack[-1], zero))
|
||||
newstack = input_.stack[:-1]
|
||||
return ResultDict(jump=jump, newstack=newstack)
|
||||
|
||||
def _if_cmp(maker, input_, iNode):
|
||||
jump = ssa_jumps.If(maker.parent, iNode.instruction[1], map(maker.blockd.get, iNode.successors), input_.stack[-2:])
|
||||
newstack = input_.stack[:-2]
|
||||
return ResultDict(jump=jump, newstack=newstack)
|
||||
|
||||
def _iinc(maker, input_, iNode):
|
||||
_, index, amount = iNode.instruction
|
||||
|
||||
oldval = input_.locals[index]
|
||||
constval = makeConstVar(maker.parent, SSA_INT, amount)
|
||||
line = ssa_ops.IAdd(maker.parent, (oldval, constval))
|
||||
|
||||
newlocals = input_.locals.copy()
|
||||
newlocals[index] = line.rval
|
||||
return ResultDict(line=line, newlocals=newlocals)
|
||||
|
||||
def _instanceof(maker, input_, iNode):
|
||||
index = iNode.instruction[1]
|
||||
desc = maker.parent.getConstPoolArgs(index)[0]
|
||||
tt = parseArrOrClassName(desc)
|
||||
line = ssa_ops.InstanceOf(maker.parent, tt, input_.stack[-1:])
|
||||
newstack = input_.stack[:-1] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _invoke(maker, input_, iNode):
|
||||
index = iNode.instruction[1]
|
||||
target, name, desc = maker.parent.getConstPoolArgs(index)
|
||||
target_tt = parseArrOrClassName(target)
|
||||
|
||||
argcnt = len(parseMethodDescriptor(desc)[0])
|
||||
if not 'static' in iNode.instruction[0]:
|
||||
argcnt += 1
|
||||
splitInd = len(input_.stack) - argcnt
|
||||
|
||||
# If we are an initializer, store a copy of the uninitialized verifier type so the Java decompiler can patch things up later
|
||||
isThisCtor = iNode.isThisCtor if iNode.op == vops.INVOKEINIT else False
|
||||
|
||||
args = [x for x in input_.stack[splitInd:] if x is not None]
|
||||
line = ssa_ops.Invoke(maker.parent, iNode.instruction, (target, name, desc),
|
||||
args=args, isThisCtor=isThisCtor, target_tt=target_tt)
|
||||
newstack = input_.stack[:splitInd] + line.returned
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _invoke_dynamic(maker, input_, iNode):
|
||||
index = iNode.instruction[1]
|
||||
desc = maker.parent.getConstPoolArgs(index)[2]
|
||||
argcnt = len(parseMethodDescriptor(desc)[0])
|
||||
splitInd = len(input_.stack) - argcnt
|
||||
|
||||
args = [x for x in input_.stack[splitInd:] if x is not None]
|
||||
line = ssa_ops.InvokeDynamic(maker.parent, desc, args)
|
||||
newstack = input_.stack[:splitInd] + line.returned
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _jsr(maker, input_, iNode):
|
||||
newstack = input_.stack + [None]
|
||||
if iNode.returnedFrom is None:
|
||||
jump = ssa_jumps.Goto(maker.parent, maker.blockd[iNode.successors[0]])
|
||||
return ResultDict(newstack=newstack, jump=jump)
|
||||
|
||||
# create output variables from callop to represent vars received from ret.
|
||||
# We can use {} for initMap since there will never be unintialized types here
|
||||
retnode = maker.iNodeD[iNode.returnedFrom]
|
||||
stack = [maker.parent.makeVarFromVtype(vt, {}) for vt in retnode.out_state.stack]
|
||||
newlocals = dict(enumerate(maker.parent.makeVarFromVtype(vt, {}) for vt in retnode.out_state.locals))
|
||||
newlocals = maker.pruneUnused(retnode.key, newlocals)
|
||||
out_slots = slots_t(locals=newlocals, stack=stack)
|
||||
|
||||
# Simply store the data for now and fix things up once all the blocks are created
|
||||
jump = subproc.ProcCallOp(maker.blockd[iNode.successors[0]], maker.blockd[iNode.next_instruction], input_, out_slots)
|
||||
return ResultDict(jump=jump, newstack=newstack)
|
||||
|
||||
def _lcmp(maker, input_, iNode):
|
||||
args = input_.stack[-4::2]
|
||||
line = ssa_ops.ICmp(maker.parent, args)
|
||||
newstack = input_.stack[:-4] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _ldc(maker, input_, iNode):
|
||||
index, cat = iNode.instruction[1:]
|
||||
entry_type = maker.parent.getConstPoolType(index)
|
||||
args = maker.parent.getConstPoolArgs(index)
|
||||
|
||||
var = None
|
||||
if entry_type == 'String':
|
||||
var = makeConstVar(maker.parent, SSA_OBJECT, args[0])
|
||||
var.decltype = objtypes.StringTT
|
||||
elif entry_type == 'Int':
|
||||
var = makeConstVar(maker.parent, SSA_INT, args[0])
|
||||
elif entry_type == 'Long':
|
||||
var = makeConstVar(maker.parent, SSA_LONG, args[0])
|
||||
elif entry_type == 'Float':
|
||||
var = makeConstVar(maker.parent, SSA_FLOAT, args[0])
|
||||
elif entry_type == 'Double':
|
||||
var = makeConstVar(maker.parent, SSA_DOUBLE, args[0])
|
||||
elif entry_type == 'Class':
|
||||
var = makeConstVar(maker.parent, SSA_OBJECT, parseArrOrClassName(args[0]))
|
||||
var.decltype = objtypes.ClassTT
|
||||
# Todo - handle MethodTypes and MethodHandles?
|
||||
|
||||
assert var
|
||||
newstack = input_.stack + [var] + [None]*(cat-1)
|
||||
return ResultDict(newstack=newstack)
|
||||
|
||||
def _load(maker, input_, iNode):
|
||||
cat = getCategory(iNode.instruction[1])
|
||||
index = iNode.instruction[2]
|
||||
newstack = input_.stack + [input_.locals[index]] + [None]*(cat-1)
|
||||
return ResultDict(newstack=newstack)
|
||||
|
||||
def _monitor(maker, input_, iNode):
|
||||
isExit = 'exit' in iNode.instruction[0]
|
||||
line = ssa_ops.Monitor(maker.parent, input_.stack[-1:], isExit)
|
||||
newstack = input_.stack[:-1]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _multinewarray(maker, input_, iNode):
|
||||
op, index, dim = iNode.instruction
|
||||
name = maker.parent.getConstPoolArgs(index)[0]
|
||||
tt = parseArrOrClassName(name)
|
||||
assert objtypes.dim(tt) >= dim
|
||||
|
||||
line = ssa_ops.MultiNewArray(maker.parent, input_.stack[-dim:], tt)
|
||||
newstack = input_.stack[:-dim] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _neg(maker, input_, iNode):
|
||||
cat = getCategory(iNode.instruction[1])
|
||||
arg = input_.stack[-cat:][0]
|
||||
|
||||
if (iNode.instruction[1] in 'DF'):
|
||||
line = ssa_ops.FNeg(maker.parent, [arg])
|
||||
else: # for integers, we can just write -x as 0 - x
|
||||
zero = makeConstVar(maker.parent, arg.type, 0)
|
||||
line = ssa_ops.ISub(maker.parent, [zero,arg])
|
||||
|
||||
newstack = input_.stack[:-cat] + [line.rval] + [None]*(cat-1)
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _new(maker, input_, iNode):
|
||||
index = iNode.instruction[1]
|
||||
classname = maker.parent.getConstPoolArgs(index)[0]
|
||||
if classname.endswith(';'):
|
||||
classname = classname[1:-1]
|
||||
|
||||
line = ssa_ops.New(maker.parent, classname, iNode.key)
|
||||
newstack = input_.stack + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _newarray(maker, input_, iNode):
|
||||
vtypes = parseFieldDescriptor(iNode.instruction[1], unsynthesize=False)
|
||||
tt = objtypes.verifierToSynthetic(vtypes[0])
|
||||
|
||||
line = ssa_ops.NewArray(maker.parent, input_.stack[-1], tt)
|
||||
newstack = input_.stack[:-1] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _nop(maker, input_, iNode):
|
||||
return ResultDict()
|
||||
|
||||
def _ret(maker, input_, iNode):
|
||||
jump = subproc.DummyRet(input_, maker.blockd[iNode.jsrTarget])
|
||||
return ResultDict(jump=jump)
|
||||
|
||||
def _return(maker, input_, iNode):
|
||||
# Our special return block expects only the return values on the stack
|
||||
rtype = iNode.instruction[1]
|
||||
if rtype is None:
|
||||
newstack = []
|
||||
else:
|
||||
newstack = input_.stack[-getCategory(rtype):]
|
||||
|
||||
# TODO: enable once structuring is smarter
|
||||
# if not maker.hasmonenter:
|
||||
# jump = ssa_jumps.Goto(maker.parent, maker.returnBlock)
|
||||
# return ResultDict(jump=jump, newstack=newstack)
|
||||
|
||||
line = ssa_ops.TryReturn(maker.parent)
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def _store(maker, input_, iNode):
|
||||
cat = getCategory(iNode.instruction[1])
|
||||
index = iNode.instruction[2]
|
||||
|
||||
newlocals = input_.locals.copy()
|
||||
newlocals[index] = input_.stack[-cat]
|
||||
newstack = input_.stack[:-cat]
|
||||
return ResultDict(newstack=newstack, newlocals=newlocals)
|
||||
|
||||
def _switch(maker, input_, iNode):
|
||||
default, raw_table = iNode.instruction[1:3]
|
||||
table = [(k, maker.blockd[v]) for k,v in raw_table]
|
||||
jump = ssa_jumps.Switch(maker.parent, maker.blockd[default], table, input_.stack[-1:])
|
||||
newstack = input_.stack[:-1]
|
||||
return ResultDict(jump=jump, newstack=newstack)
|
||||
|
||||
def _throw(maker, input_, iNode):
|
||||
line = ssa_ops.Throw(maker.parent, input_.stack[-1:])
|
||||
return ResultDict(line=line, newstack=[])
|
||||
|
||||
def _truncate(maker, input_, iNode):
|
||||
dest_c = iNode.instruction[1]
|
||||
signed, width = {'B':(True, 8), 'C':(False, 16), 'S':(True, 16)}[dest_c]
|
||||
|
||||
line = ssa_ops.Truncate(maker.parent, input_.stack[-1], signed=signed, width=width)
|
||||
newstack = input_.stack[:-1] + [line.rval]
|
||||
return ResultDict(line=line, newstack=newstack)
|
||||
|
||||
def genericStackUpdate(maker, input_, iNode):
|
||||
n = iNode.pop_amount
|
||||
stack = input_.stack
|
||||
stack, popped = stack[:-n], stack[-n:]
|
||||
|
||||
for i in iNode.stack_code:
|
||||
stack.append(popped[i])
|
||||
return ResultDict(newstack=stack)
|
||||
|
||||
instructionHandlers = {
|
||||
vops.ADD: _floatOrIntMath(ssa_ops.FAdd, ssa_ops.IAdd),
|
||||
vops.AND: _intMath(ssa_ops.IAnd, isShift=False),
|
||||
vops.ANEWARRAY: _anewarray,
|
||||
vops.ARRLEN: _arrlen,
|
||||
vops.ARRLOAD: _arrload,
|
||||
vops.ARRLOAD_OBJ: _arrload_obj,
|
||||
vops.ARRSTORE: _arrstore,
|
||||
vops.ARRSTORE_OBJ: _arrstore_obj,
|
||||
vops.CHECKCAST: _checkcast,
|
||||
vops.CONST: _const,
|
||||
vops.CONSTNULL: _constnull,
|
||||
vops.CONVERT: _convert,
|
||||
vops.DIV: _floatOrIntMath(ssa_ops.FDiv, ssa_ops.IDiv),
|
||||
vops.FCMP: _fcmp,
|
||||
vops.GETSTATIC: _field_access,
|
||||
vops.GETFIELD: _field_access,
|
||||
vops.GOTO: _goto,
|
||||
vops.IF_A: _if_a,
|
||||
vops.IF_ACMP: _if_cmp, # cmp works on ints or objs
|
||||
vops.IF_I: _if_i,
|
||||
vops.IF_ICMP: _if_cmp,
|
||||
vops.IINC: _iinc,
|
||||
vops.INSTANCEOF: _instanceof,
|
||||
vops.INVOKEINIT: _invoke,
|
||||
vops.INVOKEINTERFACE: _invoke,
|
||||
vops.INVOKESPECIAL: _invoke,
|
||||
vops.INVOKESTATIC: _invoke,
|
||||
vops.INVOKEVIRTUAL: _invoke,
|
||||
vops.INVOKEDYNAMIC: _invoke_dynamic,
|
||||
vops.JSR: _jsr,
|
||||
vops.LCMP: _lcmp,
|
||||
vops.LDC: _ldc,
|
||||
vops.LOAD: _load,
|
||||
vops.MONENTER: _monitor,
|
||||
vops.MONEXIT: _monitor,
|
||||
vops.MULTINEWARRAY: _multinewarray,
|
||||
vops.MUL: _floatOrIntMath(ssa_ops.FMul, ssa_ops.IMul),
|
||||
vops.NEG: _neg,
|
||||
vops.NEW: _new,
|
||||
vops.NEWARRAY: _newarray,
|
||||
vops.NOP: _nop,
|
||||
vops.OR: _intMath(ssa_ops.IOr, isShift=False),
|
||||
vops.PUTSTATIC: _field_access,
|
||||
vops.PUTFIELD: _field_access,
|
||||
vops.REM: _floatOrIntMath(ssa_ops.FRem, ssa_ops.IRem),
|
||||
vops.RET: _ret,
|
||||
vops.RETURN: _return,
|
||||
vops.SHL: _intMath(ssa_ops.IShl, isShift=True),
|
||||
vops.SHR: _intMath(ssa_ops.IShr, isShift=True),
|
||||
vops.STORE: _store,
|
||||
vops.SUB: _floatOrIntMath(ssa_ops.FSub, ssa_ops.ISub),
|
||||
vops.SWITCH: _switch,
|
||||
vops.THROW: _throw,
|
||||
vops.TRUNCATE: _truncate,
|
||||
vops.USHR: _intMath(ssa_ops.IUshr, isShift=True),
|
||||
vops.XOR: _intMath(ssa_ops.IXor, isShift=False),
|
||||
|
||||
vops.SWAP: genericStackUpdate,
|
||||
vops.POP: genericStackUpdate,
|
||||
vops.POP2: genericStackUpdate,
|
||||
vops.DUP: genericStackUpdate,
|
||||
vops.DUPX1: genericStackUpdate,
|
||||
vops.DUPX2: genericStackUpdate,
|
||||
vops.DUP2: genericStackUpdate,
|
||||
vops.DUP2X1: genericStackUpdate,
|
||||
vops.DUP2X2: genericStackUpdate,
|
||||
}
|
||||
|
|
@ -0,0 +1,66 @@
|
|||
import collections, itertools
|
||||
|
||||
from ... import floatutil
|
||||
from .. import objtypes
|
||||
from .int_c import IntConstraint
|
||||
from .float_c import FloatConstraint
|
||||
from .obj_c import ObjectConstraint
|
||||
|
||||
from ..ssa_types import SSA_INT, SSA_LONG, SSA_FLOAT, SSA_DOUBLE, SSA_OBJECT
|
||||
|
||||
# joins become more precise (intersection), meets become more general (union)
|
||||
# Join currently supports joining a max of two constraints
|
||||
# Meet assumes all inputs are not None
|
||||
def join(*cons):
|
||||
if None in cons:
|
||||
return None
|
||||
return cons[0].join(*cons[1:])
|
||||
|
||||
def meet(*cons):
|
||||
if not cons:
|
||||
return None
|
||||
return cons[0].meet(*cons[1:])
|
||||
|
||||
def fromConstant(env, var):
|
||||
ssa_type = var.type
|
||||
cval = var.const
|
||||
|
||||
if ssa_type[0] == SSA_INT[0]:
|
||||
return IntConstraint.const(ssa_type[1], cval)
|
||||
elif ssa_type[0] == SSA_FLOAT[0]:
|
||||
xt = floatutil.fromRawFloat(ssa_type[1], cval)
|
||||
return FloatConstraint.const(ssa_type[1], xt)
|
||||
elif ssa_type[0] == SSA_OBJECT[0]:
|
||||
if var.decltype == objtypes.NullTT:
|
||||
return ObjectConstraint.constNull(env)
|
||||
return ObjectConstraint.fromTops(env, *objtypes.declTypeToActual(env, var.decltype))
|
||||
|
||||
_bots = {
|
||||
SSA_INT: IntConstraint.bot(SSA_INT[1]),
|
||||
SSA_LONG: IntConstraint.bot(SSA_LONG[1]),
|
||||
SSA_FLOAT: FloatConstraint.bot(SSA_FLOAT[1]),
|
||||
SSA_DOUBLE: FloatConstraint.bot(SSA_DOUBLE[1]),
|
||||
}
|
||||
|
||||
def fromVariable(env, var):
|
||||
if var.const is not None:
|
||||
return fromConstant(env, var)
|
||||
ssa_type = var.type
|
||||
|
||||
try:
|
||||
return _bots[ssa_type]
|
||||
except KeyError:
|
||||
assert ssa_type == SSA_OBJECT
|
||||
isnew = var.uninit_orig_num is not None
|
||||
if var.decltype is not None:
|
||||
if var.decltype == objtypes.NullTT:
|
||||
return ObjectConstraint.constNull(env)
|
||||
return ObjectConstraint.fromTops(env, *objtypes.declTypeToActual(env, var.decltype), nonnull=isnew)
|
||||
else:
|
||||
return ObjectConstraint.fromTops(env, [objtypes.ObjectTT], [], nonnull=isnew)
|
||||
|
||||
OpReturnInfo = collections.namedtuple('OpReturnInfo', ['rval', 'eval', 'must_throw'])
|
||||
def returnOrThrow(rval, eval): return OpReturnInfo(rval, eval, False)
|
||||
def maybeThrow(eval): return OpReturnInfo(None, eval, False)
|
||||
def throw(eval): return OpReturnInfo(None, eval, True)
|
||||
def return_(rval): return OpReturnInfo(rval, None, False)
|
||||
|
|
@ -0,0 +1,59 @@
|
|||
from ... import floatutil as fu
|
||||
|
||||
from ..mixin import ValueType
|
||||
|
||||
SPECIALS = frozenset((fu.NAN, fu.INF, fu.NINF, fu.ZERO, fu.NZERO))
|
||||
|
||||
def botRange(size):
|
||||
mbits, emin, emax = size
|
||||
mag = (1<<(mbits+1))-1, emax-mbits
|
||||
return (-1,mag), (1,mag)
|
||||
|
||||
class FloatConstraint(ValueType):
|
||||
def __init__(self, size, finite, special):
|
||||
self.size = size
|
||||
self.finite = finite
|
||||
self.spec = special
|
||||
|
||||
self.isBot = (special == SPECIALS) and (finite == botRange(size))
|
||||
|
||||
@staticmethod
|
||||
def const(size, val):
|
||||
if val in SPECIALS:
|
||||
return FloatConstraint(size, (None, None), frozenset([val]))
|
||||
return FloatConstraint(size, (val, val), frozenset())
|
||||
|
||||
@staticmethod
|
||||
def bot(size):
|
||||
finite = botRange(size)
|
||||
return FloatConstraint(size, finite, SPECIALS)
|
||||
|
||||
def _key(self): return self.finite, self.spec
|
||||
|
||||
def join(*cons): # more precise (intersection)
|
||||
spec = frozenset.intersection(*[c.spec for c in cons])
|
||||
ranges = [c.finite for c in cons]
|
||||
|
||||
if (None, None) in ranges:
|
||||
xmin = xmax = None
|
||||
else:
|
||||
mins, maxs = zip(*ranges)
|
||||
xmin = max(mins, key=fu.sortkey)
|
||||
xmax = min(maxs, key=fu.sortkey)
|
||||
if fu.sortkey(xmax) < fu.sortkey(xmin):
|
||||
xmin = xmax = None
|
||||
if not xmin and not spec:
|
||||
return None
|
||||
return FloatConstraint(cons[0].size, (xmin, xmax), spec)
|
||||
|
||||
def meet(*cons):
|
||||
spec = frozenset.union(*[c.spec for c in cons])
|
||||
ranges = [c.finite for c in cons if c.finite != (None,None)]
|
||||
|
||||
if ranges:
|
||||
mins, maxs = zip(*ranges)
|
||||
xmin = min(mins, key=fu.sortkey)
|
||||
xmax = max(maxs, key=fu.sortkey)
|
||||
else:
|
||||
xmin = xmax = None
