9 commits

Author	SHA1	Message	Date
Admin	611a4e2035	ci: the tests are the platform's own, in release, in two minutes -- the user's law: "we mostly only care about the platform tests", "a max of about 2 minutes of tests", "test everything the CI does in release builds; debug builds are uselessly slow for this". ci.test now builds every test binary of its selection ONCE in release and runs them several at a time (half the cores, at least two), each one timed and capped, so a hung test costs its own cap and never the hour, and the step's detail names the slowest binaries with their counts; the doc tests follow through cargo for the packages that have a library. A selection is dirs (the packages whose manifests live under those directories, by cargo metadata), packages, package or workspace: true; a run over budget_secs (120 by default) turns the block orange and says so. The root script runs platform, draw, widgets and tools/ci by default: 49 binaries, 2535 tests, 22 s on a laptop. Every other crate's tests are the deep run, opt-in with deep_tests = true in ci.toml or ci --deep (ci.deep in scripts). The wall's cards lose a line: the time sits beside the count, 1 passed · 12s ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-22 09:50:38 +02:00
Admin	ab64a64e7e	tools: makepad-screen is makepad-agents, and its binary is agents -- the crate is the agent session manager (start, attach, list, the TUI); "screen" was a nod to GNU screen, and its other job, the pty trampoline a terminal starts its shell through, does not need the old name either. The package is makepad-agents in tools/agents, the executable is agents, so it is the command on the PATH itself and the shell wrapper that existed only to give it that name is gone (the binary already resolves the workspace's session directory the wrapper used to export); the terminal's pty spawn looks for agents beside the app, Director looks for and pins the sibling agents with the makepad-agents-v1 record and the agents version line, the session environment is MAKEPAD_AGENTS_SESSION and MAKEPAD_AGENTS_STATE_DIR, the CI scripts build makepad-agents, and Director's notes say how to build and run it ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-22 09:36:13 +02:00
Admin	c20cf99c89	ci: a judgement is never lost to the way it was phrased, and every "retry" of the bridge is retried -- the small vision model, asked to describe the desk's terminal, locked onto one phrase and repeated it to the token cap, so its answer had no verdict line and the window manager tile went red over a terminal that was fine; the decoder now stops a looping answer (the tail of the text is one short pattern over and over), and when the free answer has no verdict line the conversation is continued with "VERDICT:" already written and the model finishes that line, so the judgement is asked for outright instead of being lost. The harness retries every answer of the bridge that ends in "; retry" (a grab it could not arm as well as an input frame it could not place), where it matched one wording only and failed the browser step on the other. The workspace test suite gets three hours instead of one: it runs beside the app scripts now, and a hung test was worth a whole hour before ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-22 08:46:45 +02:00
Admin	863ca812a3	ci: the workspace script gives the wall back after warming -- after a push that touches a core crate the workspace script spent ten minutes and more alone (fifteen target checks, a release build of every app, the workspace check and the whole test suite) while every other tile sat grey, which read as a hang. It still goes first and alone for the part only it may do alone, warming the cache every other script reads, and then calls ci.shared(): its permit turns from exclusive into an ordinary one and the waiting scripts start beside its checks and tests. The runner sends it through the same pool as the rest and hands out nothing else until it holds the wall, instead of running it to the end before the pool even started. It also builds the pty helper while it is alone, so the terminal, director and window manager scripts get a cache answer instead of a cargo run that would queue behind the test build ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-22 00:07:23 +02:00
Admin	53d4960320	ci: a library's warnings turn the workspace block yellow, and a desktop tool is not warmed for the web -- warming blames nobody for an APP's trouble, since the app's own tile tells it, but a library has no tile: its warnings scrolled past in the log while the workspace block stayed green. The warm step now collects the warnings of every package that is not one of the apps being warmed, names the crates with their counts, and goes yellow. ci.warm and ci.check_targets take desktop_only, packages that are left out of the web, mobile and embedded rows; the root script lists director there, as its own script already asks only for the desktop targets, which takes its 139 compiled-out-worker warnings off the wall ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-21 23:56:25 +02:00
