Re-enable logger after catching exit

The logger is closed during at-exit hooks that get invoked by Kernel's exit method.

CHANGELOG.md

@@ -12,6 +12,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Fixed
 - Clear stubs defined with `expect().to receive()` syntax after test finishes to prevent leakage between tests.
 - Ensure stubs defined with `allow().to receive()` syntax are cleared after test finishes when used inside a test (another leakage).
+- Fix crash caused when logging is enabled after running an example that attempts to exit.
 
 ### Removed
 - Removed support for stubbing undefined (untyped) methods in lazy doubles. Avoids possible segfault.

src/spectator/system_exit.cr

@@ -20,6 +20,9 @@ class Process
   # Replace the typically used exit method with a method that raises.
   # This allows tests to catch attempts to exit the application.
   def self.exit(status = 0) : NoReturn
+    # Re-enable log that is disabled from at-exit handlers.
+    ::Log.setup_from_env(default_level: :none)
+
     raise ::Spectator::SystemExit.new(status: status)
   end
 end