|
||||
return FloatConstraint(cons[0].size, (xmin, xmax), spec)
|
||||
|
|
@ -0,0 +1,48 @@
|
|||
from ..mixin import ValueType
|
||||
|
||||
class IntConstraint(ValueType):
|
||||
__slots__ = "width min max".split()
|
||||
|
||||
def __init__(self, width, min_, max_):
|
||||
self.width = width
|
||||
self.min = min_
|
||||
self.max = max_
|
||||
# self.isBot = (-min_ == max_+1 == (1<<width)//2)
|
||||
|
||||
@staticmethod
|
||||
def range(width, min_, max_):
|
||||
if min_ > max_:
|
||||
return None
|
||||
return IntConstraint(width, min_, max_)
|
||||
|
||||
@staticmethod
|
||||
def const(width, val):
|
||||
return IntConstraint(width, val, val)
|
||||
|
||||
@staticmethod
|
||||
def bot(width):
|
||||
return IntConstraint(width, -1<<(width-1), (1<<(width-1))-1)
|
||||
|
||||
def _key(self): return self.min, self.max
|
||||
|
||||
def join(*cons):
|
||||
xmin = max(c.min for c in cons)
|
||||
xmax = min(c.max for c in cons)
|
||||
if xmin > xmax:
|
||||
return None
|
||||
res = IntConstraint(cons[0].width, xmin, xmax)
|
||||
return cons[0] if cons[0] == res else res
|
||||
|
||||
def meet(*cons):
|
||||
xmin = min(c.min for c in cons)
|
||||
xmax = max(c.max for c in cons)
|
||||
return IntConstraint(cons[0].width, xmin, xmax)
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
t = 'Int' if self.width == 32 else 'Long'
|
||||
if self.min == self.max:
|
||||
return '{}({})'.format(t, self.min)
|
||||
elif self == self.bot(self.width):
|
||||
return t
|
||||
return '{}({}, {})'.format(t, self.min, self.max)
|
||||
__repr__ = __str__
|
||||
|
|
@ -0,0 +1,6 @@
|
|||
class ValueType(object):
|
||||
'''Define _key() and inherit from this class to implement comparison and hashing'''
|
||||
# def __init__(self, *args, **kwargs): super(ValueType, self).__init__(*args, **kwargs)
|
||||
def __eq__(self, other): return type(self) == type(other) and self._key() == other._key()
|
||||
def __ne__(self, other): return type(self) != type(other) or self._key() != other._key()
|
||||
def __hash__(self): return hash(self._key())
|
||||
|
|
@ -0,0 +1,123 @@
|
|||
import itertools
|
||||
|
||||
from .. import objtypes
|
||||
from ..mixin import ValueType
|
||||
|
||||
array_supers = 'java/lang/Object','java/lang/Cloneable','java/io/Serializable'
|
||||
obj_fset = frozenset([objtypes.ObjectTT])
|
||||
|
||||
def isAnySubtype(env, x, seq):
|
||||
return any(objtypes.isSubtype(env,x,y) for y in seq)
|
||||
|
||||
class TypeConstraint(ValueType):
|
||||
__slots__ = "env supers exact isBot".split()
|
||||
def __init__(self, env, supers, exact):
|
||||
self.env, self.supers, self.exact = env, frozenset(supers), frozenset(exact)
|
||||
self.isBot = objtypes.ObjectTT in supers
|
||||
|
||||
temp = self.supers | self.exact
|
||||
assert objtypes.NullTT not in temp
|
||||
assert all(objtypes.isBaseTClass(tt) for tt in supers)
|
||||
assert all(objtypes.dim(tt) < 999 for tt in exact)
|
||||
|
||||
def _key(self): return self.supers, self.exact
|
||||
def __nonzero__(self): return bool(self.supers or self.exact)
|
||||
|
||||
def getSingleTType(self):
|
||||
# comSuper doesn't care about order so we can freely pass in nondeterministic order
|
||||
return objtypes.commonSupertype(self.env, list(self.supers) + list(self.exact))
|
||||
|
||||
def isBoolOrByteArray(self):
|
||||
if self.supers or len(self.exact) != 2:
|
||||
return False
|
||||
tt1, tt2 = self.exact
|
||||
bases = objtypes.baset(tt1), objtypes.baset(tt2)
|
||||
return objtypes.dim(tt1) == objtypes.dim(tt2) and sorted(bases) == [objtypes.baset(objtypes.BoolTT), objtypes.baset(objtypes.ByteTT)]
|
||||
|
||||
@staticmethod
|
||||
def reduce(env, supers, exact):
|
||||
newsupers = []
|
||||
for x in supers:
|
||||
if not isAnySubtype(env, x, newsupers):
|
||||
newsupers = [y for y in newsupers if not objtypes.isSubtype(env, y, x)]
|
||||
newsupers.append(x)
|
||||
|
||||
newexact = [x for x in exact if not isAnySubtype(env, x, newsupers)]
|
||||
return TypeConstraint(env, newsupers, newexact)
|
||||
|
||||
def join(*cons):
|
||||
assert len(set(map(type, cons))) == 1
|
||||
env = cons[0].env
|
||||
|
||||
# optimize for the common case of joining with itself or with bot
|
||||
cons = set(c for c in cons if not c.isBot)
|
||||
if not cons:
|
||||
return TypeConstraint(env, obj_fset, [])
|
||||
elif len(cons) == 1:
|
||||
return cons.pop()
|
||||
assert(len(cons) == 2) # joining more than 2 not currently supported
|
||||
|
||||
supers_l, exact_l = zip(*(c._key() for c in cons))
|
||||
|
||||
newsupers = set()
|
||||
for t1,t2 in itertools.product(*supers_l):
|
||||
if objtypes.isSubtype(env, t1, t2):
|
||||
newsupers.add(t1)
|
||||
elif objtypes.isSubtype(env, t2, t1):
|
||||
newsupers.add(t2)
|
||||
else: # TODO: need to add special handling for interfaces here
|
||||
pass
|
||||
|
||||
newexact = frozenset.union(*exact_l)
|
||||
for c in cons:
|
||||
newexact = [x for x in newexact if x in c.exact or isAnySubtype(env, x, c.supers)]
|
||||
return TypeConstraint.reduce(env, newsupers, newexact)
|
||||
|
||||
def meet(*cons):
|
||||
supers = frozenset.union(*(c.supers for c in cons))
|
||||
exact = frozenset.union(*(c.exact for c in cons))
|
||||
return TypeConstraint.reduce(cons[0].env, supers, exact)
|
||||
|
||||
class ObjectConstraint(ValueType):
|
||||
__slots__ = "null types isBot".split()
|
||||
def __init__(self, null, types):
|
||||
self.null, self.types = null, types
|
||||
self.isBot = null and types.isBot
|
||||
|
||||
@staticmethod
|
||||
def constNull(env):
|
||||
return ObjectConstraint(True, TypeConstraint(env, [], []))
|
||||
|
||||
@staticmethod
|
||||
def fromTops(env, supers, exact, nonnull=False):
|
||||
types = TypeConstraint(env, supers, exact)
|
||||
if nonnull and not types:
|
||||
return None
|
||||
return ObjectConstraint(not nonnull, types)
|
||||
|
||||
def _key(self): return self.null, self.types
|
||||
|
||||
def isConstNull(self): return self.null and not self.types
|
||||
|
||||
def getSingleTType(self):
|
||||
return self.types.getSingleTType() if self.types else objtypes.NullTT
|
||||
|
||||
def join(*cons):
|
||||
null = all(c.null for c in cons)
|
||||
types = TypeConstraint.join(*(c.types for c in cons))
|
||||
if not null and not types:
|
||||
return None
|
||||
|
||||
res = ObjectConstraint(null, types)
|
||||
return cons[0] if cons[0] == res else res
|
||||
|
||||
def meet(*cons):
|
||||
null = any(c.null for c in cons)
|
||||
types = TypeConstraint.meet(*(c.types for c in cons))
|
||||
return ObjectConstraint(null, types)
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
if not self.types:
|
||||
return 'Obj(null)'
|
||||
return 'Obj({}, {}, {})'.format(self.null, sorted(self.types.supers), sorted(self.types.exact))
|
||||
__repr__ = __str__
|
||||
|
|
@ -0,0 +1,207 @@
|
|||
import collections
|
||||
import itertools
|
||||
|
||||
from . import objtypes
|
||||
from .mixin import ValueType
|
||||
|
||||
class CatchSetManager(object):
|
||||
def __init__(self, env, sets, mask):
|
||||
self.env, self.sets, self.mask = env, sets, mask
|
||||
assert not self._conscheck()
|
||||
|
||||
@staticmethod # factory
|
||||
def new(env, chpairs):
|
||||
sets = collections.OrderedDict() # make this ordered since OnException relies on it
|
||||
sofar = empty = ExceptionSet.EMPTY
|
||||
for catchtype, handler in chpairs:
|
||||
old = sets.get(handler, empty)
|
||||
new = ExceptionSet.fromTops(env, catchtype)
|
||||
sets[handler] = old | (new - sofar)
|
||||
sofar = sofar | new
|
||||
return CatchSetManager(env, sets, sofar)
|
||||
|
||||
def newMask(self, mask):
|
||||
for k in self.sets:
|
||||
self.sets[k] &= mask
|
||||
self.mask &= mask
|
||||
assert not self._conscheck()
|
||||
|
||||
def pruneKeys(self):
|
||||
for handler, catchset in list(self.sets.items()):
|
||||
if not catchset:
|
||||
del self.sets[handler]
|
||||
|
||||
def copy(self):
|
||||
return CatchSetManager(self.env, self.sets.copy(), self.mask)
|
||||
|
||||
def replaceKeys(self, replace):
|
||||
self.sets = collections.OrderedDict((replace.get(key,key), val) for key, val in self.sets.items())
|
||||
|
||||
def _conscheck(self):
|
||||
temp = ExceptionSet.EMPTY
|
||||
for v in self.sets.values():
|
||||
assert not v & temp
|
||||
temp |= v
|
||||
assert temp == self.mask
|
||||
assert isinstance(self.sets, collections.OrderedDict)
|
||||
|
||||
class ExceptionSet(ValueType):
|
||||
__slots__ = "env pairs".split()
|
||||
def __init__(self, env, pairs): # assumes arguments are in reduced form
|
||||
self.env = env
|
||||
self.pairs = frozenset([(x,frozenset(y)) for x,y in pairs])
|
||||
|
||||
# We allow env to be None for the empty set so we can construct empty sets easily
|
||||
# Any operation resulting in a nonempty set will get its env from the nonempty argument
|
||||
assert self.empty() or self.env is not None
|
||||
|
||||
# make sure set is fully reduced
|
||||
parts = []
|
||||
for t, holes in pairs:
|
||||
parts.append(t)
|
||||
parts.extend(holes)
|
||||
assert len(set(parts)) == len(parts)
|
||||
|
||||
@staticmethod # factory
|
||||
def fromTops(env, *tops):
|
||||
return ExceptionSet(env, [(x, frozenset()) for x in tops])
|
||||
|
||||
def _key(self): return self.pairs
|
||||
def empty(self): return not self.pairs
|
||||
def __nonzero__(self): return bool(self.pairs)
|
||||
|
||||
def getTopTTs(self): return sorted([objtypes.TypeTT(top,0) for (top,holes) in self.pairs])
|
||||
|
||||
def __sub__(self, other):
|
||||
assert type(self) == type(other)
|
||||
if self.empty() or other.empty():
|
||||
return self
|
||||
if self == other:
|
||||
return ExceptionSet.EMPTY
|
||||
|
||||
subtest = self.env.isSubclass
|
||||
pairs = self.pairs
|
||||
|
||||
for pair2 in other.pairs:
|
||||
# Warning, due to a bug in Python, TypeErrors raised inside the gen expr will give an incorect error message
|
||||
# TypeError: type object argument after * must be a sequence, not generator
|
||||
# This can be worked around by using a list comprehension instead of a genexpr after the *
|
||||
pairs = itertools.chain(*[ExceptionSet.diffPair(subtest, pair1, pair2) for pair1 in pairs])
|
||||
return ExceptionSet.reduce(self.env, pairs)
|
||||
|
||||
def __or__(self, other):
|
||||
assert type(self) == type(other)
|
||||
if other.empty() or self == other:
|
||||
return self
|
||||
if self.empty():
|
||||
return other
|
||||
return ExceptionSet.reduce(self.env, self.pairs | other.pairs)
|
||||
|
||||
def __and__(self, other):
|
||||
assert type(self) == type(other)
|
||||
new = self - (self - other)
|
||||
return new
|
||||
|
||||
def isdisjoint(self, other):
|
||||
return (self-other) == self
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
parts = [('{} - [{}]'.format(top, ', '.join(sorted(holes))) if holes else top) for top, holes in self.pairs]
|
||||
return 'ES[{}]'.format(', '.join(parts))
|
||||
__repr__ = __str__
|
||||
|
||||
@staticmethod
|
||||
def diffPair(subtest, pair1, pair2): # subtract pair2 from pair1. Returns a list of new pairs
|
||||
# todo - find way to make this less ugly
|
||||
t1, holes1 = pair1
|
||||
t2, holes2 = pair2
|
||||
if subtest(t1,t2): # t2 >= t1
|
||||
if any(subtest(t1, h) for h in holes2):
|
||||
return pair1,
|
||||
else:
|
||||
newpairs = []
|
||||
holes2 = [h for h in holes2 if subtest(h, t1) and not any(subtest(h,h2) for h2 in holes1)]
|
||||
|
||||
for h in holes2:
|
||||
newholes = [h2 for h2 in holes1 if subtest(h2, h)]
|
||||
newpairs.append((h, newholes))
|
||||
return newpairs
|
||||
elif subtest(t2,t1): # t2 < t1
|
||||
if any(subtest(t2, h) for h in holes1):
|
||||
return pair1,
|
||||
else:
|
||||
newpairs = [(t1,ExceptionSet.reduceHoles(subtest, list(holes1)+[t2]))]
|
||||
holes2 = [h for h in holes2 if not any(subtest(h,h2) for h2 in holes1)]
|
||||
|
||||
for h in holes2:
|
||||
newholes = [h2 for h2 in holes1 if subtest(h2, h)]
|
||||
newpairs.append((h, newholes))
|
||||
return newpairs
|
||||
else:
|
||||
return pair1,
|
||||
|
||||
@staticmethod
|
||||
def mergePair(subtest, pair1, pair2): # merge pair2 into pair1 and return the union
|
||||
t1, holes1 = pair1
|
||||
t2, holes2 = pair2
|
||||
assert subtest(t2,t1)
|
||||
|
||||
if t2 in holes1:
|
||||
holes1 = list(holes1)
|
||||
holes1.remove(t2)
|
||||
return t1, holes1 + list(holes2)
|
||||
|
||||
# TODO - this can probably be made more efficient
|
||||
holes1a = set(h for h in holes1 if not subtest(h, t2))
|
||||
holes1b = [h for h in holes1 if h not in holes1a]
|
||||
|
||||
merged_holes = set()
|
||||
for h1, h2 in itertools.product(holes1b, holes2):
|
||||
if subtest(h2, h1):
|
||||
merged_holes.add(h1)
|
||||
elif subtest(h1, h2):
|
||||
merged_holes.add(h2)
|
||||
merged_holes = ExceptionSet.reduceHoles(subtest, merged_holes)
|
||||
assert len(merged_holes) <= len(holes1b) + len(holes2)
|
||||
return t1, (list(holes1a) + merged_holes)
|
||||
|
||||
@staticmethod
|
||||
def reduceHoles(subtest, holes):
|
||||
newholes = []
|
||||
for hole in holes:
|
||||
for ehole in newholes:
|
||||
if subtest(hole, ehole):
|
||||
break
|
||||
else:
|
||||
newholes = [hole] + [h for h in newholes if not subtest(h, hole)]
|
||||
return newholes
|
||||
|
||||
@staticmethod
|
||||
def reduce(env, pairs):
|
||||
subtest = env.isSubclass
|
||||
pairs = [pair for pair in pairs if pair[0] not in pair[1]] # remove all degenerate pairs
|
||||
|
||||
newpairs = []
|
||||
while pairs:
|
||||
top, holes = pair = pairs.pop()
|
||||
|
||||
# look for an existing top to merge into
|
||||
for epair in newpairs[:]:
|
||||
etop, eholes = epair
|
||||
# new pair can be merged into existing pair
|
||||
if subtest(top, etop) and (top in eholes or not any(subtest(top, ehole) for ehole in eholes)):
|
||||
new = ExceptionSet.mergePair(subtest, epair, pair)
|
||||
newpairs, pairs = [new], [p for p in newpairs if p is not epair] + pairs
|
||||
break
|
||||
# existing pair can be merged into new pair
|
||||
elif subtest(etop, top) and (etop in holes or not any(subtest(etop, hole) for hole in holes)):
|
||||
new = ExceptionSet.mergePair(subtest, pair, epair)
|
||||
newpairs, pairs = [new], [p for p in newpairs if p is not epair] + pairs
|
||||
break
|
||||
# pair is incomparable to all existing pairs
|
||||
else:
|
||||
holes = ExceptionSet.reduceHoles(subtest, holes)
|
||||
newpairs.append((top,holes))
|
||||
return ExceptionSet(env, newpairs)
|
||||
|
||||
ExceptionSet.EMPTY = ExceptionSet(None, [])
|
||||
|
|
@ -0,0 +1,11 @@
|
|||
from . import objtypes
|
||||
|
||||
# common exception types
|
||||
Arithmetic = objtypes.TypeTT('java/lang/ArithmeticException', 0)
|
||||
ArrayOOB = objtypes.TypeTT('java/lang/ArrayIndexOutOfBoundsException', 0)
|
||||
ArrayStore = objtypes.TypeTT('java/lang/ArrayStoreException', 0)
|
||||
ClassCast = objtypes.TypeTT('java/lang/ClassCastException', 0)
|
||||
MonState = objtypes.TypeTT('java/lang/IllegalMonitorStateException', 0)
|
||||
NegArrSize = objtypes.TypeTT('java/lang/NegativeArraySizeException', 0)
|
||||
NullPtr = objtypes.TypeTT('java/lang/NullPointerException', 0)
|
||||
OOM = objtypes.TypeTT('java/lang/OutOfMemoryError', 0)
|
||||
|
|
@ -0,0 +1,9 @@
|
|||
class SSAFunctionBase(object):
|
||||
def __init__(self, parent, arguments):
|
||||
self.parent = parent
|
||||
self.params = list(arguments)
|
||||
assert None not in self.params
|
||||
|
||||
def replaceVars(self, rdict):
|
||||
self.params = [rdict.get(x,x) for x in self.params]
|
||||
assert None not in self.params
|
||||
|
|
@ -0,0 +1,764 @@
|
|||
import collections
|
||||
import copy
|
||||
import functools
|
||||
import itertools
|
||||
|
||||
from .. import graph_util
|
||||
from ..verifier.descriptors import parseUnboundMethodDescriptor
|
||||
|
||||
from . import blockmaker, constraints, objtypes, ssa_jumps, ssa_ops, subproc
|
||||
from .ssa_types import BasicBlock, SSA_OBJECT, verifierToSSAType
|
||||
|
||||
class SSA_Variable(object):
|
||||
__slots__ = 'type','origin','name','const','decltype','uninit_orig_num'
|
||||
|
||||
def __init__(self, type_, origin=None, name=""):
|
||||
self.type = type_ # SSA_INT, SSA_OBJECT, etc.
