The prod differential ladder finished: staged byte VOLUME on a warm load is
the only trigger left (V1a siblings-without-lib-tables dies, V1b +120 files
survives, V1c sibling KiCad files renamed byte-for-byte dies, V1d Leonardo +
123MB of inert markdown dies on loads 3-4; 14MB never dies). 3D models,
collab/ydoc/presence, lib tables, sibling KiCad handling and file count are
all exonerated — volume only loads the dice on the underlying race.
That made the crash reproducible locally for the first time in six campaigns:
a persistent browser profile + a 110MB project fails every warm load with the
exact prod signature. Iteration is now ~12 minutes instead of a release cycle.
Shim: every fiber switch now records the departing side's remaining asyncify
buffer and its recorded rewind entry (rem=/rf=), which is what identified the
unrewindable capture and disproved buffer overflow. The deferral family is
closed for good — a microtask-deferred retry on a clean empty stack died
identically to the nested rewind, because the suspension is broken at write
time, not by nesting.
Shell: log the origin stack when wx reports the top window destroyed. That
notification fires from ~wxTopLevelWindowWasm for ANY top-level window, so a
transient frame dying mid-load navigates the user out of the editor — a real
bug in its own right, found while chasing the empty flight-recorder dumps.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
The 96-event recorder ring holds under a second of history at idle tick rate
(~110Hz), so settle-time collision pressure scrolls out before any poll can
read it — the prod dumps only caught the kill because the trap froze the
moment. Scroll-proof totals since boot: every finishContextSwitch increments
fcsTotal; every root entry inside a sleep-wake window increments rootHotTotal
(the fatal precondition). Both appear in the [wx-asyncify] STATE line, i.e.
in every trap auto-dump, every __wxAsyncifyDump() call, and every blue-screen
console — turning ANY prod load (crashing or clean) into a dose measurement
for the differential-repro experiment (which project ingredient generates
collision windows: siblings, 3D models, libs).
.ci-cache-epoch 7→8.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
v0.1.24 in prod, doubly convicted the same morning: the Leonardo open
crawled/hung (open:settled result=failed at the 60s escape, heap never past
256MB — every main-loop iteration runs INSIDE its yield-wake extent, so the
"root-owned wake" scope matched thousands of legitimate nested coroutine
Call/returns per open, each paying a deferred macrotask, throttled to ≥1s in
a background tab), AND the Nano crashed 22ms after deferrals=1 fired.
Harmful and insufficient: the fatal nested-rewind interleave and the benign
bulk are observationally identical at this layer — no discriminator exists.
Retired (second and final retraction, async/16 round 5). What stays shipped
and clean: consume-once root suspensions, the internally-parked quarantine +
laundering check, the flight recorder + beacons, the WSOD floor, the
pendingSleeps leak fix (confirmed by pendingSleeps=[] in the Nano dump). The
rare nested-rewind crash is ACCEPTED and fully observable until the
structural fix — the design-B fiber-first runtime (async/06,12,13), where
one scheduler owns every suspension and this interleave cannot exist.
.ci-cache-epoch 6→7.
Local: fiber 2/2 (one refusal beacon) + timer + firefox sweep 21 passed,
chromium scenarios 11 passed/4 quarantine-skips, web fatal+follow 2/2.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
The v0.1.23 flight recorder caught the kill live (console-export-2026-8-1_19-16-8):
dozens of benign fiber round-trips at w=0, the yield cycling healthily on its
buffer — then "fcs … ROOT w=1" and the trap, state frozen at Rewinding with
currData=root+20. The fatal condition, observed rather than inferred: a fiber
round-trip inside the ROOT's OWN sleep-wake continuation re-suspends and
re-rewinds the root nested inside its live wake rewind. Consume-once passed
correctly — it guards a different corruption and stays.
The round-3 deferral was aimed right but unscoped (taxed fiber-owned wakes,
flaked S4). Final form: every fresh sleep is tagged root- or fiber-owned
(fiber ⇔ started inside a finishContextSwitch fiber slice or a fiber-owned
wake; root entries don't count as slices); finishContextSwitch(root) defers
one macrotask ONLY while a root-owned wake is live (Asyncify.__wakingRoot).
Beacon: root-entry-deferred. Verified inert where it must be: zero beacons
across all 13 drift-trio-scenarios logs (26/26 + 25/26-then-26/26 stress —
the single miss carried no beacons, i.e. the pre-existing under-load flake).
Also: resume re-entries no longer push sleep contexts (the v0.1.23 dump
carried ~380 leaked zero-linked entries), and wake events in the recorder are
tagged R/f for ownership.