Admin	c091430d32	ci: a quiet header, a build that visibly moves, and scripts that leave the machine alone -- the window's top is one band: the branch and its tip, the test progress large (17 / 29) with failed and warning counts only when there are any, the run's elapsed time, Run now and Stop; the line under it is the run progress bar and the tiles start right below it, the rows of counts, "now:" and "Watching" are gone and next poll and the model's state moved to the one footer line; a running step carries what its command is doing ("compiled makepad_draw · 212 crates", told at most twice a second from cargo's artifact messages) and the running tile shows it, so a ten-minute warm build no longer looks stuck; director embeds a terminal, so its script builds the pty helper as the terminal's does; ci.launch takes app_env, and Files runs on its synthetic tree (MAKEPAD_FILES_DEMO) instead of walking the real home, which made macOS ask the person at the CI box for access to Downloads on behalf of "release"; and the workspace tests run with --no-fail-fast, so a run names every failing test target instead of the first ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-21 23:41:02 +02:00
Admin	fbe8c2e6c4	ci: the wall shows THIS run and is fit for an OLED -- a tile is grey until its script is tested, grey-blue with a progress bar while it runs, then grey-green, grey-amber or the one bright thing on the screen, a failure red; what the last finished run said is a small dot, never the tile's colour, and a script the user stopped or a restart interrupted goes back to untested instead of red; tiles carry their short name (wm, scope, workspace) in a named font family, since an empty family only rendered where a system fallback happened to exist and the tiles came up blank on the CI box; the packing picks the columns that give the largest whole name; a progress line says how many are tested, passed, warned and failed and what runs now; the pulse and the clocks tick on an 8 Hz timer only while something runs, never per frame on a 240 Hz display; the window no longer maximizes into native fullscreen over its left-half geometry, the background is near-black, the detail sits under the grid so it fits 960 points, a watcher problem is one header line, and the install button shows only when the vision model is missing; a compiler warning is orange and never red in the root script, and a desktop-only app with no library is not applicable on web and mobile targets instead of a standing warning ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-21 22:10:19 +02:00
Admin	0458f0c2b7	ci: a run is one target dir and one cargo batch per target, and the wall draws -- the root script warms a run-level cache with ONE cargo check per target covering every app package (the matrix's BuildTy per row, diagnostics attributed by package_id, --keep-going then per-package fallback when one package fails) and one release build of all bins, so the 24 app scripts answer check_targets and build from the cache; the tile grid registers its draw type on DrawQuad and merges its shader into the widget, which is what makes the green / orange / red / grey tiles appear at all; the window takes the left half of the main display (system_profiler, overridable with window = [x, y, w, h]); only apps/<name>/Cargo.toml is an app, so a private app's nested deps are its libraries; scope is no longer skipped by default; a failed launch hands the script an inert app instead of nil and a failure a step already carries is said once; the log keeps the first 20 diagnostics per cargo invocation and hub-install counts a present file once ... Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-21 21:52:43 +02:00
Admin	229cdd4d2e	ci: tools/ci, a Splash runtime that watches a branch and runs every ci.splash it finds -- each script checks its app on every other platform, builds and runs it here, drives it over --remote and has a small local vision model judge the grabs ... A test is a ci.splash beside what it tests; the script decides every input and the model only ever judges one picture against one acceptance text. mod.ci: launch (hidden, --remote, user_seq preserved), key, type_text, click, get, snap, wait_log (a * is a gap inside one line), no_errors, grab, quit; step, sleep, check, run; cargo, check_targets (the cargo makepad check matrix, check only for platforms we are not on, a test fails if the two tables drift), test, build, machine (another box over the makepad tunnel), exclusive; judge, accept, ask. The watcher polls git ls-remote once a minute for work and any extra branches, syncs a checkout the CI owns, runs the root script first and alone, then the rest up to a parallel limit behind one shared model judge. The window is a wall of squares, one per script: green passed, orange warnings, red failures, with a detail panel for the selected one. Scripts: the root ci.splash (workspace check with core warnings denied, the tests), apps/wm (desktop up, switch to macOS by Cmd+Space / type / Return, launch the terminal and the browser, each waited for by the WM's own first-frame line), and one per main app in the default shape. Proven here: apps/wm/ci.splash green in 280 s, fifteen target checks and seven vision verdicts. Models come from Hugging Face through the hub: registry entries qwen3.5-4b-vision and qwen3.5-9b-vision with exact revisions, sizes and digests, and hub-install, a command line over LocalModels::start_install. vlm-probe reads PNG. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-21 21:23:11 +02:00