|
||||
self.origin = origin
|
||||
self.name = name
|
||||
self.const = None
|
||||
self.decltype = None # for objects, the inferred type from the verifier if any
|
||||
self.uninit_orig_num = None # if uninitialized, the bytecode offset of the new instr
|
||||
|
||||
# for debugging
|
||||
def __str__(self): # pragma: no cover
|
||||
return self.name if self.name else super(SSA_Variable, self).__str__()
|
||||
|
||||
def __repr__(self): # pragma: no cover
|
||||
name = self.name if self.name else "@" + hex(id(self))
|
||||
return "Var {}".format(name)
|
||||
|
||||
# This class is the main IR for bytecode level methods. It consists of a control
|
||||
# flow graph (CFG) in static single assignment form (SSA). Each node in the
|
||||
# graph is a BasicBlock. This consists of a list of phi statements representing
|
||||
# inputs, a list of operations, and a jump statement. Exceptions are represented
|
||||
# explicitly in the graph with the OnException jump. Each block also keeps track
|
||||
# of the unary constraints on the variables in that block.
|
||||
|
||||
# Handling of subprocedures is rather annoying. Each complete subproc has an associated
|
||||
# ProcInfo while jsrs and rets are represented by ProcCallOp and DummyRet respectively.
|
||||
# The jsrblock has the target and fallthrough as successors, while the fallthrough has
|
||||
# the jsrblock as predecessor, but not the retblock. Control flow paths where the proc
|
||||
# never returns are represented by ordinary jumps from blocks in the procedure to outside
|
||||
# Successful completion of the proc is represented by the fallthrough edge. The fallthrough
|
||||
# block gets its variables from the jsrblock, including skip vars which don't depend on the
|
||||
# proc, and variables from jsr.output which represent what would have been returned from ret
|
||||
# Every proc has a reachable retblock. Jsrs with no associated ret are simply turned
|
||||
# into gotos during the initial basic block creation.
|
||||
|
||||
class SSA_Graph(object):
|
||||
entryKey = blockmaker.ENTRY_KEY
|
||||
|
||||
def __init__(self, code):
|
||||
self.code = code
|
||||
self.class_ = code.class_
|
||||
self.env = self.class_.env
|
||||
|
||||
self.inputArgs = None
|
||||
self.entryBlock = None
|
||||
self.blocks = None
|
||||
self.procs = None # used to store information on subprocedues (from the JSR instructions)
|
||||
|
||||
self.block_numberer = itertools.count(-4,-1)
|
||||
|
||||
def condenseBlocks(self):
|
||||
assert not self.procs
|
||||
old = self.blocks
|
||||
# Can't do a consistency check on entry as the graph may be in an inconsistent state at this point
|
||||
# Since the purpose of this function is to prune unreachable blocks from self.blocks
|
||||
sccs = graph_util.tarjanSCC([self.entryBlock], lambda block:block.jump.getSuccessors())
|
||||
self.blocks = list(itertools.chain.from_iterable(map(reversed, sccs[::-1])))
|
||||
|
||||
assert set(self.blocks) <= set(old)
|
||||
if len(self.blocks) < len(old):
|
||||
kept = set(self.blocks)
|
||||
for block in self.blocks:
|
||||
for pair in block.predecessors[:]:
|
||||
if pair[0] not in kept:
|
||||
block.removePredPair(pair)
|
||||
return [b for b in old if b not in kept]
|
||||
return []
|
||||
|
||||
def removeUnusedVariables(self):
|
||||
assert not self.procs
|
||||
roots = [x for x in self.inputArgs if x is not None]
|
||||
for block in self.blocks:
|
||||
roots += block.jump.params
|
||||
for op in block.lines:
|
||||
if op.has_side_effects:
|
||||
roots += op.params
|
||||
|
||||
reachable = graph_util.topologicalSort(roots, lambda var:(var.origin.params if var.origin else []))
|
||||
|
||||
keepset = set(reachable)
|
||||
assert None not in keepset
|
||||
def filterOps(oldops):
|
||||
newops = []
|
||||
for op in oldops:
|
||||
# if any of the params is being removed due to being unreachable, we can assume the whole function can be removed
|
||||
keep = keepset.issuperset(op.params) and (op.has_side_effects or not keepset.isdisjoint(op.getOutputs()))
|
||||
if keep:
|
||||
newops.append(op)
|
||||
for v in op.getOutputs():
|
||||
if v and v not in keepset:
|
||||
op.removeOutput(v)
|
||||
else:
|
||||
assert keepset.isdisjoint(op.getOutputs())
|
||||
assert not op.has_side_effects
|
||||
return newops
|
||||
|
||||
for block in self.blocks:
|
||||
block.phis = filterOps(block.phis)
|
||||
block.lines = filterOps(block.lines)
|
||||
block.filterVarConstraints(keepset)
|
||||
assert self._conscheck() is None
|
||||
|
||||
def mergeSingleSuccessorBlocks(self):
|
||||
assert(not self.procs) # Make sure that all single jsr procs are inlined first
|
||||
assert self._conscheck() is None
|
||||
|
||||
removed = set()
|
||||
for block in self.blocks:
|
||||
if block in removed:
|
||||
continue
|
||||
while isinstance(block.jump, ssa_jumps.Goto):
|
||||
jump = block.jump
|
||||
block2 = jump.getNormalSuccessors()[0]
|
||||
fromkey = block, False
|
||||
if block2.predecessors != [fromkey]:
|
||||
break
|
||||
|
||||
jump2 = block2.jump
|
||||
ucs = block.unaryConstraints
|
||||
ucs2 = block2.unaryConstraints
|
||||
replace = {phi.rval: phi.get(fromkey) for phi in block2.phis}
|
||||
for var2, var in replace.items():
|
||||
ucs[var] = constraints.join(ucs[var], ucs2.pop(var2))
|
||||
ucs.update(ucs2)
|
||||
|
||||
for op in block2.lines:
|
||||
op.replaceVars(replace)
|
||||
block.lines += block2.lines
|
||||
|
||||
jump2.replaceVars(replace)
|
||||
block.jump = jump2
|
||||
|
||||
# remember to update phis of blocks referring to old child!
|
||||
for successor, t in block.jump.getSuccessorPairs():
|
||||
successor.replacePredPair((block2, t), (block, t))
|
||||
for phi in successor.phis:
|
||||
phi.replaceVars(replace)
|
||||
removed.add(block2)
|
||||
self.blocks = [b for b in self.blocks if b not in removed]
|
||||
assert self._conscheck() is None
|
||||
|
||||
def disconnectConstantVariables(self):
|
||||
for block in self.blocks:
|
||||
for var, uc in block.unaryConstraints.items():
|
||||
if var.origin is not None:
|
||||
newval = None
|
||||
if var.type[0] == 'int':
|
||||
if uc.min == uc.max:
|
||||
newval = uc.min
|
||||
elif var.type[0] == 'obj':
|
||||
if uc.isConstNull():
|
||||
newval = 'null'
|
||||
|
||||
if newval is not None:
|
||||
var.origin.removeOutput(var)
|
||||
var.origin = None
|
||||
var.const = newval
|
||||
block.phis = [phi for phi in block.phis if phi.rval is not None]
|
||||
assert self._conscheck() is None
|
||||
|
||||
def _conscheck(self):
|
||||
'''Sanity check'''
|
||||
for block in self.blocks:
|
||||
assert block.jump is not None
|
||||
for phi in block.phis:
|
||||
assert phi.rval is None or phi.rval in block.unaryConstraints
|
||||
for k,v in phi.dict.items():
|
||||
assert v in k[0].unaryConstraints
|
||||
|
||||
keys = [block.key for block in self.blocks]
|
||||
assert len(set(keys)) == len(keys)
|
||||
temp = [self.entryBlock]
|
||||
for proc in self.procs:
|
||||
temp += [proc.retblock]
|
||||
temp += proc.jsrblocks
|
||||
assert len(set(temp)) == len(temp)
|
||||
|
||||
def copyPropagation(self):
|
||||
# Loop aware copy propagation
|
||||
assert not self.procs
|
||||
assert self._conscheck() is None
|
||||
|
||||
# The goal is to propagate constants that would never be inferred pessimistically
|
||||
# due to the prescence of loops. Variables that aren't derived from a constant or phi
|
||||
# are treated as opaque and variables are processed by SCC in topological order.
|
||||
# For each scc, we can infer that it is the meet of all inputs that come from variables
|
||||
# in different sccs that come before it in topological order, thus ignoring variables
|
||||
# in the current scc (the loop problem).
|
||||
v2b = {}
|
||||
assigns = collections.OrderedDict()
|
||||
for block in self.blocks:
|
||||
for var in block.unaryConstraints:
|
||||
v2b[var] = block
|
||||
for phi in block.phis:
|
||||
assigns[phi.rval] = map(phi.get, block.predecessors)
|
||||
|
||||
UCs = {}
|
||||
sccs = graph_util.tarjanSCC(assigns, lambda v:assigns.get(v, []))
|
||||
for scc in sccs:
|
||||
if all(var in assigns for var in scc):
|
||||
invars = sum(map(assigns.get, scc), [])
|
||||
inputs = [UCs[invar] for invar in invars if invar in UCs]
|
||||
assert inputs
|
||||
uc = constraints.meet(*inputs)
|
||||
|
||||
for var in scc:
|
||||
old = v2b[var].unaryConstraints[var]
|
||||
new = constraints.join(uc, old) or old # temporary hack
|
||||
v2b[var].unaryConstraints[var] = UCs[var] = new
|
||||
else:
|
||||
# There is a root in this scc, so we can't do anything
|
||||
for var in scc:
|
||||
UCs[var] = v2b[var].unaryConstraints[var]
|
||||
|
||||
assert self._conscheck() is None
|
||||
|
||||
def abstractInterpert(self):
|
||||
# Sparse conditional constant propagation and type inference
|
||||
assert not self.procs
|
||||
assert self._conscheck() is None
|
||||
|
||||
visit_counts = collections.defaultdict(int)
|
||||
dirty_phis = set(itertools.chain.from_iterable(block.phis for block in self.blocks))
|
||||
while dirty_phis:
|
||||
for block in self.blocks:
|
||||
assert block in self.blocks
|
||||
UCs = block.unaryConstraints
|
||||
assert None not in UCs.values()
|
||||
dirty = visit_counts[block] == 0
|
||||
for phi in block.phis:
|
||||
if phi in dirty_phis:
|
||||
dirty_phis.remove(phi)
|
||||
inputs = [key[0].unaryConstraints[phi.get(key)] for key in block.predecessors]
|
||||
out = constraints.meet(*inputs)
|
||||
old = UCs[phi.rval]
|
||||
UCs[phi.rval] = out = constraints.join(old, out)
|
||||
dirty = dirty or out != old
|
||||
assert out
|
||||
|
||||
if not dirty or visit_counts[block] >= 5:
|
||||
continue
|
||||
visit_counts[block] += 1
|
||||
|
||||
must_throw = False
|
||||
dirty_vars = set()
|
||||
last_line = block.lines[-1] if block.lines else None # Keep reference handy to exception phi, if any
|
||||
for i, op in enumerate(block.lines):
|
||||
if hasattr(op, 'propagateConstraints'):
|
||||
output_vars = op.getOutputs()
|
||||
inputs = [UCs[var] for var in op.params]
|
||||
assert None not in inputs
|
||||
output_info = op.propagateConstraints(*inputs)
|
||||
|
||||
for var, out in zip(output_vars, [output_info.rval, output_info.eval]):
|
||||
if var is None:
|
||||
continue
|
||||
old = UCs[var]
|
||||
UCs[var] = out = constraints.join(old, out)
|
||||
if out is None:
|
||||
if var is op.outException:
|
||||
assert isinstance(last_line, ssa_ops.ExceptionPhi)
|
||||
last_line.params.remove(var)
|
||||
op.removeOutput(var) # Note, this must be done after the op.outException check!
|
||||
del UCs[var]
|
||||
elif out != old:
|
||||
dirty_vars.add(var)
|
||||
|
||||
if output_info.must_throw:
|
||||
must_throw = True
|
||||
# Remove all code after this in the basic block and adjust exception code
|
||||
# at end as appropriate
|
||||
assert isinstance(last_line, ssa_ops.ExceptionPhi)
|
||||
assert i < len(block.lines) and op.outException
|
||||
removed = block.lines[i+1:-1]
|
||||
block.lines = block.lines[:i+1] + [last_line]
|
||||
for op2 in removed:
|
||||
if op2.outException:
|
||||
last_line.params.remove(op2.outException)
|
||||
for var in op2.getOutputs():
|
||||
if var is not None:
|
||||
del UCs[var]
|
||||
break
|
||||
|
||||
# now handle end of block
|
||||
if isinstance(last_line, ssa_ops.ExceptionPhi):
|
||||
inputs = map(UCs.get, last_line.params)
|
||||
out = constraints.meet(*inputs)
|
||||
old = UCs[last_line.outException]
|
||||
assert out is None or not out.null
|
||||
UCs[last_line.outException] = out = constraints.join(old, out)
|
||||
if out is None:
|
||||
del UCs[last_line.outException]
|
||||
block.lines.pop()
|
||||
elif out != old:
|
||||
dirty_vars.add(last_line.outException)
|
||||
|
||||
# prune jumps
|
||||
dobreak = False
|
||||
if hasattr(block.jump, 'constrainJumps'):
|
||||
assert block.jump.params
|
||||
oldEdges = block.jump.getSuccessorPairs()
|
||||
inputs = map(UCs.get, block.jump.params)
|
||||
block.jump = block.jump.constrainJumps(*inputs)
|
||||
# No exception case ordinarily won't be pruned, so we have to handle it explicitly
|
||||
if must_throw and isinstance(block.jump, ssa_jumps.OnException):
|
||||
if block.jump.getNormalSuccessors(): # make sure it wasn't already pruned
|
||||
fallthrough = block.jump.getNormalSuccessors()[0]
|
||||
block.jump = block.jump.reduceSuccessors([(fallthrough, False)])
|
||||
|
||||
newEdges = block.jump.getSuccessorPairs()
|
||||
if newEdges != oldEdges:
|
||||
pruned = [x for x in oldEdges if x not in newEdges]
|
||||
for (child,t) in pruned:
|
||||
child.removePredPair((block,t))
|
||||
|
||||
removed_blocks = self.condenseBlocks()
|
||||
# In case where no blocks were removed, self.blocks will possibly be in a different
|
||||
# order than the version of self.blocks we are iterating over, but it still has the
|
||||
# same contents, so this should be safe. If blocks were removed, we break out of the
|
||||
# list and restart to avoid the possibility of processing an unreachable block.
|
||||
dobreak = len(removed_blocks) > 0
|
||||
for removed in removed_blocks:
|
||||
for phi in removed.phis:
|
||||
dirty_phis.discard(phi)
|
||||
|
||||
# update dirty set
|
||||
for child, t in block.jump.getSuccessorPairs():
|
||||
assert child in self.blocks
|
||||
for phi in child.phis:
|
||||
if phi.get((block, t)) in dirty_vars:
|
||||
dirty_phis.add(phi)
|
||||
if dobreak:
|
||||
break
|
||||
|
||||
# Try to turn switches into if statements - note that this may
|
||||
# introduce a new variable and this modify block.unaryConstraints
|
||||
# However, it won't change the control flow graph structure
|
||||
for block in self.blocks:
|
||||
if isinstance(block.jump, ssa_jumps.Switch):
|
||||
block.jump = block.jump.simplifyToIf(block)
|
||||
|
||||
def simplifyThrows(self):
|
||||
# Try to turn throws into gotos where possible. This primarily helps with certain patterns of try-with-resources
|
||||
# To do this, the exception must be known to be non null and there must be only one target that can catch it
|
||||
# As a heuristic, we also restrict it to cases where every predecessor of the target can be converted
|
||||
candidates = collections.defaultdict(list)
|
||||
for block in self.blocks:
|
||||
if not isinstance(block.jump, ssa_jumps.OnException) or len(block.jump.getSuccessorPairs()) != 1:
|
||||
continue
|
||||
if len(block.lines[-1].params) != 1 or not isinstance(block.lines[-2], ssa_ops.Throw):
|
||||
continue
|
||||
if block.unaryConstraints[block.lines[-2].params[0]].null:
|
||||
continue
|
||||
|
||||
candidates[block.jump.getExceptSuccessors()[0]].append(block)
|
||||
|
||||
for child in self.blocks:
|
||||
if not candidates[child] or len(candidates[child]) < len(child.predecessors):
|
||||
continue
|
||||
|
||||
for parent in candidates[child]:
|
||||
ephi = parent.lines.pop()
|
||||
throw_op = parent.lines.pop()
|
||||
|
||||
var1 = throw_op.params[0]
|
||||
var2 = throw_op.outException
|
||||
assert ephi.params == [var2]
|
||||
var3 = ephi.outException
|
||||
assert parent.jump.params[0] == var3
|
||||
|
||||
for phi in child.phis:
|
||||
phi.replaceVars({var3: var1})
|
||||
child.replacePredPair((parent, True), (parent, False))
|
||||
|
||||
del parent.unaryConstraints[var2]
|
||||
del parent.unaryConstraints[var3]
|
||||
parent.jump = ssa_jumps.Goto(self, child)
|
||||
|
||||
def simplifyCatchIgnored(self):
|
||||
# When there is a single throwing instruction, which is garuenteed to throw, has a single handler, and
|
||||
# the caught exception is unused, turn it into a goto. This simplifies a pattern used by some obfuscators
|
||||
# that do stuff like try{new int[-1];} catch(Exception e) {...}
|
||||
candidates = collections.defaultdict(list)
|
||||
for block in self.blocks:
|
||||
if not isinstance(block.jump, ssa_jumps.OnException) or len(block.jump.getSuccessorPairs()) != 1:
|
||||
continue
|
||||
if len(block.lines[-1].params) != 1:
|
||||
continue
|
||||
candidates[block.jump.getExceptSuccessors()[0]].append(block)
|
||||
|
||||
for child in self.blocks:
|
||||
if not candidates[child] or len(candidates[child]) < len(child.predecessors):
|
||||
continue
|
||||
|
||||
# Make sure caught exception is unused
|
||||
temp = candidates[child][0].lines[-1].outException
|
||||
if any(temp in phi.params for phi in child.phis):
|
||||
continue
|
||||
|
||||
for parent in candidates[child]:
|
||||
ephi = parent.lines.pop()
|
||||
throw_op = parent.lines.pop()
|
||||
del parent.unaryConstraints[throw_op.outException]
|
||||
del parent.unaryConstraints[ephi.outException]
|
||||
child.replacePredPair((parent, True), (parent, False))
|
||||
parent.jump = ssa_jumps.Goto(self, child)
|
||||
|
||||
# Subprocedure stuff #####################################################
|
||||
def _newBlockFrom(self, block):
|
||||
b = BasicBlock(next(self.block_numberer))
|
||||
self.blocks.append(b)
|
||||
return b
|
||||
|
||||
def _copyVar(self, var, vard=None):
|
||||
v = copy.copy(var)
|
||||
v.name = v.origin = None # TODO - generate new names?
|
||||
if vard is not None:
|
||||
vard[var] = v
|
||||
return v
|
||||
|
||||
def _region(self, proc):
|
||||
# Find the set of blocks 'in' a subprocedure, i.e. those reachable from the target that can reach the ret block
|
||||
region = graph_util.topologicalSort([proc.retblock], lambda block:[] if block == proc.target else [b for b,t in block.predecessors])
|
||||
temp = set(region)
|
||||
assert self.entryBlock not in temp and proc.target in temp and temp.isdisjoint(proc.jsrblocks)
|
||||
return region
|
||||
|
||||
def _duplicateBlocks(self, region, excludedPreds):
|
||||
# Duplicate a region of blocks. All inedges will be redirected to the new blocks
|
||||
# except for those from excludedPreds
|
||||
excludedPreds = excludedPreds | set(region)
|
||||
outsideBlocks = [b for b in self.blocks if b not in excludedPreds]
|
||||
|
||||
blockd, vard = {}, {}
|
||||
for oldb in region:
|
||||
block = blockd[oldb] = self._newBlockFrom(oldb)
|
||||
block.unaryConstraints = {self._copyVar(k, vard):v for k, v in oldb.unaryConstraints.items()}
|
||||
block.phis = [ssa_ops.Phi(block, vard[oldphi.rval]) for oldphi in oldb.phis]
|
||||
|
||||
for op in oldb.lines:
|
||||
new = copy.copy(op)
|
||||
new.replaceVars(vard)
|
||||
new.replaceOutVars(vard)
|
||||
assert new.getOutputs().count(None) == op.getOutputs().count(None)
|
||||
for outv in new.getOutputs():
|
||||
if outv is not None:
|
||||
assert outv.origin is None
|
||||
outv.origin = new
|
||||
block.lines.append(new)
|
||||
|
||||
assert set(vard).issuperset(oldb.jump.params)
|
||||
block.jump = oldb.jump.clone()
|
||||
block.jump.replaceVars(vard)
|
||||
|
||||
# Fix up blocks outside the region that jump into the region.
|
||||
for key in oldb.predecessors[:]:
|
||||
pred = key[0]
|
||||
if pred not in excludedPreds:
|
||||
for phi1, phi2 in zip(oldb.phis, block.phis):
|
||||
phi2.add(key, phi1.get(key))
|
||||
del phi1.dict[key]
|
||||
oldb.predecessors.remove(key)
|
||||
block.predecessors.append(key)
|
||||
|
||||
# fix up jump targets of newly created blocks
|
||||
for oldb, block in blockd.items():
|
||||
block.jump.replaceBlocks(blockd)
|
||||
for suc, t in block.jump.getSuccessorPairs():
|
||||
suc.predecessors.append((block, t))
|
||||
|
||||
# update the jump targets of predecessor blocks
|
||||
for block in outsideBlocks:
|
||||
block.jump.replaceBlocks(blockd)
|
||||
|
||||
for old, new in vard.items():
|
||||
assert type(old.origin) == type(new.origin)
|
||||
|
||||
# Fill in phi args in successors of new blocks
|
||||
for oldb, block in blockd.items():
|
||||
for oldc, t in oldb.jump.getSuccessorPairs():
|
||||
child = blockd.get(oldc, oldc)
|
||||
assert len(child.phis) == len(oldc.phis)
|
||||
for phi1, phi2 in zip(oldc.phis, child.phis):
|
||||
phi2.add((block, t), vard[phi1.get((oldb, t))])
|
||||
|
||||
assert self._conscheck() is None
|
||||
return blockd
|
||||
|
||||
def _splitSubProc(self, proc):
|
||||
# Splits a proc into two, with one callsite using the new proc instead
|
||||
# this involves duplicating the body of the procedure
|
||||
# the new proc is appended to the list of procs so it can work properly
|
||||
# with the stack processing in inlineSubprocs
|
||||
assert len(proc.jsrblocks) > 1
|
||||
target, retblock = proc.target, proc.retblock
|
||||
region = self._region(proc)
|
||||
|
||||
split_jsrs = [proc.jsrblocks.pop()]
|
||||
blockd = self._duplicateBlocks(region, set(proc.jsrblocks))
|
||||
|
||||
newproc = subproc.ProcInfo(blockd[proc.retblock], blockd[proc.target])
|
||||
newproc.jsrblocks = split_jsrs
|
||||
# Sanity check
|
||||
for temp in self.procs + [newproc]:
|
||||
for jsr in temp.jsrblocks:
|
||||
assert jsr.jump.target == temp.target
|
||||
return newproc
|
||||
|
||||
def _inlineSubProc(self, proc):
|
||||
# Inline a proc with single callsite inplace
|
||||
assert len(proc.jsrblocks) == 1
|
||||
target, retblock = proc.target, proc.retblock
|
||||
region = self._region(proc)
|
||||
|
||||
jsrblock = proc.jsrblocks[0]
|
||||
jsrop = jsrblock.jump
|
||||
ftblock = jsrop.fallthrough
|
||||
|
||||
# first we find any vars that bypass the proc since we have to pass them through the new blocks
|
||||
skipvars = [phi.get((jsrblock, False)) for phi in ftblock.phis]
|
||||
skipvars = [var for var in skipvars if var.origin is not jsrop]
|
||||
|
||||
svarcopy = {(var, block):self._copyVar(var) for var, block in itertools.product(skipvars, region)}
|
||||
for var, block in itertools.product(skipvars, region):
|
||||
# Create a new phi for the passed through var for this block
|
||||
rval = svarcopy[var, block]
|
||||
phi = ssa_ops.Phi(block, rval)
|
||||
block.phis.append(phi)
|
||||
block.unaryConstraints[rval] = jsrblock.unaryConstraints[var]
|
||||
|
||||
for key in block.predecessors:
|
||||
if key == (jsrblock, False):
|
||||
phi.add(key, var)
|
||||
else:
|
||||
phi.add(key, svarcopy[var, key[0]])
|
||||
|
||||
outreplace = {jv:rv for jv, rv in zip(jsrblock.jump.output.stack, retblock.jump.input.stack) if jv is not None}
|
||||
outreplace.update({jv:retblock.jump.input.locals[i] for i, jv in jsrblock.jump.output.locals.items() if jv is not None})
|
||||
for var in outreplace: # don't need jsrop's out vars anymore
|
||||
del jsrblock.unaryConstraints[var]
|
||||
|
||||
for var in skipvars:
|
||||
outreplace[var] = svarcopy[var, retblock]
|
||||
jsrblock.jump = ssa_jumps.Goto(self, target)
|
||||
retblock.jump = ssa_jumps.Goto(self, ftblock)
|
||||
|
||||
ftblock.replacePredPair((jsrblock, False), (retblock, False))
|
||||
for phi in ftblock.phis:
|
||||
phi.replaceVars(outreplace)
|
||||
|
||||
def inlineSubprocs(self):
|
||||
assert self._conscheck() is None
|
||||
assert self.procs
|
||||
|
||||
# establish DAG of subproc callstacks if we're doing nontrivial inlining, since we can only inline leaf procs
|
||||
regions = {proc:frozenset(self._region(proc)) for proc in self.procs}
|
||||
parents = {proc:[] for proc in self.procs}
|
||||
for x,y in itertools.product(self.procs, repeat=2):
|
||||
if not regions[y].isdisjoint(x.jsrblocks):
|
||||
parents[x].append(y)
|
||||
|
||||
self.procs = graph_util.topologicalSort(self.procs, parents.get)
|
||||
if any(parents.values()):
|
||||
print 'Warning, nesting subprocedures detected! This method may take a long time to decompile.'