.ci-cache-epoch 5→6.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
The deferral (9ca2ac1) modeled the wrong condition and taxed every parked
fiber completion with a macrotask hop: under CI load that stretched
three-client apply chains and flaked drift-trio S4 twice consecutively
(26/26 green locally under stress) — retracted.
The actual fatal state, readable in all four prod stacks once seen: a SECOND
rewind of the same root suspension. Root suspends once per fiber_swap out of
it; two parked fibers completing against one root suspension epoch (a tool
fiber + a collab fiber both waking around open:settled) each drive
finishContextSwitch(root) — the second rewinds already-consumed data →
"unreachable executed" → poisoned runtime, with the wake-side "index out of
bounds" as the sibling symptom.
Cure: stop exempting root from the validity check the shim already keeps.
First consumption proceeds synchronously — zero added latency anywhere; the
second is refused ([wx-asyncify] "root suspension already consumed") — the
yielded fiber stays properly suspended and resumable, root continues via its
real pending resume, libcontext's ghost-epoch contract enforced one layer
lower. Root remains exempt only from the internally-parked quarantine (its
yield park is routine).
.ci-cache-epoch 4→5 (the epoch-4 cache holds the retracted deferral shim).
Local: fiber 2/2 + timer 1/1, firefox sweep 20 passed, drift-trio-scenarios
kicad-chromium 26/26 under 3-worker stress, web fatal+follow 2/2.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
v0.1.22 still trapped with BOTH guards silent: the fatal rewind's target is
the ROOT context, which layer 2 exempted. All four prod stacks are the same
collision — a fiber completes its yield-back to main while main's sleep-wake
rewind is still on the stack (maybeStopUnwind → trampoline →
finishContextSwitch → doRewind(root) → unreachable), two "resume main" paths
interleaved in one tick; the 8ms-earlier "index out of bounds" is the wake
side of the same event.
Root entry is legal and constant in healthy flow; only the wake-window
overlap is fatal. So: serialize, don't refuse. The shim marks the
synchronous wake window (Asyncify.__inSleepWake around wakeUp) and DEFERS a
root finishContextSwitch landing inside it by one macrotask
([wx-asyncify] root-entry-deferred beacon, trampoline retry) — an ordering
change only, nothing dropped. Suspension recording happens before the
deferral branch, so the yielding fiber's validity survives the wake chain
nulling currData.
Plus a flight recorder: a 96-entry ring of asyncify/fiber events (sleeps,
wakes, every context switch with ROOT/wake-depth, refusals, deferrals),
silent in normal operation, auto-dumped with full machine state next to the
first trap signature in the console; window.__wxAsyncifyDump() on demand.
The next prod export reads like a black box, not a stack-shape puzzle.
.ci-cache-epoch 3→4 (wasm cache key omits scripts/**).
Local: fiber-resume-park 2/2 (one refusal beacon), timer-park 1/1, sweep 20
passed, web fatal+follow 2/2.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
v0.1.21 still trapped with ZERO jump-refused beacons: the fatal swap PASSED
the C++ swap_suspended guard. Mechanism (async/16 round 2): a fresh JS entry
executing while g_current_context still points at a parked fiber gets
attributed to that fiber — fiber_swap writes a fresh, valid-LOOKING foreign
suspension into the parked fiber's struct and re-marks the flag. The flag
lies; the resume rewinds garbage.
This guard tracks truth at the emscripten-fiber layer (handlesleep.js wraps
Fibers.finishContextSwitch):
- valid suspensions = real swap-outs (currData == oldFiber+20 when the
trampoline runs), consumed on rewind;
- internally-parked = an entered slice that ended in a handleSleep park
(currData set, no nextFiber) — quarantined until a GENUINE swap-out,
where genuine means the fiber's pending sleep has resolved
(__pendingSleepContexts), so a laundering write cannot lift it;
- entering a quarantined or suspension-less fiber is REFUSED
([wx-asyncify] fiber-resume-refused, ghost contract).
The ROOT context is exempt from quarantine and refusal: its rewound
continuation runs the whole main loop, whose routine yield park says nothing
about a fiber body — the first build of this guard quarantined main off that
signal and starved every coroutine return (19 collab e2e reds, empty
results). Root = the old side of the first switch ever.
.ci-cache-epoch 2→3: the wasm output cache key omits scripts/**.
Red/green: fiber-resume-park.spec.ts scenario 2 (laundered resume → exactly
one refusal beacon, both coroutines complete); full fiber-heavy sweep green
(21 passed).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
The shim has been silently restoring Asyncify.currData across overlapping
parks since it was written — so a prod console dump could never say whether
the #9153 family fired, mislinked, or was bypassed. Rate-limited [wx-asyncify]
reporting, calibrated against the timer-park e2e (2 rounds, 2026-07-31):
- concurrent-park: a FRESH park (state 0) starting while another chain's
currData is live. state-2 entries are NOT reported — every legal resume
re-enters handleSleep while Rewinding (~100/s measured on a healthy run).