|
||||
print 'Subprocedures for', self.code.method.name + ':', self.procs
|
||||
|
||||
# now inline the procs
|
||||
while self.procs:
|
||||
proc = self.procs.pop()
|
||||
while len(proc.jsrblocks) > 1:
|
||||
print 'splitting', proc
|
||||
# push new subproc onto stack
|
||||
self.procs.append(self._splitSubProc(proc))
|
||||
assert self._conscheck() is None
|
||||
# When a subprocedure has only one call point, it can just be inlined instead of splitted
|
||||
print 'inlining', proc
|
||||
self._inlineSubProc(proc)
|
||||
assert self._conscheck() is None
|
||||
##########################################################################
|
||||
def splitDualInedges(self):
|
||||
# Split any blocks that have both normal and exceptional in edges
|
||||
assert not self.procs
|
||||
for block in self.blocks[:]:
|
||||
if block is self.entryBlock:
|
||||
continue
|
||||
types = set(zip(*block.predecessors)[1])
|
||||
if len(types) <= 1:
|
||||
continue
|
||||
assert not isinstance(block.jump, (ssa_jumps.Return, ssa_jumps.Rethrow))
|
||||
|
||||
new = self._newBlockFrom(block)
|
||||
print 'Splitting', block, '->', new
|
||||
# first fix up CFG edges
|
||||
badpreds = [t for t in block.predecessors if t[1]]
|
||||
new.predecessors = badpreds
|
||||
for t in badpreds:
|
||||
block.predecessors.remove(t)
|
||||
|
||||
for pred, _ in badpreds:
|
||||
assert isinstance(pred.jump, ssa_jumps.OnException)
|
||||
pred.jump.replaceExceptTarget(block, new)
|
||||
|
||||
new.jump = ssa_jumps.Goto(self, block)
|
||||
block.predecessors.append((new, False))
|
||||
|
||||
# fix up variables
|
||||
new.phis = []
|
||||
new.unaryConstraints = {}
|
||||
for phi in block.phis:
|
||||
newrval = self._copyVar(phi.rval)
|
||||
new.unaryConstraints[newrval] = block.unaryConstraints[phi.rval]
|
||||
newphi = ssa_ops.Phi(new, newrval)
|
||||
new.phis.append(newphi)
|
||||
|
||||
for t in badpreds:
|
||||
arg = phi.get(t)
|
||||
phi.delete(t)
|
||||
newphi.add(t, arg)
|
||||
phi.add((new, False), newrval)
|
||||
assert self._conscheck() is None
|
||||
|
||||
def fixLoops(self):
|
||||
assert not self.procs
|
||||
todo = self.blocks[:]
|
||||
while todo:
|
||||
newtodo = []
|
||||
temp = set(todo)
|
||||
sccs = graph_util.tarjanSCC(todo, lambda block:[x for x,t in block.predecessors if x in temp])
|
||||
|
||||
for scc in sccs:
|
||||
if len(scc) <= 1:
|
||||
continue
|
||||
|
||||
scc_pair_set = {(x, False) for x in scc} | {(x, True) for x in scc}
|
||||
entries = [n for n in scc if not scc_pair_set.issuperset(n.predecessors)]
|
||||
|
||||
if len(entries) <= 1:
|
||||
head = entries[0]
|
||||
else:
|
||||
# if more than one entry point into the loop, we have to choose one as the head and duplicate the rest
|
||||
print 'Warning, multiple entry point loop detected. Generated code may be extremely large',
|
||||
print '({} entry points, {} blocks)'.format(len(entries), len(scc))
|
||||
def loopSuccessors(head, block):
|
||||
if block == head:
|
||||
return []
|
||||
return [x for x in block.jump.getSuccessors() if (x, False) in scc_pair_set]
|
||||
|
||||
reaches = [(n, graph_util.topologicalSort(entries, functools.partial(loopSuccessors, n))) for n in scc]
|
||||
for head, reachable in reaches:
|
||||
reachable.remove(head)
|
||||
|
||||
head, reachable = min(reaches, key=lambda t:(len(t[1]), -len(t[0].predecessors)))
|
||||
assert head not in reachable
|
||||
print 'Duplicating {} nodes'.format(len(reachable))
|
||||
blockd = self._duplicateBlocks(reachable, set(scc) - set(reachable))
|
||||
newtodo += map(blockd.get, reachable)
|
||||
newtodo.extend(scc)
|
||||
newtodo.remove(head)
|
||||
todo = newtodo
|
||||
assert self._conscheck() is None
|
||||
|
||||
# Functions called by children ###########################################
|
||||
# assign variable names for debugging
|
||||
varnum = collections.defaultdict(itertools.count)
|
||||
def makeVariable(self, *args, **kwargs):
|
||||
# Note: Make sure this doesn't hold on to created variables in any way,
|
||||
# since this func may be called for temporary results that are discarded
|
||||
var = SSA_Variable(*args, **kwargs)
|
||||
# pref = args[0][0][0].replace('o','a')
|
||||
# var.name = pref + str(next(self.varnum[pref]))
|
||||
return var
|
||||
|
||||
def setObjVarData(self, var, vtype, initMap):
|
||||
vtype2 = initMap.get(vtype, vtype)
|
||||
tt = objtypes.verifierToSynthetic(vtype2)
|
||||
assert var.decltype is None or var.decltype == tt
|
||||
var.decltype = tt
|
||||
# if uninitialized, record the offset of originating new instruction for later
|
||||
if vtype.tag == '.new':
|
||||
assert var.uninit_orig_num is None or var.uninit_orig_num == vtype.extra
|
||||
var.uninit_orig_num = vtype.extra
|
||||
|
||||
def makeVarFromVtype(self, vtype, initMap):
|
||||
vtype2 = initMap.get(vtype, vtype)
|
||||
type_ = verifierToSSAType(vtype2)
|
||||
if type_ is not None:
|
||||
var = self.makeVariable(type_)
|
||||
if type_ == SSA_OBJECT:
|
||||
self.setObjVarData(var, vtype, initMap)
|
||||
return var
|
||||
return None
|
||||
|
||||
def getConstPoolArgs(self, index):
|
||||
return self.class_.cpool.getArgs(index)
|
||||
|
||||
def getConstPoolType(self, index):
|
||||
return self.class_.cpool.getType(index)
|
||||
|
||||
def ssaFromVerified(code, iNodes, opts):
|
||||
method = code.method
|
||||
inputTypes, returnTypes = parseUnboundMethodDescriptor(method.descriptor, method.class_.name, method.static)
|
||||
|
||||
parent = SSA_Graph(code)
|
||||
data = blockmaker.BlockMaker(parent, iNodes, inputTypes, returnTypes, code.except_raw, opts=opts)
|
||||
|
||||
parent.blocks = blocks = data.blocks
|
||||
parent.entryBlock = data.entryBlock
|
||||
parent.inputArgs = data.inputArgs
|
||||
assert parent.entryBlock in blocks
|
||||
|
||||
# create subproc info
|
||||
procd = {block.jump.target: subproc.ProcInfo(block, block.jump.target) for block in blocks if isinstance(block.jump, subproc.DummyRet)}
|
||||
for block in blocks:
|
||||
if isinstance(block.jump, subproc.ProcCallOp):
|
||||
procd[block.jump.target].jsrblocks.append(block)
|
||||
parent.procs = sorted(procd.values(), key=lambda p:p.target.key)
|
||||
|
||||
# Intern constraints to save a bit of memory for long methods
|
||||
def makeConstraint(var, _cache={}):
|
||||
key = var.type, var.const, var.decltype, var.uninit_orig_num is None
|
||||
try:
|
||||
return _cache[key]
|
||||
except KeyError:
|
||||
_cache[key] = temp = constraints.fromVariable(parent.env, var)
|
||||
return temp
|
||||
|
||||
# create unary constraints for each variable
|
||||
for block in blocks:
|
||||
bvars = []
|
||||
if isinstance(block.jump, subproc.ProcCallOp):
|
||||
bvars += block.jump.flatOutput()
|
||||
# entry block has no phis
|
||||
if block is parent.entryBlock:
|
||||
bvars += parent.inputArgs
|
||||
|
||||
bvars = [v for v in bvars if v is not None]
|
||||
bvars += [phi.rval for phi in block.phis]
|
||||
for op in block.lines:
|
||||
bvars += op.params
|
||||
bvars += [x for x in op.getOutputs() if x is not None]
|
||||
bvars += block.jump.params
|
||||
|
||||
for suc, t in block.jump.getSuccessorPairs():
|
||||
for phi in suc.phis:
|
||||
bvars.append(phi.get((block, t)))
|
||||
assert None not in bvars
|
||||
# Note that makeConstraint can indirectly cause class loading
|
||||
block.unaryConstraints = {var:makeConstraint(var) for var in bvars}
|
||||
|
||||
parent._conscheck()
|
||||
return parent
|
||||
|
|
@ -0,0 +1,6 @@
|
|||
class ValueType(object):
|
||||
'''Define _key() and inherit from this class to implement comparison and hashing'''
|
||||
# def __init__(self, *args, **kwargs): super(ValueType, self).__init__(*args, **kwargs)
|
||||
def __eq__(self, other): return type(self) == type(other) and self._key() == other._key()
|
||||
def __ne__(self, other): return type(self) != type(other) or self._key() != other._key()
|
||||
def __hash__(self): return hash(self._key())
|
||||
|
|
@ -0,0 +1,126 @@
|
|||
from ..verifier import verifier_types as vtypes
|
||||
|
||||
# types are represented by classname, dimension
|
||||
# primitive types are .int, etc since these cannot be valid classnames since periods are forbidden
|
||||
def TypeTT(baset, dim):
|
||||
assert dim >= 0
|
||||
return baset, dim
|
||||
|
||||
# Not real types
|
||||
VoidTT = TypeTT('.void', 0)
|
||||
NullTT = TypeTT('.null', 0)
|
||||
|
||||
ObjectTT = TypeTT('java/lang/Object', 0)
|
||||
StringTT = TypeTT('java/lang/String', 0)
|
||||
ThrowableTT = TypeTT('java/lang/Throwable', 0)
|
||||
ClassTT = TypeTT('java/lang/Class', 0)
|
||||
|
||||
BoolTT = TypeTT('.boolean', 0)
|
||||
IntTT = TypeTT('.int', 0)
|
||||
LongTT = TypeTT('.long', 0)
|
||||
FloatTT = TypeTT('.float', 0)
|
||||
DoubleTT = TypeTT('.double', 0)
|
||||
|
||||
ByteTT = TypeTT('.byte', 0)
|
||||
CharTT = TypeTT('.char', 0)
|
||||
ShortTT = TypeTT('.short', 0)
|
||||
|
||||
BExpr = '.bexpr' # bool or byte
|
||||
|
||||
def baset(tt): return tt[0]
|
||||
def dim(tt): return tt[1]
|
||||
def withDimInc(tt, inc): return TypeTT(baset(tt), dim(tt)+inc)
|
||||
def withNoDim(tt): return TypeTT(baset(tt), 0)
|
||||
|
||||
def isBaseTClass(tt): return not baset(tt).startswith('.')
|
||||
def className(tt): return baset(tt) if not baset(tt).startswith('.') else None
|
||||
def primName(tt): return baset(tt)[1:] if baset(tt).startswith('.') else None
|
||||
|
||||
###############################################################################
|
||||
|
||||
def isSubtype(env, x, y):
|
||||
if x == y or y == ObjectTT or x == NullTT:
|
||||
return True
|
||||
elif y == NullTT:
|
||||
return False
|
||||
|
||||
xname, xdim = baset(x), dim(x)
|
||||
yname, ydim = baset(y), dim(y)
|
||||
if ydim > xdim:
|
||||
return False
|
||||
elif xdim > ydim: # TODO - these constants should be defined in one place to reduce risk of typos
|
||||
return yname in ('java/lang/Object','java/lang/Cloneable','java/io/Serializable')
|
||||
else:
|
||||
return isBaseTClass(x) and isBaseTClass(y) and env.isSubclass(xname, yname)
|
||||
|
||||
# Will not return interface unless all inputs are same interface or null
|
||||
def commonSupertype(env, tts):
|
||||
assert(hasattr(env, 'getClass')) # catch common errors where we forget the env argument
|
||||
|
||||
tts = set(tts)
|
||||
tts.discard(NullTT)
|
||||
|
||||
if len(tts) == 1:
|
||||
return tts.pop()
|
||||
elif not tts:
|
||||
return NullTT
|
||||
|
||||
dims = map(dim, tts)
|
||||
newdim = min(dims)
|
||||
if max(dims) > newdim or any(baset(tt) == 'java/lang/Object' for tt in tts):
|
||||
return TypeTT('java/lang/Object', newdim)
|
||||
# if any are primitive arrays, result is object array of dim-1
|
||||
if not all(isBaseTClass(tt) for tt in tts):
|
||||
return TypeTT('java/lang/Object', newdim-1)
|
||||
|
||||
# find common superclass of base types
|
||||
bases = sorted(map(baset, tts))
|
||||
superclass = reduce(env.commonSuperclass, bases)
|
||||
return TypeTT(superclass, newdim)
|
||||
|
||||
######################################################################################################
|
||||
_verifierConvert = {vtypes.T_INT:IntTT, vtypes.T_FLOAT:FloatTT, vtypes.T_LONG:LongTT,
|
||||
vtypes.T_DOUBLE:DoubleTT, vtypes.T_SHORT:ShortTT, vtypes.T_CHAR:CharTT,
|
||||
vtypes.T_BYTE:ByteTT, vtypes.T_BOOL:BoolTT, vtypes.T_NULL:NullTT,
|
||||
vtypes.OBJECT_INFO:ObjectTT}
|
||||
|
||||
def verifierToSynthetic_seq(vts):
|
||||
return [verifierToSynthetic(vt) for vt in vts if vt != vtypes.T_INVALID]
|
||||
|
||||
def verifierToSynthetic(vtype):
|
||||
assert vtype.tag not in (None, '.address', '.new', '.init')
|
||||
vtype = vtypes.withNoConst(vtype)
|
||||
|
||||
if vtype in _verifierConvert:
|
||||
return _verifierConvert[vtype]
|
||||
|
||||
base = vtypes.withNoDimension(vtype)
|
||||
if base in _verifierConvert:
|
||||
return withDimInc(_verifierConvert[base], vtype.dim)
|
||||
|
||||
return TypeTT(vtype.extra, vtype.dim)
|
||||
|
||||
# returns supers, exacts
|
||||
def declTypeToActual(env, decltype):
|
||||
name, newdim = baset(decltype), dim(decltype)
|
||||
|
||||
# Verifier treats bool[]s and byte[]s as interchangeable, so it could really be either
|
||||
if newdim and (name == baset(ByteTT) or name == baset(BoolTT)):
|
||||
return [], [withDimInc(ByteTT, newdim), withDimInc(BoolTT, newdim)]
|
||||
elif not isBaseTClass(decltype): # primitive types can't be subclassed anyway
|
||||
return [], [decltype]
|
||||
|
||||
# Verifier doesn't fully verify interfaces so they could be anything
|
||||
if env.isInterface(name):
|
||||
return [withDimInc(ObjectTT, newdim)], []
|
||||
# If class is final, return it as exact, not super
|
||||
elif env.isFinal(name):
|
||||
return [], [decltype]
|
||||
else:
|
||||
return [decltype], []
|
||||
|
||||
def removeInterface(env, decltype):
|
||||
name, newdim = baset(decltype), dim(decltype)
|
||||
if isBaseTClass(decltype) and env.isInterface(name):
|
||||
return withDimInc(ObjectTT, newdim)
|
||||
return decltype
|
||||
|
|
@ -0,0 +1,10 @@
|
|||
from .base import BaseJump
|
||||
|
||||
from .onexception import OnException
|
||||
from .goto import Goto
|
||||
from .ifcmp import If
|
||||
from .exit import Return, Rethrow
|
||||
from .switch import Switch
|
||||
|
||||
from . import placeholder
|
||||
OnAbscond = Ret = placeholder.Placeholder
|
||||
|
|
@ -0,0 +1,18 @@
|
|||
import copy
|
||||
|
||||
from ..functionbase import SSAFunctionBase
|
||||
|
||||
class BaseJump(SSAFunctionBase):
|
||||
def __init__(self, parent, arguments=()):
|
||||
super(BaseJump, self).__init__(parent,arguments)
|
||||
|
||||
def replaceBlocks(self, blockDict):
|
||||
assert not self.getSuccessors()
|
||||
|
||||
def getNormalSuccessors(self): return []
|
||||
def getExceptSuccessors(self): return []
|
||||
def getSuccessors(self): return self.getNormalSuccessors() + self.getExceptSuccessors()
|
||||
def getSuccessorPairs(self): return [(x,False) for x in self.getNormalSuccessors()] + [(x,True) for x in self.getExceptSuccessors()]
|
||||
def reduceSuccessors(self, pairsToRemove): return self
|
||||
|
||||
def clone(self): return copy.copy(self) # overriden by classes which need to do a deep copy
|
||||
|
|
@ -0,0 +1,9 @@
|
|||
from .base import BaseJump
|
||||
|
||||
class Return(BaseJump):
|
||||
def __init__(self, parent, arguments):
|
||||
super(Return, self).__init__(parent, arguments)
|
||||
|
||||
class Rethrow(BaseJump):
|
||||
def __init__(self, parent, arguments):
|
||||
super(Rethrow, self).__init__(parent, arguments)
|
||||
|
|
@ -0,0 +1,17 @@
|
|||
from .base import BaseJump
|
||||
|
||||
class Goto(BaseJump):
|
||||
def __init__(self, parent, target):
|
||||
super(Goto, self).__init__(parent, [])
|
||||
self.successors = [target]
|
||||
|
||||
def replaceBlocks(self, blockDict):
|
||||
self.successors = [blockDict.get(key,key) for key in self.successors]
|
||||
|
||||
def getNormalSuccessors(self):
|
||||
return self.successors
|
||||
|
||||
def reduceSuccessors(self, pairsToRemove):
|
||||
if (self.successors[0], False) in pairsToRemove:
|
||||
return None
|
||||
return self
|
||||
|
|
@ -0,0 +1,101 @@
|
|||
from .. import ssa_types
|
||||
from ..constraints import IntConstraint, ObjectConstraint
|
||||
|
||||
from .base import BaseJump
|
||||
from .goto import Goto
|
||||
|
||||
class If(BaseJump):
|
||||
opposites = {'eq':'ne', 'ne':'eq', 'lt':'ge', 'ge':'lt', 'gt':'le', 'le':'gt'}
|
||||
|
||||
def __init__(self, parent, cmp, successors, arguments):
|
||||
super(If, self).__init__(parent, arguments)
|
||||
assert cmp in ('eq','ne','lt','ge','gt','le')
|
||||
self.cmp = cmp
|
||||
self.successors = successors
|
||||
self.isObj = (arguments[0].type == ssa_types.SSA_OBJECT)
|
||||
assert None not in successors
|
||||
|
||||
def replaceBlocks(self, blockDict):
|
||||
self.successors = [blockDict.get(key,key) for key in self.successors]
|
||||
|
||||
def getNormalSuccessors(self):
|
||||
return self.successors
|
||||
|
||||
def reduceSuccessors(self, pairsToRemove):
|
||||
temp = set(self.successors)
|
||||
for (child, t) in pairsToRemove:
|
||||
temp.remove(child)
|
||||
|
||||
if len(temp) == 0:
|
||||
return None
|
||||
elif len(temp) == 1:
|
||||
return Goto(self.parent, temp.pop())
|
||||
return self
|
||||
|
||||
###############################################################################
|
||||
def constrainJumps(self, x, y):
|
||||
impossible = []
|
||||
for child in self.successors:
|
||||
func = self.getSuccessorConstraints((child,False))
|
||||
|
||||
results = func(x,y)
|
||||
if None in results:
|
||||
assert results == (None,None)
|
||||
impossible.append((child,False))
|
||||
return self.reduceSuccessors(impossible)
|
||||
|
||||
def getSuccessorConstraints(self, (block, t)):
|
||||
assert t is False
|
||||
cmp_t = If.opposites[self.cmp] if block == self.successors[0] else self.cmp
|
||||
|
||||
if self.isObj:
|
||||
def propagateConstraints_obj(x, y):
|
||||
if x is None or y is None:
|
||||
return None, None
|
||||
if cmp_t == 'eq':
|
||||
z = x.join(y)
|
||||
return z,z
|
||||
else:
|
||||
x2, y2 = x, y
|
||||
if x.isConstNull():
|
||||
yt = y.types
|
||||
y2 = ObjectConstraint.fromTops(yt.env, yt.supers, yt.exact, nonnull=True)
|
||||
if y.isConstNull():
|
||||
xt = x.types
|
||||
x2 = ObjectConstraint.fromTops(xt.env, xt.supers, xt.exact, nonnull=True)
|
||||
return x2, y2
|
||||
return propagateConstraints_obj
|
||||
else:
|
||||
def propagateConstraints_int(x, y):
|
||||
if x is None or y is None:
|
||||
return None, None
|
||||
x1, x2, y1, y2 = x.min, x.max, y.min, y.max
|
||||
if cmp_t == 'ge' or cmp_t == 'gt':
|
||||
x1, x2, y1, y2 = y1, y2, x1, x2
|
||||
|
||||
# treat greater like less than swap before and afterwards
|
||||
if cmp_t == 'lt' or cmp_t == 'gt':
|
||||
x2 = min(x2, y2-1)
|
||||
y1 = max(x1+1, y1)
|
||||
elif cmp_t == 'le' or cmp_t == 'ge':
|
||||
x2 = min(x2, y2)
|
||||
y1 = max(x1, y1)
|
||||
elif cmp_t == 'eq':
|
||||
x1 = y1 = max(x1, y1)
|
||||
x2 = y2 = min(x2, y2)
|
||||
elif cmp_t == 'ne':
|
||||
if x1 == x2 == y1 == y2:
|
||||
return None, None
|
||||
if x1 == x2:
|
||||
y1 = y1 if y1 != x1 else y1+1
|
||||
y2 = y2 if y2 != x2 else y2-1
|
||||
if y1 == y2:
|
||||
x1 = x1 if x1 != y1 else x1+1
|
||||
x2 = x2 if x2 != y2 else x2-1
|
||||
|
||||
if cmp_t == 'ge' or cmp_t == 'gt':
|
||||
x1, x2, y1, y2 = y1, y2, x1, x2
|
||||
con1 = IntConstraint.range(x.width, x1, x2) if x1 <= x2 else None
|
||||
con2 = IntConstraint.range(y.width, y1, y2) if y1 <= y2 else None
|
||||
return con1, con2
|
||||
return propagateConstraints_int
|
||||
|
|
@ -0,0 +1,59 @@
|
|||
from .. import objtypes
|
||||
from ..constraints import ObjectConstraint
|
||||
from ..exceptionset import CatchSetManager, ExceptionSet
|
||||
|
||||
from .base import BaseJump
|
||||
from .goto import Goto
|
||||
|
||||
class OnException(BaseJump):
|
||||
def __init__(self, parent, throwvar, chpairs, fallthrough=None):
|
||||
super(OnException, self).__init__(parent, [throwvar])
|
||||
self.default = fallthrough
|
||||
self.cs = CatchSetManager.new(parent.env, chpairs)
|
||||
self.cs.pruneKeys()
|
||||
|
||||
def replaceExceptTarget(self, old, new):
|
||||
self.cs.replaceKeys({old:new})
|
||||
|
||||
def replaceNormalTarget(self, old, new):
|
||||
self.default = new if self.default == old else self.default
|
||||
|
||||
def replaceBlocks(self, blockDict):
|
||||
self.cs.replaceKeys(blockDict)
|
||||
if self.default is not None:
|
||||
self.default = blockDict.get(self.default, self.default)
|
||||
|
||||
def reduceSuccessors(self, pairsToRemove):
|
||||
for (child, t) in pairsToRemove:
|
||||
if t:
|
||||
self.cs.mask -= self.cs.sets[child]
|
||||
del self.cs.sets[child]
|
||||
else:
|
||||
self.replaceNormalTarget(child, None)
|
||||
|
||||
self.cs.pruneKeys()
|
||||
if not self.cs.sets:
|
||||
if not self.default:
|
||||
return None
|
||||
return Goto(self.parent, self.default)
|
||||
return self
|
||||
|
||||
def getNormalSuccessors(self):
|
||||
return [self.default] if self.default is not None else []
|
||||
|
||||
def getExceptSuccessors(self):
|
||||
return self.cs.sets.keys()
|
||||
|
||||
def clone(self):
|
||||
new = super(OnException, self).clone()
|
||||
new.cs = self.cs.copy()
|
||||
return new
|
||||
|
||||
###############################################################################
|
||||
def constrainJumps(self, x):
|
||||
if x is None:
|
||||
mask = ExceptionSet.EMPTY
|
||||
else:
|
||||
mask = ExceptionSet(x.types.env, [(objtypes.className(tt),()) for tt in x.types.supers | x.types.exact])
|
||||
self.cs.newMask(mask)
|
||||
return self.reduceSuccessors([])
|
||||
|
|
@ -0,0 +1,5 @@
|
|||
from .base import BaseJump
|
||||
|
||||
class Placeholder(BaseJump):
|
||||
def __init__(self, parent, *args, **kwargs):
|
||||
super(Placeholder, self).__init__(parent)
|
||||
|
|
@ -0,0 +1,86 @@
|
|||
import collections
|
||||
|
||||
from ..constraints import IntConstraint
|
||||
from ..ssa_types import SSA_INT
|
||||
|
||||
from .base import BaseJump
|
||||
from .goto import Goto
|
||||
from .ifcmp import If
|
||||
|
||||
class Switch(BaseJump):
|
||||
def __init__(self, parent, default, table, arguments):
|
||||
super(Switch, self).__init__(parent, arguments)
|
||||
|
||||
# get ordered successors since our map will be unordered. Default is always first successor
|
||||
if not table:
|
||||
ordered = [default]
|
||||
else:
|
||||
tset = set()
|
||||
ordered = [x for x in (default,) + zip(*table)[1] if not x in tset and not tset.add(x)]
|
||||
|
||||
self.successors = ordered
|
||||
reverse = collections.defaultdict(set)
|
||||
for k,v in table:
|
||||
if v != default:
|
||||
reverse[v].add(k)
|
||||
self.reverse = {k: frozenset(v) for k, v in reverse.items()}
|
||||
|
||||
def getNormalSuccessors(self):
|
||||
return self.successors
|
||||
|
||||
def replaceBlocks(self, blockDict):
|
||||
self.successors = [blockDict.get(key,key) for key in self.successors]
|
||||
self.reverse = {blockDict.get(k,k):v for k,v in self.reverse.items()}
|
||||
|
||||
def reduceSuccessors(self, pairsToRemove):
|
||||
temp = list(self.successors)
|
||||
for (child, t) in pairsToRemove:
|
||||
temp.remove(child)
|
||||
|
||||
if len(temp) == 0:
|
||||
return None
|
||||
elif len(temp) == 1:
|
||||
return Goto(self.parent, temp.pop())
|
||||
|
||||
if len(temp) < len(self.successors):
|
||||
self.successors = temp
|
||||
self.reverse = {v:self.reverse[v] for v in temp[1:]}
|
||||
return self
|
||||
|
||||
def simplifyToIf(self, block):
|
||||
# Try to replace with an if statement if possible
|
||||
# e.g. switch(x) {case C: ... default: ...} -> if (x == C) {...} else {...}
|
||||
if len(self.successors) == 2:
|
||||
cases = self.reverse[self.successors[-1]]
|
||||
if len(cases) == 1:
|
||||
const = self.parent.makeVariable(SSA_INT)
|
||||
const.const = min(cases)
|
||||
block.unaryConstraints[const] = IntConstraint.const(32, const.const)
|
||||
return If(self.parent, 'eq', self.successors, self.params + [const])
|
||||
return self
|
||||
|
||||
###############################################################################
|
||||
def constrainJumps(self, x):
|
||||
impossible = []
|
||||
for child in self.successors:
|
||||
func = self.getSuccessorConstraints((child,False))
|
||||
results = func(x)
|
||||
if results[0] is None:
|
||||
impossible.append((child,False))
|
||||
return self.reduceSuccessors(impossible)
|
||||
|
||||
def getSuccessorConstraints(self, (block, t)):
|
||||
if block in self.reverse:
|
||||
cmin = min(self.reverse[block])
|
||||
cmax = max(self.reverse[block])
|
||||
def propagateConstraints(x):
|
||||
if x is None:
|
||||
return None,
|
||||
return IntConstraint.range(x.width, max(cmin, x.min), min(cmax, x.max)),
|
||||
else:
|
||||
allcases = set().union(*self.reverse.values())
|
||||
def propagateConstraints(x):
|
||||
if x is None or (x.min == x.max and x.min in allcases):
|
||||
return None,
|
||||
return x,
|
||||
return propagateConstraints
|
||||
|
|
@ -0,0 +1,16 @@
|
|||
from .base import BaseOp
|
||||
|
||||
from .array import ArrLoad, ArrStore, ArrLength
|
||||
from .checkcast import CheckCast, InstanceOf
|
||||
from .convert import Convert
|
||||
from .fieldaccess import FieldAccess
|
||||
from .fmath import FAdd, FDiv, FMul, FRem, FSub, FNeg, FCmp
|
||||
from .invoke import Invoke, InvokeDynamic
|
||||
from .imath import IAdd, IDiv, IMul, IRem, ISub, IAnd, IOr, IShl, IShr, IUshr, IXor, ICmp
|
||||
from .monitor import Monitor
|
||||
from .new import New, NewArray, MultiNewArray
|
||||
from .throw import Throw, MagicThrow
|
||||
from .truncate import Truncate
|
||||
from .tryreturn import TryReturn
|
||||
|
||||
from .phi import Phi, ExceptionPhi
|
||||
|
|
@ -0,0 +1,79 @@
|
|||
from .. import excepttypes, objtypes
|
||||
from ..constraints import FloatConstraint, IntConstraint, ObjectConstraint, maybeThrow, returnOrThrow, throw
|
||||
from ..ssa_types import SSA_INT
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
def getElementTypes(env, tops):
|
||||
types = [objtypes.withDimInc(tt, -1) for tt in tops]
|
||||
# temporary hack
|
||||
types = [objtypes.removeInterface(env, tt) for tt in types]
|
||||
|
||||
supers = [tt for tt in types if objtypes.isBaseTClass(tt)]
|
||||
exact = [tt for tt in types if not objtypes.isBaseTClass(tt)]
|
||||
return ObjectConstraint.fromTops(env, supers, exact)
|
||||
|
||||
class ArrLoad(BaseOp):
|
||||
def __init__(self, parent, args, ssatype):
|
||||
super(ArrLoad, self).__init__(parent, args, makeException=True)
|
||||
self.env = parent.env
|
||||
self.rval = parent.makeVariable(ssatype, origin=self)
|
||||
self.ssatype = ssatype
|
||||
|
||||
def propagateConstraints(self, a, i):
|
||||
etypes = (excepttypes.ArrayOOB,)
|
||||
if a.null:
|
||||
etypes += (excepttypes.NullPtr,)
|
||||
if a.isConstNull():
|
||||
return throw(ObjectConstraint.fromTops(self.env, [], [excepttypes.NullPtr], nonnull=True))
|
||||
|
||||
if self.ssatype[0] == 'int':
|
||||
rout = IntConstraint.bot(self.ssatype[1])
|
||||
elif self.ssatype[0] == 'float':
|
||||
rout = FloatConstraint.bot(self.ssatype[1])
|
||||
elif self.ssatype[0] == 'obj':
|
||||
rout = getElementTypes(self.env, a.types.supers | a.types.exact)
|
||||
|
||||
eout = ObjectConstraint.fromTops(self.env, [], etypes, nonnull=True)
|
||||
return returnOrThrow(rout, eout)
|
||||
|
||||
class ArrStore(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, args):
|
||||
super(ArrStore, self).__init__(parent, args, makeException=True)
|
||||
self.env = parent.env
|
||||
|
||||
def propagateConstraints(self, a, i, x):
|
||||
etypes = (excepttypes.ArrayOOB,)
|
||||
if a.null:
|
||||
etypes += (excepttypes.NullPtr,)
|
||||
if a.isConstNull():
|
||||
return throw(ObjectConstraint.fromTops(self.env, [], [excepttypes.NullPtr], nonnull=True))
|
||||
|
||||
if isinstance(x, ObjectConstraint):
|
||||
# If the type of a is known exactly to be the single possibility T[]
|
||||
# and x is assignable to T, we can assume there is no ArrayStore exception
|
||||
# if a's type has multiple possibilities, then there can be an exception
|
||||
known_type = a.types.exact if len(a.types.exact) == 1 else frozenset()
|
||||
allowed = getElementTypes(self.env, known_type)
|
||||
if allowed.meet(x) != allowed:
|
||||
etypes += (excepttypes.ArrayStore,)
|
||||
|
||||
return maybeThrow(ObjectConstraint.fromTops(self.env, [], etypes, nonnull=True))
|
||||
|
||||
class ArrLength(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(ArrLength, self).__init__(parent, args, makeException=True)
|
||||
self.env = parent.env
|
||||
self.rval = parent.makeVariable(SSA_INT, origin=self)
|
||||
|
||||
def propagateConstraints(self, x):
|
||||
etypes = ()
|
||||
if x.null:
|
||||
etypes += (excepttypes.NullPtr,)
|
||||
if x.isConstNull():
|
||||
return throw(ObjectConstraint.fromTops(self.env, [], [excepttypes.NullPtr], nonnull=True))
|
||||
|
||||
excons = ObjectConstraint.fromTops(self.env, [], etypes, nonnull=True)
|
||||
return returnOrThrow(IntConstraint.range(32, 0, (1<<31)-1), excons)
|
||||
|
|
@ -0,0 +1,30 @@
|
|||
from ..functionbase import SSAFunctionBase
|
||||
from ..ssa_types import SSA_OBJECT
|
||||
|
||||
class BaseOp(SSAFunctionBase):
|
||||
has_side_effects = False
|
||||
|
||||
def __init__(self, parent, arguments, makeException=False):
|
||||
super(BaseOp, self).__init__(parent, arguments)
|
||||
|
||||
self.rval = None
|
||||
self.outException = None
|
||||
|
||||
if makeException:
|
||||
self.outException = parent.makeVariable(SSA_OBJECT, origin=self)
|
||||
|
||||
def getOutputs(self):
|
||||
return self.rval, self.outException
|
||||
|
||||
def removeOutput(self, var):
|
||||
outs = self.rval, self.outException
|
||||
assert var is not None and var in outs
|
||||
self.rval, self.outException = [(x if x != var else None) for x in outs]
|
||||
|
||||
def replaceOutVars(self, vardict):
|
||||
self.rval, self.outException = map(vardict.get, (self.rval, self.outException))
|
||||
|
||||
# Given input constraints, return constraints on outputs. Output is (rval, exception)
|
||||
# With None returned for unused or impossible values. This should only be defined if it is
|
||||
# actually implemented.
|
||||
# def propagateConstraints(self, *cons):
|
||||
|
|
@ -0,0 +1,54 @@
|
|||
import itertools
|
||||
import operator
|
||||
|
||||
from ..constraints import IntConstraint
|
||||
|
||||
def split_pow2ranges(x,y):
|
||||
'''split given range into power of two ranges of form [x, x+2^k)'''
|
||||
out = []
|
||||
while x<=y:
|
||||
# The largest power of two range of the form x,k
|
||||
# has k min of number of zeros at end of x
|
||||
# and the largest power of two that fits in y-x
|
||||
bx = bin(x)
|
||||
numzeroes = float('inf') if x==0 else (len(bx)-bx.rindex('1')-1)
|
||||
k = min(numzeroes, (y-x+1).bit_length()-1)
|
||||
out.append((x,k))
|
||||
x += 1<<k
|
||||
assert x == y+1
|
||||
return out
|
||||
|
||||
def propagateBitwise(arg1, arg2, op, usemin, usemax):
|
||||
ranges1 = split_pow2ranges(arg1.min, arg1.max)
|
||||
ranges2 = split_pow2ranges(arg2.min, arg2.max)
|
||||
|
||||
vals = []
|
||||
for (s1,k1),(s2,k2) in itertools.product(ranges1, ranges2):
|
||||
# there are three parts. The high bits fixed in both arguments,
|
||||
# the middle bits fixed in one but not the other, and the
|
||||
# lowest bits which can be chosen freely for both arguments
|
||||
# high = op(h1,h2) and low goes from 0 to 1... but the range of
|
||||
# the middle depends on the particular operation
|
||||
# 0-x, x-1 and 0-1 for and, or, and xor respectively
|
||||
if k1 > k2:
|
||||
(s1,k1),(s2,k2) = (s2,k2),(s1,k1)
|
||||
|
||||
mask1 = (1<<k1) - 1
|
||||
mask2 = (1<<k2) - 1 - mask1
|
||||
|
||||
high = op(s1, s2) & ~(mask1 | mask2)
|
||||
midmin = (s1 & mask2) if usemin else 0
|
||||
midmax = (s1 & mask2) if usemax else mask2
|
||||
|
||||
vals.append(high | midmin)
|
||||
vals.append(high | midmax | mask1)
|
||||
return IntConstraint.range(arg1.width, min(vals), max(vals))
|
||||
|
||||
def propagateAnd(x, y):
|
||||
return propagateBitwise(x, y, operator.__and__, False, True)
|
||||
|
||||
def propagateOr(x, y):
|
||||
return propagateBitwise(x, y, operator.__or__, True, False)
|
||||
|
||||
def propagateXor( x, y):
|
||||
return propagateBitwise(x, y, operator.__xor__, False, False)
|
||||
|
|
@ -0,0 +1,40 @@
|
|||
from .. import excepttypes, objtypes, ssa_types
|
||||
from ..constraints import IntConstraint, ObjectConstraint, join, maybeThrow, return_, throw
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class CheckCast(BaseOp):
|
||||
def __init__(self, parent, target, args):
|
||||
super(CheckCast, self).__init__(parent, args, makeException=True)
|
||||
self.env = parent.env
|
||||
self.target_tt = target
|
||||
# Temporary hack
|
||||
target = objtypes.removeInterface(self.env, target)
|
||||
|
||||
if objtypes.isBaseTClass(target):
|
||||
self.outCasted = ObjectConstraint.fromTops(parent.env, [target], [])
|
||||
else:
|
||||
# Primative array types need to be in exact, not supers
|
||||
self.outCasted = ObjectConstraint.fromTops(parent.env, [], [target])
|
||||
self.outExceptionCons = ObjectConstraint.fromTops(parent.env, [], (excepttypes.ClassCast,), nonnull=True)
|
||||
|
||||
def propagateConstraints(self, x):
|
||||
intersect = join(x, self.outCasted)
|
||||
if intersect is None:
|
||||
return throw(self.outExceptionCons)
|
||||
elif intersect != x:
|
||||
assert not x.isConstNull()
|
||||
return maybeThrow(self.outExceptionCons)
|
||||
else:
|
||||
return return_(None)
|
||||
|
||||
class InstanceOf(BaseOp):
|
||||
def __init__(self, parent, target, args):
|
||||
super(InstanceOf, self).__init__(parent, args)
|
||||
self.env = parent.env
|
||||
self.target_tt = target
|
||||
self.rval = parent.makeVariable(ssa_types.SSA_INT, origin=self)
|
||||
|
||||
def propagateConstraints(self, x):
|
||||
rvalcons = IntConstraint.range(32, 0, 1)
|
||||
return return_(rvalcons)
|
||||
|
|
@ -0,0 +1,8 @@
|
|||
from .base import BaseOp
|
||||
|
||||
class Convert(BaseOp):
|
||||
def __init__(self, parent, arg, source_ssa, target_ssa):
|
||||
super(Convert, self).__init__(parent, [arg])
|
||||
self.source = source_ssa
|
||||
self.target = target_ssa
|
||||
self.rval = parent.makeVariable(target_ssa, origin=self)
|
||||
|
|
@ -0,0 +1,62 @@
|
|||
from ...verifier.descriptors import parseFieldDescriptor
|
||||
|
||||
from .. import constraints, excepttypes, objtypes
|
||||
from ..constraints import IntConstraint, ObjectConstraint, returnOrThrow, throw
|
||||
from ..ssa_types import SSA_INT, SSA_OBJECT, verifierToSSAType
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
# Empirically, Hotspot does enfore size restrictions on short fields
|
||||
# Except that bool is still a byte
|
||||
_short_constraints = {
|
||||
objtypes.ByteTT: IntConstraint.range(32, -128, 127),
|
||||
objtypes.CharTT: IntConstraint.range(32, 0, 65535),
|
||||
objtypes.ShortTT: IntConstraint.range(32, -32768, 32767),
|
||||
objtypes.IntTT: IntConstraint.bot(32)
|
||||
}
|
||||
_short_constraints[objtypes.BoolTT] = _short_constraints[objtypes.ByteTT]
|
||||
# Assume no linkage errors occur, so only exception that can be thrown is NPE
|
||||
class FieldAccess(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, instr, info, args):
|
||||
super(FieldAccess, self).__init__(parent, args, makeException=('field' in instr[0]))
|
||||
|
||||
self.instruction = instr
|
||||
self.target, self.name, self.desc = info
|
||||
|
||||
dtype = None
|
||||
if 'get' in instr[0]:
|
||||
vtypes = parseFieldDescriptor(self.desc)
|
||||
stype = verifierToSSAType(vtypes[0])
|
||||
dtype = objtypes.verifierToSynthetic(vtypes[0]) # todo, find way to merge this with Invoke code?
|
||||
cat = len(vtypes)
|
||||
|
||||
self.rval = parent.makeVariable(stype, origin=self)
|
||||
self.returned = [self.rval] + [None]*(cat-1)
|
||||
else:
|
||||
self.returned = []
|
||||
|
||||
# just use a fixed constraint until we can do interprocedural analysis
|
||||
# output order is rval, exception, defined by BaseOp.getOutputs
|
||||
env = parent.env
|
||||
self.eout = ObjectConstraint.fromTops(env, [excepttypes.NullPtr], [], nonnull=True)
|
||||
if self.rval is not None:
|
||||
if self.rval.type == SSA_OBJECT:
|
||||
supers, exact = objtypes.declTypeToActual(env, dtype)
|
||||
self.rout = ObjectConstraint.fromTops(env, supers, exact)
|
||||
elif self.rval.type == SSA_INT:
|
||||
self.rout = _short_constraints[dtype]
|
||||
else:
|
||||
self.rout = constraints.fromVariable(env, self.rval)
|
||||
else:
|
||||
self.rout = None
|
||||
|
||||
def propagateConstraints(self, *incons):
|
||||
eout = None # no NPE
|
||||
if 'field' in self.instruction[0] and incons[0].null:
|
||||
eout = self.eout
|
||||
if incons[0].isConstNull():
|
||||
return throw(eout)
|
||||
|
||||
return returnOrThrow(self.rout, eout)
|
||||
|
|
@ -0,0 +1,40 @@
|
|||
from .. import ssa_types
|
||||
from ..constraints import IntConstraint, return_
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class FAdd(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
class FDiv(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
class FMul(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
class FRem(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
class FSub(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
# Unary, unlike the others
|
||||
class FNeg(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
class FCmp(BaseOp):
|
||||
def __init__(self, parent, args, NaN_val):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(ssa_types.SSA_INT, origin=self)
|
||||
self.NaN_val = NaN_val
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
return return_(IntConstraint.range(32, -1, 1))
|
||||
|
|
@ -0,0 +1,208 @@
|
|||
import itertools
|
||||
|
||||
from .. import excepttypes, ssa_types
|
||||
from ..constraints import IntConstraint, ObjectConstraint, returnOrThrow, return_, throw
|
||||
|
||||
from . import bitwise_util
|
||||
from .base import BaseOp
|
||||
|
||||
def getNewRange(w, zmin, zmax):
|
||||
HN = 1 << w-1
|
||||
zmin = zmin + HN
|
||||
zmax = zmax + HN
|
||||
split = (zmin>>w != zmax>>w)
|
||||
|
||||
if split:
|
||||
return return_(IntConstraint.range(w, -HN, HN-1))
|
||||
else:
|
||||
N = 1<<w
|
||||
return return_(IntConstraint.range(w, (zmin % N)-HN, (zmax % N)-HN))
|
||||
|
||||
class IAdd(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(IAdd, self).__init__(parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
return getNewRange(x.width, x.min+y.min, x.max+y.max)
|
||||
|
||||
class IMul(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(IMul, self).__init__(parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
vals = x.min*y.min, x.min*y.max, x.max*y.min, x.max*y.max
|
||||
return getNewRange(x.width, min(vals), max(vals))
|
||||
|
||||
class ISub(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(ISub, self).__init__(parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
return getNewRange(x.width, x.min-y.max, x.max-y.min)
|
||||
|
||||
#############################################################################################
|
||||
class IAnd(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
return return_(bitwise_util.propagateAnd(x,y))
|
||||
|
||||
class IOr(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
return return_(bitwise_util.propagateOr(x,y))
|
||||
|
||||
class IXor(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
return return_(bitwise_util.propagateXor(x,y))
|
||||
|
||||
#############################################################################################
|
||||
# Shifts currently only propogate ranges in the case where the shift is a known constant
|
||||
# TODO - make this handle the general case
|
||||
def getMaskedRange(x, bits):
|
||||
assert bits < x.width
|
||||
y = IntConstraint.const(x.width, (1<<bits) - 1)
|
||||
x = bitwise_util.propagateAnd(x,y)
|
||||
|
||||
H = 1<<(bits-1)
|
||||
M = 1<<bits
|
||||
|
||||
m1 = x.min if (x.max <= H-1) else -H
|
||||
m2 = x.max if (x.min >= -H) else H-1
|
||||
return m1, m2
|
||||
|
||||
class IShl(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
if y.min < y.max:
|
||||
return return_(IntConstraint.bot(x.width))
|
||||
shift = y.min % x.width
|
||||
if not shift:
|
||||
return return_(x)
|
||||
m1, m2 = getMaskedRange(x, x.width - shift)
|
||||
return return_(IntConstraint.range(x.width, m1<<shift, m2<<shift))
|
||||
|
||||
class IShr(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
if y.min < y.max:
|
||||
return return_(IntConstraint.range(x.width, min(x.min, 0), max(x.max, 0)))
|
||||
shift = y.min % x.width
|
||||
if not shift:
|
||||
return return_(x)
|
||||
m1, m2 = x.min, x.max
|
||||
return return_(IntConstraint.range(x.width, m1>>shift, m2>>shift))
|
||||
|
||||
class IUshr(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
M = 1<<x.width
|
||||
if y.min < y.max:
|
||||
intmax = (M//2)-1
|
||||
return return_(IntConstraint.range(x.width, min(x.min, 0), max(x.max, intmax)))
|
||||
shift = y.min % x.width
|
||||
if not shift:
|
||||
return return_(x)
|
||||
|
||||
parts = [x.min, x.max]
|
||||
if x.min <= -1 <= x.max:
|
||||
parts.append(-1)
|
||||
if x.min <= 0 <= x.max:
|
||||
parts.append(0)
|
||||
parts = [p % M for p in parts]
|
||||
m1, m2 = min(parts), max(parts)
|
||||
|
||||
return return_(IntConstraint.range(x.width, m1>>shift, m2>>shift))
|
||||
|
||||
#############################################################################################
|
||||
exec_tts = excepttypes.Arithmetic,
|
||||
class IDiv(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(IDiv, self).__init__(parent, args, makeException=True)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
self.outExceptionCons = ObjectConstraint.fromTops(parent.env, [], exec_tts, nonnull=True)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
excons = self.outExceptionCons if (y.min <= 0 <= y.max) else None
|
||||
if y.min == 0 == y.max:
|
||||
return throw(excons)
|
||||
|
||||
# Calculate possible extremes for division, taking into account special case of intmin/-1
|
||||
intmin = -1<<(x.width - 1)
|
||||
xvals = set([x.min, x.max])
|
||||
yvals = set([y.min, y.max])
|
||||
|
||||
for val in (intmin+1, 0):
|
||||
if x.min <= val <= x.max:
|
||||
xvals.add(val)
|
||||
for val in (-2,-1,1):
|
||||
if y.min <= val <= y.max:
|
||||
yvals.add(val)
|
||||
yvals.discard(0)
|
||||
|
||||
vals = set()
|
||||
for xv, yv in itertools.product(xvals, yvals):
|
||||
if xv == intmin and yv == -1:
|
||||
vals.add(intmin)
|
||||
elif xv*yv < 0: # Unlike Python, Java rounds to 0 so opposite sign case must be handled specially
|
||||
vals.add(-(-xv//yv))
|
||||
else:
|
||||
vals.add(xv//yv)
|
||||
|
||||
rvalcons = IntConstraint.range(x.width, min(vals), max(vals))
|
||||
return returnOrThrow(rvalcons, excons)
|
||||
|
||||
class IRem(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(IRem, self).__init__(parent, args, makeException=True)
|
||||
self.rval = parent.makeVariable(args[0].type, origin=self)
|
||||
self.outExceptionCons = ObjectConstraint.fromTops(parent.env, [], exec_tts, nonnull=True)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
excons = self.outExceptionCons if (y.min <= 0 <= y.max) else None
|
||||
if y.min == 0 == y.max:
|
||||
return throw(excons)
|
||||
# only do an exact result if both values are constants, and otherwise
|
||||
# just approximate the range as -(y-1) to (y-1) (or 0 to y-1 if it's positive)
|
||||
if x.min == x.max and y.min == y.max:
|
||||
val = abs(x.min) % abs(y.min)
|
||||
val = val if x.min >= 0 else -val
|
||||
return return_(IntConstraint.range(x.width, val, val))
|
||||
|
||||
mag = max(abs(y.min), abs(y.max)) - 1
|
||||
rmin = -min(mag, abs(x.min)) if x.min < 0 else 0
|
||||
rmax = min(mag, abs(x.max)) if x.max > 0 else 0
|
||||
|
||||
rvalcons = IntConstraint.range(x.width, rmin, rmax)
|
||||
return returnOrThrow(rvalcons, excons)
|
||||
|
||||
###############################################################################
|
||||
class ICmp(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
BaseOp.__init__(self, parent, args)
|
||||
self.rval = parent.makeVariable(ssa_types.SSA_INT, origin=self)
|
||||
|
||||
def propagateConstraints(self, x, y):
|
||||
rvalcons = IntConstraint.range(32, -1, 1)
|
||||
return return_(rvalcons)
|
||||
|
|
@ -0,0 +1,87 @@
|
|||
from ...verifier.descriptors import parseMethodDescriptor
|
||||
|
||||
from .. import constraints, excepttypes, objtypes
|
||||
from ..constraints import ObjectConstraint, returnOrThrow, throw
|
||||
from ..ssa_types import SSA_OBJECT, verifierToSSAType
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class Invoke(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, instr, info, args, isThisCtor, target_tt):
|
||||
super(Invoke, self).__init__(parent, args, makeException=True)
|
||||
|
||||
self.instruction = instr
|
||||
self.target, self.name, self.desc = info
|
||||
self.isThisCtor = isThisCtor # whether this is a ctor call for the current class
|
||||
self.target_tt = target_tt
|
||||
vtypes = parseMethodDescriptor(self.desc)[1]
|
||||
|
||||
dtype = None
|
||||
if vtypes:
|
||||
stype = verifierToSSAType(vtypes[0])
|
||||
dtype = objtypes.verifierToSynthetic(vtypes[0])
|
||||
cat = len(vtypes)
|
||||
# clone() on an array type is known to always return that type, rather than any Object
|
||||
if self.name == "clone" and target_tt[1] > 0:
|
||||
dtype = target_tt
|
||||
|
||||
self.rval = parent.makeVariable(stype, origin=self)
|
||||
self.returned = [self.rval] + [None]*(cat-1)
|
||||
else:
|
||||
self.rval, self.returned = None, []
|
||||
|
||||
# just use a fixed constraint until we can do interprocedural analysis
|
||||
# output order is rval, exception, defined by BaseOp.getOutputs
|
||||
env = parent.env
|
||||
self.eout = ObjectConstraint.fromTops(env, [objtypes.ThrowableTT], [], nonnull=True)
|
||||
self.eout_npe = ObjectConstraint.fromTops(env, [excepttypes.NullPtr], [], nonnull=True)
|
||||
if self.rval is not None:
|
||||
if self.rval.type == SSA_OBJECT:
|
||||
supers, exact = objtypes.declTypeToActual(env, dtype)
|
||||
self.rout = ObjectConstraint.fromTops(env, supers, exact)
|
||||
else:
|
||||
self.rout = constraints.fromVariable(env, self.rval)
|
||||
else:
|
||||
self.rout = None
|
||||
|
||||
def propagateConstraints(self, *incons):
|
||||
if self.instruction[0] != 'invokestatic' and incons[0].isConstNull():
|
||||
return throw(self.eout_npe)
|
||||
return returnOrThrow(self.rout, self.eout)
|
||||
|
||||
# TODO - cleanup
|
||||
class InvokeDynamic(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, desc, args):
|
||||
super(InvokeDynamic, self).__init__(parent, args, makeException=True)
|
||||
self.desc = desc
|
||||
vtypes = parseMethodDescriptor(self.desc)[1]
|
||||
|
||||
dtype = None
|
||||
if vtypes:
|
||||
stype = verifierToSSAType(vtypes[0])
|
||||
dtype = objtypes.verifierToSynthetic(vtypes[0])
|
||||
cat = len(vtypes)
|
||||
self.rval = parent.makeVariable(stype, origin=self)
|
||||
self.returned = [self.rval] + [None]*(cat-1)
|
||||
else:
|
||||
self.rval, self.returned = None, []
|
||||
|
||||
# just use a fixed constraint until we can do interprocedural analysis
|
||||
# output order is rval, exception, defined by BaseOp.getOutputs
|
||||
env = parent.env
|
||||
self.eout = ObjectConstraint.fromTops(env, [objtypes.ThrowableTT], [], nonnull=True)
|
||||
if self.rval is not None:
|
||||
if self.rval.type == SSA_OBJECT:
|
||||
supers, exact = objtypes.declTypeToActual(env, dtype)
|
||||
self.rout = ObjectConstraint.fromTops(env, supers, exact)
|
||||
else:
|
||||
self.rout = constraints.fromVariable(env, self.rval)
|
||||
else:
|
||||
self.rout = None
|
||||
|
||||
def propagateConstraints(self, *incons):
|
||||
return returnOrThrow(self.rout, self.eout)
|
||||
|
|
@ -0,0 +1,21 @@
|
|||
from .. import excepttypes
|
||||
from ..constraints import ObjectConstraint, maybeThrow
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class Monitor(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, args, isExit):
|
||||
BaseOp.__init__(self, parent, args, makeException=True)
|
||||
self.exit = isExit
|
||||
self.env = parent.env
|
||||
|
||||
def propagateConstraints(self, x):
|
||||
etypes = ()
|
||||
if x.null:
|
||||
etypes += (excepttypes.NullPtr,)
|
||||
if self.exit and not x.isConstNull():
|
||||
etypes += (excepttypes.MonState,)
|
||||
eout = ObjectConstraint.fromTops(self.env, [], etypes, nonnull=True)
|
||||
return maybeThrow(eout)
|
||||
|
|
@ -0,0 +1,68 @@
|
|||
from .. import excepttypes, objtypes
|
||||
from ..constraints import ObjectConstraint, returnOrThrow, throw
|
||||
from ..ssa_types import SSA_OBJECT
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class New(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, name, inode_key):
|
||||
super(New, self).__init__(parent, [], makeException=True)
|
||||
self.env = parent.env
|
||||
self.tt = objtypes.TypeTT(name, 0)
|
||||
self.rval = parent.makeVariable(SSA_OBJECT, origin=self)
|
||||
self.rval.uninit_orig_num = inode_key
|
||||
|
||||
def propagateConstraints(self):
|
||||
eout = ObjectConstraint.fromTops(self.env, [], (excepttypes.OOM,), nonnull=True)
|
||||
rout = ObjectConstraint.fromTops(self.env, [], [self.tt], nonnull=True)
|
||||
return returnOrThrow(rout, eout)
|
||||
|
||||
class NewArray(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, param, baset):
|
||||
super(NewArray, self).__init__(parent, [param], makeException=True)
|
||||
self.baset = baset
|
||||
self.rval = parent.makeVariable(SSA_OBJECT, origin=self)
|
||||
self.tt = objtypes.withDimInc(baset, 1)
|
||||
self.env = parent.env
|
||||
|
||||
def propagateConstraints(self, i):
|
||||
if i.max < 0:
|
||||
eout = ObjectConstraint.fromTops(self.env, [], (excepttypes.NegArrSize,), nonnull=True)
|
||||
return throw(eout)
|
||||
|
||||
etypes = (excepttypes.OOM,)
|
||||
if i.min < 0:
|
||||
etypes += (excepttypes.NegArrSize,)
|
||||
|
||||
eout = ObjectConstraint.fromTops(self.env, [], etypes, nonnull=True)
|
||||
rout = ObjectConstraint.fromTops(self.env, [], [self.tt], nonnull=True)
|
||||
return returnOrThrow(rout, eout)
|
||||
|
||||
class MultiNewArray(BaseOp):
|
||||
has_side_effects = True
|
||||
|
||||
def __init__(self, parent, params, type_):
|
||||
super(MultiNewArray, self).__init__(parent, params, makeException=True)
|
||||
self.tt = type_
|
||||
self.rval = parent.makeVariable(SSA_OBJECT, origin=self)
|
||||
self.env = parent.env
|
||||
|
||||
def propagateConstraints(self, *dims):
|
||||
for i in dims:
|
||||
if i.max < 0: # ignore possibility of OOM here
|
||||
eout = ObjectConstraint.fromTops(self.env, [], (excepttypes.NegArrSize,), nonnull=True)
|
||||
return throw(eout)
|
||||
|
||||
etypes = (excepttypes.OOM,)
|
||||
for i in dims:
|
||||
if i.min < 0:
|
||||
etypes += (excepttypes.NegArrSize,)
|
||||
break
|
||||
|
||||
eout = ObjectConstraint.fromTops(self.env, [], etypes, nonnull=True)
|
||||
rout = ObjectConstraint.fromTops(self.env, [], [self.tt], nonnull=True)
|
||||
return returnOrThrow(rout, eout)
|
||||
|
|
@ -0,0 +1,46 @@
|
|||
from .base import BaseOp
|
||||
|
||||
class Phi(object):
|
||||
__slots__ = 'block dict rval'.split()
|
||||
has_side_effects = False
|
||||
|
||||
def __init__(self, block, rval):
|
||||
self.block = block # used in constraint propagation
|
||||
self.dict = {}
|
||||
self.rval = rval
|
||||
assert rval is not None and rval.origin is None
|
||||
rval.origin = self
|
||||
|
||||
def add(self, key, val):
|
||||
assert key not in self.dict
|
||||
assert val.type == self.rval.type
|
||||
assert val is not None
|
||||
self.dict[key] = val
|
||||
|
||||
@property
|
||||
def params(self): return [self.dict[k] for k in self.block.predecessors]
|
||||
|
||||
def get(self, key): return self.dict[key]
|
||||
def delete(self, key): del self.dict[key]
|
||||
|
||||
# Copy these over from BaseOp so we don't need to inherit
|
||||
def replaceVars(self, rdict):
|
||||
for k in self.dict:
|
||||
self.dict[k] = rdict.get(self.dict[k], self.dict[k])
|
||||
|
||||
def getOutputs(self):
|
||||
return self.rval, None, None
|
||||
|
||||
def removeOutput(self, var):
|
||||
assert var == self.rval
|
||||
self.rval = None
|
||||
|
||||
# An extended basic block can contain multiple throwing instructions
|
||||
# but the OnException jump expects a single param. The solution is
|
||||
# to create a dummy op that behaves like a phi function, selecting
|
||||
# among the possible thrown exceptions in the block. This is always
|
||||
# the last op in block.lines when there are exceptions.
|
||||
# As this is a phi, params can be variable length
|
||||
class ExceptionPhi(BaseOp):
|
||||
def __init__(self, parent, params):
|
||||
BaseOp.__init__(self, parent, params, makeException=True)
|
||||
|
|
@ -0,0 +1,25 @@
|
|||
from .. import excepttypes, objtypes
|
||||
from ..constraints import ObjectConstraint, maybeThrow, throw
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class Throw(BaseOp):
|
||||
def __init__(self, parent, args):
|
||||
super(Throw, self).__init__(parent, args, makeException=True)
|
||||
self.env = parent.env
|
||||
|
||||
def propagateConstraints(self, x):
|
||||
if x.null:
|
||||
t = x.types
|
||||
exact = list(t.exact) + [excepttypes.NullPtr]
|
||||
return throw(ObjectConstraint.fromTops(t.env, t.supers, exact, nonnull=True))
|
||||
return throw(x)
|
||||
|
||||
# Dummy instruction that can throw anything
|
||||
class MagicThrow(BaseOp):
|
||||
def __init__(self, parent):
|
||||
super(MagicThrow, self).__init__(parent, [], makeException=True)
|
||||
self.eout = ObjectConstraint.fromTops(parent.env, [objtypes.ThrowableTT], [], nonnull=True)
|
||||
|
||||
def propagateConstraints(self):
|
||||
return maybeThrow(self.eout)
|
||||
|
|
@ -0,0 +1,37 @@
|
|||
from ..constraints import IntConstraint, return_
|
||||
|
||||
from . import bitwise_util
|
||||
from .base import BaseOp
|
||||
|
||||
class Truncate(BaseOp):
|
||||
def __init__(self, parent, arg, signed, width):
|
||||
super(Truncate, self).__init__(parent, [arg])
|
||||
|
||||
self.signed, self.width = signed, width
|
||||
self.rval = parent.makeVariable(arg.type, origin=self)
|
||||
|
||||
def propagateConstraints(self, x):
|
||||
# get range of target type
|
||||
w = self.width
|
||||
intw = x.width
|
||||
assert w < intw
|
||||
M = 1<<w
|
||||
|
||||
mask = IntConstraint.const(intw, M-1)
|
||||
x = bitwise_util.propagateAnd(x,mask)
|
||||
|
||||
# We have the mods in the range [0,M-1], but we want it in the range
|
||||
# [-M/2, M/2-1] so we need to find the new min and max
|
||||
if self.signed:
|
||||
HM = M>>1
|
||||
|
||||
parts = [(i-M if i>=HM else i) for i in (x.min, x.max)]
|
||||
if x.min <= HM-1 <= x.max:
|
||||
parts.append(HM-1)
|
||||
if x.min <= HM <= x.max:
|
||||
parts.append(-HM)
|
||||
|
||||
assert -HM <= min(parts) <= max(parts) <= HM-1
|
||||
return return_(IntConstraint.range(intw, min(parts), max(parts)))
|
||||
else:
|
||||
return return_(x)
|
||||
|
|
@ -0,0 +1,12 @@
|
|||
from .. import excepttypes
|
||||
from ..constraints import ObjectConstraint, maybeThrow
|
||||
|
||||
from .base import BaseOp
|
||||
|
||||
class TryReturn(BaseOp):
|
||||
def __init__(self, parent, canthrow=True):
|
||||
super(TryReturn, self).__init__(parent, [], makeException=True)
|
||||
self.outExceptionCons = ObjectConstraint.fromTops(parent.env, [], (excepttypes.MonState,), nonnull=True)
|
||||
|
||||
def propagateConstraints(self):
|
||||
return maybeThrow(self.outExceptionCons)
|
||||
|
|
@ -0,0 +1,77 @@
|
|||
from collections import namedtuple as nt
|
||||
|
||||
from .. import floatutil as fu
|
||||
from ..verifier import verifier_types as vtypes
|
||||
|
||||
slots_t = nt('slots_t', ('locals', 'stack'))
|
||||
|
||||
def _localsAsList(self): return [t[1] for t in sorted(self.locals.items())]
|
||||
slots_t.localsAsList = property(_localsAsList)
|
||||
|
||||
# types
|
||||
SSA_INT = 'int', 32
|
||||
SSA_LONG = 'int', 64
|
||||
SSA_FLOAT = 'float', fu.FLOAT_SIZE
|
||||
SSA_DOUBLE = 'float', fu.DOUBLE_SIZE
|
||||
SSA_OBJECT = 'obj',
|
||||
|
||||
def verifierToSSAType(vtype):
|
||||
vtype_dict = {vtypes.T_INT:SSA_INT,
|
||||
vtypes.T_LONG:SSA_LONG,
|
||||
vtypes.T_FLOAT:SSA_FLOAT,
|
||||
vtypes.T_DOUBLE:SSA_DOUBLE}
|
||||
# These should never be passed in here
|
||||
assert vtype.tag not in ('.new','.init')
|
||||
vtype = vtypes.withNoConst(vtype)
|
||||
if vtypes.objOrArray(vtype):
|
||||
return SSA_OBJECT
|
||||
elif vtype in vtype_dict:
|
||||
return vtype_dict[vtype]
|
||||
return None
|
||||
|
||||
# Note: This is actually an Extended Basic Block. A normal basic block has to end whenever there is
|
||||
# an instruction that can throw. This means that there is a seperate basic block for every throwing
|
||||
# method, which causes horrible performance, especially in a large method with otherwise linear code.
|
||||
# The solution is to use extended basic blocks, which are like normal basic blocks except that they
|
||||
# can contain multiple throwing instructions as long as every throwing instruction has the same
|
||||
# handlers. Due to the use of SSA, we also require that there are no changes to the locals between the
|
||||
# first and last throwing instruction.
|
||||
class BasicBlock(object):
|
||||
__slots__ = "key phis lines jump unaryConstraints predecessors inslots throwvars chpairs except_used locals_at_except".split()
|
||||
|
||||
def __init__(self, key):
|
||||
self.key = key
|
||||
self.phis = None # The list of phi statements merging incoming variables
|
||||
self.lines = [] # List of operations in the block
|
||||
self.jump = None # The exit point (if, goto, etc)
|
||||
|
||||
# Holds constraints (range and type information) for each variable in the block.
|
||||
# If the value is None, this variable cannot be reached
|
||||
self.unaryConstraints = None
|
||||
# List of predecessor pairs in deterministic order
|
||||
self.predecessors = []
|
||||
|
||||
# temp vars used during graph creation
|
||||
self.inslots = None
|
||||
self.throwvars = []
|
||||
self.chpairs = None
|
||||
self.except_used = None
|
||||
self.locals_at_except = None
|
||||
|
||||
def filterVarConstraints(self, keepvars):
|
||||
self.unaryConstraints = {k:v for k,v in self.unaryConstraints.items() if k in keepvars}
|
||||
|
||||
def removePredPair(self, pair):
|
||||
self.predecessors.remove(pair)
|
||||
for phi in self.phis:
|
||||
del phi.dict[pair]
|
||||
|
||||
def replacePredPair(self, oldp, newp):
|
||||
self.predecessors[self.predecessors.index(oldp)] = newp
|
||||
for phi in self.phis:
|
||||
phi.dict[newp] = phi.dict[oldp]
|
||||
del phi.dict[oldp]
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
return 'Block ' + str(self.key)
|
||||
__repr__ = __str__
|
||||
|
|
@ -0,0 +1,59 @@
|
|||
import copy
|
||||
|
||||
from .ssa_types import slots_t
|
||||
|
||||
class ProcInfo(object):
|
||||
def __init__(self, retblock, target):
|
||||
self.retblock = retblock
|
||||
self.target = target
|
||||
self.jsrblocks = []
|
||||
assert target is retblock.jump.target
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
return 'Proc{}<{}>'.format(self.target.key, ', '.join(str(b.key) for b in self.jsrblocks))
|
||||
__repr__ = __str__
|
||||
|
||||
###########################################################################################
|
||||
class ProcJumpBase(object):
|
||||
@property
|
||||
def params(self):
|
||||
return [v for v in self.input.stack + self.input.localsAsList if v is not None]
|
||||
# [v for v in self.input.stack if v] + [v for k, v in sorted(self.input.locals.items()) if v]
|
||||
|
||||
def replaceBlocks(self, blockDict):
|
||||
self.target = blockDict.get(self.target, self.target)
|
||||
|
||||
def getExceptSuccessors(self): return ()
|
||||
def getSuccessors(self): return self.getNormalSuccessors()
|
||||
def getSuccessorPairs(self): return [(x,False) for x in self.getNormalSuccessors()]
|
||||
def reduceSuccessors(self, pairsToRemove): return self
|
||||
|
||||
class ProcCallOp(ProcJumpBase):
|
||||
def __init__(self, target, fallthrough, inslots, outslots):
|
||||
self.fallthrough = fallthrough
|
||||
self.target = target
|
||||
self.input = inslots
|
||||
self.output = outslots
|
||||
|
||||
for var in self.output.stack + self.output.locals.values():
|
||||
if var is not None:
|
||||
assert var.origin is None
|
||||
var.origin = self
|
||||
|
||||
# def flatOutput(self): return [v for v in self.output.stack if v] + [v for k, v in sorted(self.output.locals.items()) if v]
|
||||
def flatOutput(self): return self.output.stack + self.output.localsAsList
|
||||
|
||||
def getNormalSuccessors(self): return self.fallthrough, self.target
|
||||
|
||||
class DummyRet(ProcJumpBase):
|
||||
def __init__(self, inslots, target):
|
||||
self.target = target
|
||||
self.input = inslots
|
||||
|
||||
def replaceVars(self, varDict):
|
||||
newstack = [varDict.get(v, v) for v in self.input.stack]
|
||||
newlocals = {k: varDict.get(v, v) for k, v in self.input.locals.items()}
|
||||
self.input = slots_t(stack=newstack, locals=newlocals)
|
||||
|
||||
def getNormalSuccessors(self): return ()
|
||||
def clone(self): return copy.copy(self) # target and input will be replaced later by calls to replaceBlocks/Vars
|
||||
|
|
@ -0,0 +1,7 @@
|
|||
def thunk(initial):
|
||||
stack = [initial]
|
||||
while stack:
|
||||
try:
|
||||
stack.append(next(stack[-1]))
|
||||
except StopIteration:
|
||||
stack.pop()
|
||||
|
|
@ -0,0 +1,85 @@
|
|||
from .verifier_types import T_ARRAY, T_BOOL, T_BYTE, T_CHAR, T_DOUBLE, T_FLOAT, T_INT, T_INVALID, T_LONG, T_OBJECT, T_SHORT, unSynthesizeType
|
||||
|
||||
_cat2tops = T_LONG, T_DOUBLE
|
||||
|
||||
def parseFieldDescriptors(desc_str, unsynthesize=True):
|
||||
baseTypes = {'B':T_BYTE, 'C':T_CHAR, 'D':T_DOUBLE, 'F':T_FLOAT,
|
||||
'I':T_INT, 'J':T_LONG, 'S':T_SHORT, 'Z':T_BOOL}
|
||||
|
||||
fields = []
|
||||
while desc_str:
|
||||
oldlen = len(desc_str)
|
||||
desc_str = desc_str.lstrip('[')
|
||||
dim = oldlen - len(desc_str)
|
||||
if dim > 255:
|
||||
raise ValueError('Dimension {} > 255 in descriptor'.format(dim))
|
||||
if not desc_str:
|
||||
raise ValueError('Descriptor contains [s at end of string')
|
||||
|
||||
if desc_str[0] == 'L':
|
||||
end = desc_str.find(';')
|
||||
if end == -1:
|
||||
raise ValueError('Unmatched L in descriptor')
|
||||
|
||||
name = desc_str[1:end]
|
||||
desc_str = desc_str[end+1:]
|
||||
baset = T_OBJECT(name)
|
||||
else:
|
||||
if desc_str[0] not in baseTypes:
|
||||
raise ValueError('Unrecognized code {} in descriptor'.format(desc_str[0]))
|
||||
baset = baseTypes[desc_str[0]]
|
||||
desc_str = desc_str[1:]
|
||||
|
||||
if dim:
|
||||
# Hotspot considers byte[] and bool[] identical for type checking purposes
|
||||
if unsynthesize and baset == T_BOOL:
|
||||
baset = T_BYTE
|
||||
baset = T_ARRAY(baset, dim)
|
||||
elif unsynthesize:
|
||||
# synthetics are only meaningful as basetype of an array
|
||||
# if they are by themselves, convert to int.
|
||||
baset = unSynthesizeType(baset)
|
||||
|
||||
fields.append(baset)
|
||||
if baset in _cat2tops:
|
||||
fields.append(T_INVALID)
|
||||
return fields
|
||||
|
||||
# get a single descriptor
|
||||
def parseFieldDescriptor(desc_str, unsynthesize=True):
|
||||
rval = parseFieldDescriptors(desc_str, unsynthesize)
|
||||
|
||||
cat = 2 if (rval and rval[0] in _cat2tops) else 1
|
||||
if len(rval) != cat:
|
||||
raise ValueError('Incorrect number of fields in descriptor, expected {} but found {}'.format(cat, len(rval)))
|
||||
return rval
|
||||
|
||||
# Parse a string to get a Java Method Descriptor
|
||||
def parseMethodDescriptor(desc_str, unsynthesize=True):
|
||||
if not desc_str.startswith('('):
|
||||
raise ValueError('Method descriptor does not start with (')
|
||||
|
||||
# we need to split apart the argument list and return value
|
||||
# this is greatly complicated by the fact that ) is a legal
|
||||
# character that can appear in class names
|
||||
|
||||
lp_pos = desc_str.rfind(')') # this case will work if return type is not an object
|
||||
if desc_str.endswith(';'):
|
||||
lbound = max(desc_str.rfind(';', 1, -1), 1)
|
||||
lp_pos = desc_str.find(')', lbound, -1)
|
||||
if lp_pos < 0 or desc_str[lp_pos] != ')':
|
||||
raise ValueError('Unable to split method descriptor into arguments and return type')
|
||||
|
||||
arg_str = desc_str[1:lp_pos]
|
||||
rval_str = desc_str[lp_pos+1:]
|
||||
|
||||
args = parseFieldDescriptors(arg_str, unsynthesize)
|
||||
rval = [] if rval_str == 'V' else parseFieldDescriptor(rval_str, unsynthesize)
|
||||
return args, rval
|
||||
|
||||
# Adds self argument for nonstatic. Constructors must be handled seperately
|
||||
def parseUnboundMethodDescriptor(desc_str, target, isstatic):
|
||||
args, rval = parseMethodDescriptor(desc_str)
|
||||
if not isstatic:
|
||||
args = [T_OBJECT(target)] + args
|
||||
return args, rval
|
||||
|
|
@ -0,0 +1,508 @@
|
|||
import itertools
|
||||
|
||||
from .. import bytecode, error as error_types, opnames as ops
|
||||
|
||||
from .descriptors import parseFieldDescriptor, parseMethodDescriptor, parseUnboundMethodDescriptor
|
||||
from .verifier_types import OBJECT_INFO, T_ADDRESS, T_ARRAY, T_DOUBLE, T_FLOAT, T_INT, T_INT_CONST, T_INVALID, T_LONG, T_NULL, T_OBJECT, T_UNINIT_OBJECT, T_UNINIT_THIS, decrementDim, exactArrayFrom, fullinfo_t, mergeTypes
|
||||
|
||||
class VerifierTypesState(object):
|
||||
def __init__(self, stack, locals, masks):
|
||||
self.stack = stack
|
||||
self.locals = locals
|
||||
self.masks = masks
|
||||
|
||||
def copy(self): return VerifierTypesState(self.stack, self.locals, self.masks)
|
||||
|
||||
def withExcept(self, t): return VerifierTypesState([t], self.locals, self.masks)
|
||||
|
||||
def pop(self, n):
|
||||
if n == 0:
|
||||
return []
|
||||
self.stack, popped = self.stack[:-n], self.stack[-n:]
|
||||
return popped
|
||||
|
||||
def push(self, vals):
|
||||
self.stack = self.stack + list(vals)
|
||||
|
||||
def setLocal(self, i, v):
|
||||
if len(self.locals) < i:
|
||||
self.locals = self.locals + [T_INVALID]*(i - len(self.locals))
|
||||
self.locals = self.locals[:i] + [v] + self.locals[i+1:]
|
||||
|
||||
new = frozenset([i])
|
||||
self.masks = [(addr, old | new) for addr, old in self.masks]
|
||||
|
||||
def local(self, i):
|
||||
if len(self.locals) <= i:
|
||||
return T_INVALID
|
||||
return self.locals[i]
|
||||
|
||||
def jsr(self, target):
|
||||
self.masks = self.masks + [(target, frozenset())]
|
||||
|
||||
def replace(self, old, new):
|
||||
self.stack = [(new if v == old else v) for v in self.stack]
|
||||
|
||||
mask = frozenset(i for i, v in enumerate(self.locals) if v == old)
|
||||
self.locals = [(new if v == old else v) for v in self.locals]
|
||||
self.masks = [(addr, oldmask | mask) for addr, oldmask in self.masks]
|
||||
|
||||
def invalidateNews(self):
|
||||
# Doesn't need to update mask
|
||||
self.stack = [(T_INVALID if v.tag == '.new' else v) for v in self.stack]
|
||||
self.locals = [(T_INVALID if v.tag == '.new' else v) for v in self.locals]
|
||||
|
||||
def maskFor(self, called):
|
||||
self.masks = self.masks[:]
|
||||
target, mask = self.masks.pop()
|
||||
while target != called:
|
||||
target, mask = self.masks.pop()
|
||||
return mask
|
||||
|
||||
def returnTo(self, called, jsrstate):
|
||||
mask = self.maskFor(called)
|
||||
# merge locals using mask
|
||||
zipped = itertools.izip_longest(self.locals, jsrstate.locals, fillvalue=T_INVALID)
|
||||
self.locals = [(x if i in mask else y) for i,(x,y) in enumerate(zipped)]
|
||||
|
||||
def merge(self, other, env):
|
||||
old_triple = self.stack, self.locals, self.masks
|
||||
assert len(self.stack) == len(other.stack)
|
||||
self.stack = [mergeTypes(env, new, old) for old, new in zip(self.stack, other.stack)]
|
||||
self.locals = [mergeTypes(env, new, old) for old, new in zip(self.locals, other.locals)]
|
||||
while self.locals and self.locals[-1] == T_INVALID:
|
||||
self.locals.pop()
|
||||
|
||||
# Merge Masks
|
||||
last_match = -1
|
||||
mergedmasks = []
|
||||
for entry1, mask1 in self.masks:
|
||||
for j, (entry2, mask2) in enumerate(other.masks):
|
||||
if j > last_match and entry1 == entry2:
|
||||
item = entry1, (mask1 | mask2)
|
||||
mergedmasks.append(item)
|
||||
last_match = j
|
||||
self.masks = mergedmasks
|
||||
return (self.stack, self.locals, self.masks) != old_triple
|
||||
|
||||
def stateFromInitialArgs(args): return VerifierTypesState([], args[:], [])
|
||||
|
||||
|
||||
|
||||
_invoke_ops = (ops.INVOKESPECIAL, ops.INVOKESTATIC, ops.INVOKEVIRTUAL, ops.INVOKEINTERFACE, ops.INVOKEINIT, ops.INVOKEDYNAMIC)
|
||||
|
||||
def _loadFieldDesc(cpool, ind):
|
||||
target, name, desc = cpool.getArgsCheck('Field', ind)
|
||||
return parseFieldDescriptor(desc)
|
||||
|
||||
def _loadMethodDesc(cpool, ind):
|
||||
target, name, desc = cpool.getArgs(ind)
|
||||
return parseMethodDescriptor(desc)
|
||||
|
||||
def _indexToCFMInfo(cpool, ind, typen):
|
||||
actual = cpool.getType(ind)
|
||||
# JVM_GetCPMethodClassNameUTF accepts both
|
||||
assert actual == typen or actual == 'InterfaceMethod' and typen == 'Method'
|
||||
|
||||
cname = cpool.getArgs(ind)[0]
|
||||
if cname.startswith('[') or cname.endswith(';'):
|
||||
try:
|
||||
return parseFieldDescriptor(cname)[0]
|
||||
except ValueError as e:
|
||||
return T_INVALID
|
||||
else:
|
||||
return T_OBJECT(cname)
|
||||
|
||||
# Instructions which pop a fixed amount
|
||||
_popAmount = {
|
||||
ops.ARRLOAD_OBJ: 2,
|
||||
ops.ARRSTORE_OBJ: 3,
|
||||
ops.ARRLOAD: 2,
|
||||
ops.TRUNCATE: 1,
|
||||
ops.LCMP: 4,
|
||||
ops.IF_A: 1,
|
||||
ops.IF_I: 1,
|
||||
ops.IF_ACMP: 2,
|
||||
ops.IF_ICMP: 2,
|
||||
ops.SWITCH: 1,
|
||||
ops.NEWARRAY: 1,
|
||||
ops.ANEWARRAY: 1,
|
||||
ops.ARRLEN: 1,
|
||||
ops.THROW: 1,
|
||||
ops.CHECKCAST: 1,
|
||||
ops.INSTANCEOF: 1,
|
||||
ops.MONENTER: 1,
|
||||
ops.MONEXIT: 1,
|
||||
ops.GETFIELD: 1,
|
||||
|
||||
ops.NOP: 0,
|
||||
ops.CONSTNULL: 0,
|
||||
ops.CONST: 0,
|
||||
ops.LDC: 0,
|
||||
ops.LOAD: 0,
|
||||
ops.IINC: 0,
|
||||
ops.GOTO: 0,
|
||||
ops.JSR: 0,
|
||||
ops.RET: 0,
|
||||
ops.NEW: 0,
|
||||
ops.GETSTATIC: 0,
|
||||
}
|
||||
# Instructions which pop a variable amount depending on whether type is category 2
|
||||
_popAmountVar = {
|
||||
ops.STORE: (1, 0),
|
||||
ops.NEG: (1, 0),
|
||||
ops.CONVERT: (1, 0),
|
||||
|
||||
ops.ADD: (2, 0),
|
||||
ops.SUB: (2, 0),
|
||||
ops.MUL: (2, 0),
|
||||
ops.DIV: (2, 0),
|
||||
ops.REM: (2, 0),
|
||||
ops.XOR: (2, 0),
|
||||
ops.OR: (2, 0),
|
||||
ops.AND: (2, 0),
|
||||
ops.FCMP: (2, 0),
|
||||
|
||||
ops.SHL: (1, 1),
|
||||
ops.SHR: (1, 1),
|
||||
ops.USHR: (1, 1),
|
||||
|
||||
ops.ARRSTORE: (1, 2),
|
||||
}
|
||||
# Generic stack codes
|
||||
genericStackCodes = {
|
||||
ops.POP: (1, []),
|
||||
ops.POP2: (2, []),
|
||||
ops.DUP: (1, [0, 0]),
|
||||
ops.DUPX1: (2, [1, 0, 1]),
|
||||
ops.DUPX2: (3, [2, 0, 1, 2]),
|
||||
ops.DUP2: (2, [0, 1, 0, 1]),
|
||||
ops.DUP2X1: (3, [1, 2, 0, 1, 2]),
|
||||
ops.DUP2X2: (4, [2, 3, 0, 1, 2, 3]),
|
||||
ops.SWAP: (2, [1, 0]),
|
||||
}
|
||||
|
||||
def _getPopAmount(cpool, instr, method):
|
||||
op = instr[0]
|
||||
if op in _popAmount:
|
||||
return _popAmount[op]
|
||||
if op in _popAmountVar:
|
||||
a, b = _popAmountVar[op]
|
||||
cat = 2 if instr[1] in 'JD' else 1
|
||||
return a * cat + b
|
||||
if op in genericStackCodes:
|
||||
return genericStackCodes[op][0]
|
||||
|
||||
if op == ops.MULTINEWARRAY:
|
||||
return instr[2]
|
||||
elif op == ops.RETURN:
|
||||
return len(parseMethodDescriptor(method.descriptor)[1])
|
||||
elif op in (ops.PUTFIELD, ops.PUTSTATIC):
|
||||
args = len(_loadFieldDesc(cpool, instr[1]))
|
||||
if op == ops.PUTFIELD:
|
||||
args += 1
|
||||
return args
|
||||
elif op in _invoke_ops:
|
||||
args = len(_loadMethodDesc(cpool, instr[1])[0])
|
||||
if op != ops.INVOKESTATIC and op != ops.INVOKEDYNAMIC:
|
||||
args += 1
|
||||
return args
|
||||
|
||||
codes = dict(zip('IFJD', [T_INT, T_FLOAT, T_LONG, T_DOUBLE]))
|
||||
def _getStackResult(cpool, instr, key):
|
||||
op = instr[0]
|
||||
|
||||
if op in (ops.TRUNCATE, ops.LCMP, ops.FCMP, ops.ARRLEN, ops.INSTANCEOF):
|
||||
return T_INT
|
||||
elif op in (ops.ADD, ops.SUB, ops.MUL, ops.DIV, ops.REM, ops.XOR, ops.AND, ops.OR, ops.SHL, ops.SHR, ops.USHR, ops.NEG):
|
||||
return codes[instr[1]]
|
||||
elif op == ops.CONSTNULL:
|
||||
return T_NULL
|
||||
elif op == ops.CONST:
|
||||
if instr[1] == 'I':
|
||||
return T_INT_CONST(instr[2])
|
||||
return codes[instr[1]]
|
||||
elif op == ops.ARRLOAD:
|
||||
return codes.get(instr[1], T_INT)
|
||||
elif op == ops.CONVERT:
|
||||
return codes[instr[2]]
|
||||
|
||||
elif op == ops.LDC:
|
||||
return {
|
||||
'Int': T_INT,
|
||||
'Long': T_LONG,
|
||||
'Float': T_FLOAT,
|
||||
'Double': T_DOUBLE,
|
||||
'String': T_OBJECT('java/lang/String'),
|
||||
'Class': T_OBJECT('java/lang/Class'),
|
||||
'MethodType': T_OBJECT('java/lang/invoke/MethodType'),
|
||||
'MethodHandle': T_OBJECT('java/lang/invoke/MethodHandle'),
|
||||
}[cpool.getType(instr[1])]
|
||||
|
||||
elif op == ops.JSR:
|
||||
return T_ADDRESS(instr[1])
|
||||
|
||||
elif op in (ops.CHECKCAST, ops.NEW, ops.ANEWARRAY, ops.MULTINEWARRAY):
|
||||
target = _indexToCFMInfo(cpool, instr[1], 'Class')
|
||||
if op == ops.ANEWARRAY:
|
||||
return T_ARRAY(target)
|
||||
elif op == ops.NEW:
|
||||
return T_UNINIT_OBJECT(key)
|
||||
return target
|
||||
elif op == ops.NEWARRAY:
|
||||
return parseFieldDescriptor('[' + instr[1])[0]
|
||||
|
||||
elif op in (ops.GETFIELD, ops.GETSTATIC):
|
||||
return _loadFieldDesc(cpool, instr[1])[0]
|
||||
elif op in _invoke_ops:
|
||||
out = _loadMethodDesc(cpool, instr[1])[1]
|
||||
assert 0 <= len(out) <= 2
|
||||
return out[0] if out else None
|
||||
|
||||
class InstructionNode(object):
|
||||
__slots__ = "key code env class_ cpool instruction op visited changed offsetToIndex indexToOffset state out_state jsrTarget next_instruction returnedFrom successors pop_amount stack_push stack_code target_type isThisCtor".split()
|
||||
|
||||
def __init__(self, code, offsetToIndex, indexToOffset, key):
|
||||
self.key = key
|
||||
assert(self.key is not None) # if it is this will cause problems with origin tracking
|
||||
|
||||
self.code = code
|
||||
self.env = code.class_.env
|
||||
self.class_ = code.class_
|
||||
self.cpool = self.class_.cpool
|
||||
|
||||
self.instruction = code.bytecode[key]
|
||||
self.op = self.instruction[0]
|
||||
|
||||
self.visited, self.changed = False, False
|
||||
# store for usage calculating JSRs, finding successor instructions and the like
|
||||
self.offsetToIndex = offsetToIndex
|
||||
self.indexToOffset = indexToOffset
|
||||
|
||||
self.state = None
|
||||
|
||||
# Fields to be assigned later
|
||||
self.jsrTarget = None
|
||||
self.next_instruction = None
|
||||
self.returnedFrom = None
|
||||
self.successors = None
|
||||
|
||||
self.pop_amount = -1
|
||||
self.stack_push = []
|
||||
self.stack_code = None
|
||||
|
||||
# for blockmaker
|
||||
self.target_type = None
|
||||
self.isThisCtor = False
|
||||
self.out_state = None # store out state for JSR/RET instructions
|
||||
|
||||
self._precomputeValues()
|
||||
|
||||
def _removeInterface(self, vt):
|
||||
if vt.tag == '.obj' and vt.extra is not None and self.env.isInterface(vt.extra, forceCheck=True):
|
||||
return T_ARRAY(OBJECT_INFO, vt.dim)
|
||||
return vt
|
||||
|
||||
def _precomputeValues(self):
|
||||
# parsed_desc, successors
|
||||
off_i = self.offsetToIndex[self.key]
|
||||
self.next_instruction = self.indexToOffset[off_i+1] # None if end of code
|
||||
op = self.instruction[0]
|
||||
|
||||
self.pop_amount = _getPopAmount(self.cpool, self.instruction, self.code.method)
|
||||
|
||||
# cache these, since they're not state dependent
|
||||
result = _getStackResult(self.cpool, self.instruction, self.key)
|
||||
# temporary hack
|
||||
if op == ops.CHECKCAST:
|
||||
result = self._removeInterface(result)
|
||||
|
||||
if result is not None:
|
||||
self.stack_push = [result]
|
||||
if result in (T_LONG, T_DOUBLE):
|
||||
self.stack_push.append(T_INVALID)
|
||||
|
||||
if op in genericStackCodes:
|
||||
self.stack_code = genericStackCodes[op][1]
|
||||
|
||||
if op == ops.NEW:
|
||||
self.target_type = _indexToCFMInfo(self.cpool, self.instruction[1], 'Class')
|
||||
|
||||
# Now get successors
|
||||
next_ = self.next_instruction
|
||||
if op in (ops.IF_A, ops.IF_I, ops.IF_ICMP, ops.IF_ACMP):
|
||||
self.successors = next_, self.instruction[2]
|
||||
elif op in (ops.JSR, ops.GOTO):
|
||||
self.successors = self.instruction[1],
|
||||
elif op in (ops.RETURN, ops.THROW):
|
||||
self.successors = ()
|
||||
elif op == ops.RET:
|
||||
self.successors = None # calculate it when the node is reached
|
||||
elif op == ops.SWITCH:
|
||||
opname, default, jumps = self.instruction
|
||||
targets = (default,)
|
||||
if jumps:
|
||||
targets += zip(*jumps)[1]
|
||||
self.successors = targets
|
||||
else:
|
||||
self.successors = next_,
|
||||
|
||||
def _getNewState(self, iNodes):
|
||||
state = self.state.copy()
|
||||
popped = state.pop(self.pop_amount)
|
||||
|
||||
# Local updates/reading
|
||||
op = self.instruction[0]
|
||||
if op == ops.LOAD:
|
||||
state.push([state.local(self.instruction[2])])
|
||||
if self.instruction[1] in 'JD':
|
||||
state.push([T_INVALID])
|
||||
elif op == ops.STORE:
|
||||
for i, val in enumerate(popped):
|
||||
state.setLocal(self.instruction[2] + i, val)
|
||||
elif op == ops.IINC:
|
||||
state.setLocal(self.instruction[1], T_INT) # Make sure to clobber constants
|
||||
elif op == ops.JSR:
|
||||
state.jsr(self.instruction[1])
|
||||
elif op == ops.NEW:
|
||||
# This should never happen, but better safe than sorry.
|
||||
state.replace(self.stack_push[0], T_INVALID)
|
||||
elif op == ops.INVOKEINIT:
|
||||
old = popped[0]
|
||||
if old.tag == '.new':
|
||||
new = _indexToCFMInfo(self.cpool, self.instruction[1], 'Method')
|
||||
else: # .init
|
||||
new = T_OBJECT(self.class_.name)
|
||||
self.isThisCtor = True
|
||||
state.replace(old, new)
|
||||
|
||||
# Make sure that push happens after local replacement in case of new/invokeinit
|
||||
if self.stack_code is not None:
|
||||
state.push(popped[i] for i in self.stack_code)
|
||||
elif op == ops.ARRLOAD_OBJ:
|
||||
# temporary hack
|
||||
result = self._removeInterface(decrementDim(popped[0]))
|
||||
state.push([result])
|
||||
|
||||
elif op == ops.NEWARRAY or op == ops.ANEWARRAY:
|
||||
arrt = self.stack_push[0]
|
||||
size = popped[0].const
|
||||
if size is not None:
|
||||
arrt = exactArrayFrom(arrt, size)
|
||||
state.push([arrt])
|
||||
else:
|
||||
state.push(self.stack_push)
|
||||
|
||||
if self.op in (ops.RET, ops.JSR):
|
||||
state.invalidateNews()
|
||||
self.out_state = state # store for later convienence
|
||||
|
||||
assert all(isinstance(vt, fullinfo_t) for vt in state.stack)
|
||||
assert all(isinstance(vt, fullinfo_t) for vt in state.locals)
|
||||
return state
|
||||
|
||||
def _mergeSingleSuccessor(self, other, newstate, iNodes, isException):
|
||||
if self.op == ops.RET and not isException:
|
||||
# Get the instruction before other
|
||||
off_i = self.offsetToIndex[other.key]
|
||||
jsrnode = iNodes[self.indexToOffset[off_i-1]]
|
||||
jsrnode.returnedFrom = self.key
|
||||
|
||||
if jsrnode.visited: # if not, skip for later
|
||||
newstate = newstate.copy()
|
||||
newstate.returnTo(jsrnode.instruction[1], jsrnode.state)
|
||||
else:
|
||||
return
|
||||
|
||||
if not other.visited:
|
||||
other.state = newstate.copy()
|
||||
other.visited = other.changed = True
|
||||
else:
|
||||
changed = other.state.merge(newstate, self.env)
|
||||
other.changed = other.changed or changed
|
||||
|
||||
def update(self, iNodes, exceptions):
|
||||
assert self.visited
|
||||
self.changed = False
|
||||
newstate = self._getNewState(iNodes)
|
||||
|
||||
successors = self.successors
|
||||
if self.op == ops.JSR and self.returnedFrom is not None:
|
||||
iNodes[self.returnedFrom].changed = True
|
||||
|
||||
if successors is None:
|
||||
assert self.op == ops.RET
|
||||
called = self.state.local(self.instruction[1]).extra
|
||||
temp = [n.next_instruction for n in iNodes.values() if (n.op == ops.JSR and n.instruction[1] == called)]
|
||||
successors = self.successors = tuple(temp)
|
||||
self.jsrTarget = called # store for later use in ssa creation
|
||||
|
||||
# Merge into exception handlers first
|
||||
for (start, end, handler, except_info) in exceptions:
|
||||
if start <= self.key < end:
|
||||
self._mergeSingleSuccessor(handler, self.state.withExcept(except_info), iNodes, True)
|
||||
if self.op == ops.INVOKEINIT: # two cases since the ctor may suceed or fail before throwing
|
||||
self._mergeSingleSuccessor(handler, newstate.withExcept(except_info), iNodes, True)
|
||||
|
||||
# Now regular successors
|
||||
for k in self.successors:
|
||||
self._mergeSingleSuccessor(iNodes[k], newstate, iNodes, False)
|
||||
|
||||
def __str__(self): # pragma: no cover
|
||||
lines = ['{}: {}'.format(self.key, bytecode.printInstruction(self.instruction))]
|
||||
if self.visited:
|
||||
lines.append('Stack: ' + ', '.join(map(str, self.state.stack)))
|
||||
lines.append('Locals: ' + ', '.join(map(str, self.state.locals)))
|
||||
if self.state.masks:
|
||||
lines.append('Masks:')
|
||||
lines += ['\t{}: {}'.format(entry, sorted(cset)) for entry, cset in self.state.masks]
|
||||
else:
|
||||
lines.append('\tunvisited')
|
||||
return '\n'.join(lines) + '\n'
|
||||
|
||||
def verifyBytecode(code):
|
||||
method, class_ = code.method, code.class_
|
||||
args, rval = parseUnboundMethodDescriptor(method.descriptor, class_.name, method.static)
|
||||
env = class_.env
|
||||
|
||||
# Object has no superclass to construct, so it doesn't get an uninit this
|
||||
if method.isConstructor and class_.name != 'java/lang/Object':
|
||||
assert args[0] == T_OBJECT(class_.name)
|
||||
args[0] = T_UNINIT_THIS
|
||||
assert len(args) <= 255 and len(args) <= code.locals
|
||||
|
||||
offsets = sorted(code.bytecode.keys())
|
||||
offset_rmap = {v:i for i,v in enumerate(offsets)}
|
||||
offsets.append(None) # Sentinel for end of code
|
||||
iNodes = [InstructionNode(code, offset_rmap, offsets, key) for key in offsets[:-1]]
|
||||
iNodeLookup = {n.key:n for n in iNodes}
|
||||
|
||||
keys = frozenset(iNodeLookup)
|
||||
for raw in code.except_raw:
|
||||
if not ((0 <= raw.start < raw.end) and (raw.start in keys) and
|
||||
(raw.handler in keys) and (raw.end in keys or raw.end == code.codelen)):
|
||||
|
||||
keylist = sorted(keys) + [code.codelen]
|
||||
msg = "Illegal exception handler: {}\nValid offsets are: {}".format(raw, ', '.join(map(str, keylist)))
|
||||
raise error_types.VerificationError(msg)
|
||||
|
||||
def makeException(rawdata):
|
||||
if rawdata.type_ind:
|
||||
typen = class_.cpool.getArgsCheck('Class', rawdata.type_ind)
|
||||
else:
|
||||
typen = 'java/lang/Throwable'
|
||||
return (rawdata.start, rawdata.end, iNodeLookup[rawdata.handler], T_OBJECT(typen))
|
||||
exceptions = map(makeException, code.except_raw)
|
||||
|
||||
start = iNodes[0]
|
||||
start.state = stateFromInitialArgs(args)
|
||||
start.visited, start.changed = True, True
|
||||
|
||||
done = False
|
||||
while not done:
|
||||
done = True
|
||||
for node in iNodes:
|
||||
if node.changed:
|
||||
node.update(iNodeLookup, exceptions)
|
||||
done = False
|
||||
return iNodes
|
||||
|
|
@ -0,0 +1,141 @@
|
|||
from collections import namedtuple as nt
|
||||
|
||||
# Define types for Inference
|
||||
# Extra stores address for .new and .address and class name for object types
|
||||
# Const stores value for int constants and length for exact arrays. This isn't needed for normal verification but is
|
||||
# useful for optimizing the code later.
|
||||
fullinfo_t = nt('fullinfo_t', ['tag','dim','extra','const'])
|
||||
|
||||
valid_tags = ['.'+_x for _x in 'int float double long obj new init address byte short char boolean'.split()]
|
||||
valid_tags = frozenset([None] + valid_tags)
|
||||
|
||||
def _makeinfo(tag, dim=0, extra=None, const=None):
|
||||
assert tag in valid_tags
|
||||
return fullinfo_t(tag, dim, extra, const)
|
||||
|
||||
T_INVALID = _makeinfo(None)
|
||||
T_INT = _makeinfo('.int')
|
||||
T_FLOAT = _makeinfo('.float')
|
||||
T_DOUBLE = _makeinfo('.double')
|
||||
T_LONG = _makeinfo('.long')
|
||||
|
||||
T_NULL = _makeinfo('.obj')
|
||||
T_UNINIT_THIS = _makeinfo('.init')
|
||||
|
||||
T_BYTE = _makeinfo('.byte')
|
||||
T_SHORT = _makeinfo('.short')
|
||||
T_CHAR = _makeinfo('.char')
|
||||
T_BOOL = _makeinfo('.boolean') # Hotspot doesn't have a bool type, but we can use this elsewhere
|
||||
|
||||
# types with arguments
|
||||
def T_ADDRESS(entry):
|
||||
return _makeinfo('.address', extra=entry)
|
||||
|
||||
def T_OBJECT(name):
|
||||
return _makeinfo('.obj', extra=name)
|
||||
|
||||
def T_ARRAY(baset, newDimensions=1):
|
||||
assert 0 <= baset.dim <= 255-newDimensions
|
||||
return _makeinfo(baset.tag, baset.dim+newDimensions, baset.extra)
|
||||
|
||||
def T_UNINIT_OBJECT(origin):
|
||||
return _makeinfo('.new', extra=origin)
|
||||
|
||||
def T_INT_CONST(val):
|
||||
assert -0x80000000 <= val < 0x80000000
|
||||
return _makeinfo(T_INT.tag, const=val)
|
||||
|
||||
OBJECT_INFO = T_OBJECT('java/lang/Object')
|
||||
CLONE_INFO = T_OBJECT('java/lang/Cloneable')
|
||||
SERIAL_INFO = T_OBJECT('java/io/Serializable')
|
||||
THROWABLE_INFO = T_OBJECT('java/lang/Throwable')
|
||||
|
||||
def objOrArray(fi): # False on uninitialized
|
||||
return fi.tag == '.obj' or fi.dim > 0
|
||||
|
||||
def unSynthesizeType(t):
|
||||
if t in (T_BOOL, T_BYTE, T_CHAR, T_SHORT):
|
||||
return T_INT
|
||||
return t
|
||||
|
||||
def decrementDim(fi):
|
||||
if fi == T_NULL:
|
||||
return T_NULL
|
||||
assert fi.dim
|
||||
|
||||
tag = unSynthesizeType(fi).tag if fi.dim <= 1 else fi.tag
|
||||
return _makeinfo(tag, fi.dim-1, fi.extra)
|
||||
|
||||
def exactArrayFrom(fi, size):
|
||||
assert fi.dim > 0
|
||||
if size >= 0:
|
||||
return _makeinfo(fi.tag, fi.dim, fi.extra, size)
|
||||
return fi
|
||||
|
||||
def withNoDimension(fi):
|
||||
return _makeinfo(fi.tag, 0, fi.extra)
|
||||
|
||||
def withNoConst(fi):
|
||||
if fi.const is None:
|
||||
return fi
|
||||
return _makeinfo(fi.tag, fi.dim, fi.extra)
|
||||
|
||||
def _decToObjArray(fi):
|
||||
return fi if fi.tag == '.obj' else T_ARRAY(OBJECT_INFO, fi.dim-1)
|
||||
|
||||
def mergeTypes(env, t1, t2):
|
||||
if t1 == t2:
|
||||
return t1
|
||||
|
||||
t1 = withNoConst(t1)
|
||||
t2 = withNoConst(t2)
|
||||
if t1 == t2:
|
||||
return t1
|
||||
|
||||
# non objects must match exactly
|
||||
if not objOrArray(t1) or not objOrArray(t2):
|
||||
return T_INVALID
|
||||
|
||||
if t1 == T_NULL:
|
||||
return t2
|
||||
elif t2 == T_NULL:
|
||||
return t1
|
||||
|
||||
if t1 == OBJECT_INFO or t2 == OBJECT_INFO:
|
||||
return OBJECT_INFO
|
||||
|
||||
if t1.dim or t2.dim:
|
||||
for x in (t1,t2):
|
||||
if x in (CLONE_INFO,SERIAL_INFO):
|
||||
return x
|
||||
t1 = _decToObjArray(t1)
|
||||
t2 = _decToObjArray(t2)
|
||||
|
||||
if t1.dim > t2.dim:
|
||||
t1, t2 = t2, t1
|
||||
|
||||
if t1.dim == t2.dim:
|
||||
res = mergeTypes(env, withNoDimension(t1), withNoDimension(t2))
|
||||
return res if res == T_INVALID else _makeinfo('.obj', t1.dim, res.extra)
|
||||
else: # t1.dim < t2.dim
|
||||
return t1 if withNoDimension(t1) in (CLONE_INFO, SERIAL_INFO) else T_ARRAY(OBJECT_INFO, t1.dim)
|
||||
else: # neither is array
|
||||
if env.isInterface(t2.extra, forceCheck=True):
|
||||
return OBJECT_INFO
|
||||
return T_OBJECT(env.commonSuperclass(t1.extra, t2.extra))
|
||||
|
||||
|
||||
# Make verifier types printable for easy debugging
|
||||
def vt_toStr(self): # pragma: no cover
|
||||
if self == T_INVALID:
|
||||
return '.none'
|
||||
elif self == T_NULL:
|
||||
return '.null'
|
||||
if self.tag == '.obj':
|
||||
base = self.extra
|
||||
elif self.extra is not None:
|
||||
base = '{}<{}>'.format(self.tag, self.extra)
|
||||
else:
|
||||
base = self.tag
|
||||
return base + '[]'*self.dim
|
||||
fullinfo_t.__str__ = fullinfo_t.__repr__ = vt_toStr
|
||||
|
|
@ -671,4 +671,4 @@ into proprietary programs. If your program is a subroutine library, you
|
|||
may consider it more useful to permit linking proprietary applications with
|
||||
the library. If this is what you want to do, use the GNU Lesser General
|
||||
Public License instead of this License. But first, please read
|
||||
<http://www.gnu.org/philosophy/why-not-lgpl.html>.
|
||||
<http://www.gnu.org/philosophy/why-not-lgpl.html>.
|
||||
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in New Issue