- reentrant-state: handleSleep entered mid-unwind (state 1) — never observed;
if it ever fires it IS the bug.
- aliased-wake-live / overlapped-wake: the restore actually repairing —
dangerous (other chain still parked) vs benign (other chain done). The
e2e observed 8 live cross-restores in one run, all survived.
Console.log only, no printf plumbing needed; stacks name the parker
(__asyncjs__wxWasmYieldToBrowser vs _emscripten_sleep etc.).
.ci-cache-epoch 1→2: the wasm output cache key omits scripts/**.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
Move OpenCASCADE out of the merged editor image into occ_service: a separate
emscripten module (-sASYNCIFY=0, MODULARIZE, in-container -Oz finalize, 2N+8
pre-warmed pthread pool) booted lazily in a dedicated Web Worker on the first
STEP export or STEP/IGES model parse. kicad_editor.wasm ~190 MB -> 130 MB;
sessions that never touch OCC never fetch its 57 MB. STEP export works in the
browser for the first time: the unchanged desktop dialog runs EXPORTER_STEP,
whose wasm shadow suspends into globalThis.occService and the export bytes go
straight to a browser download (never entering the editor heap). STEP/IGES 3D
models parse in the worker via the oce shadow (S3D WriteCache/ReadCache wire).
- wasm/occ-service/: service CMake target (hooked from the kicad fork's
top-level CMakeLists, wasm/editor pattern), embind entry
(occExport/occLoadModel), wxConfig pre-js.
- wasm/stubs/{exporter_step,oce_plugin}_stub.cpp: EM_ASYNC_JS worker bridges
(callee-shadowing; no caller #ifdefs).
- web/standalone: provider installed whenever the kicad_editor bundle boots
(cross-face safe); ONE shared worker-boot source occ-worker.js (vite ?raw;
the e2e stub reads the same file) — blob worker with locateFile absolutized
against the glue URL; export download-name guard.
- deps: OCC builds with RapidJSON so its glTF/GLB writer exists — pinned to
the vcpkg master snapshot 2025-02-26 (24b5e7a8b27f), the same code official
KiCad consumes via vcpkg.json's opencascade[rapidjson]; rapidjson's latest
tag (v1.1.0, 2016) is ill-formed under modern clang.
- tests: occ-export dialog e2e (lazy-fetch boundary + STEP download bytes),
occ-probe incl. a 9-format matrix (step/stpz/brep/xao/ply/stl/glb/u3d/pdf),
3d-viewer-models hard-asserts the worker parse; occ provider stub installed
ambiently by the kicad fixtures.
Validated against desktop kicad-cli 10.0.4: geometric exact equality (bbox
delta 0 um, volume delta 0.0000%) for STEP/GLB/STL/BREP/STPZ across three
boards and option sweeps — with desktop OCC 7.9 vs wasm OCC 7.8; PLY/XAO/PDF
structurally equal; U3D same-size (quantizer float LSBs differ). Full kicad
e2e green on Firefox and Chromium; standalone verified end to end (lazy fetch
only on the Export click; export.step 60,628 B ISO-10303-21; loadModel 700 KB
STEP -> 569 KB scenegraph cache).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Binaryen wasm-opt shrink after Asyncify was the only -O2 in the
pipeline: -O2 by default (local + tag release) but -O1 in main CI. The
C++/wx/deps compile is already -O1 everywhere (DEBUG_BUILD defaults to 1;
nothing passes --release in CI/release). Pin the tail to -O1 too:
- apply-asyncify.sh: BINARYEN_OPT_LEVEL default -O2 -> -O1
- release.yml: opt_level -O2 -> -O1 (demo now ships -O1)
- ci-ubicloud.yml / wasm-build.yml / docker/build.sh: refresh stale -O2 comments
Because both callers now build at -O1, the FINAL cache key converges, so a
tag release FINAL-cache-hits main's build and skips the asyncify-tail
rebuild entirely (previously the ~1-2h -O2 wasm-opt was rerun per release).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The emscripten glue's `var ENV = {};` never merged Module.ENV, so boot.ts's
Module.ENV.KICAD_TRACE was a silent no-op — and since environ_get on a pthread
proxies to the main thread, the app/UI thread's getenv read the (empty) main
ENV. Net: KICAD_TRACE never arrived and every KI_TRACE was a no-op.
scripts/common/patch-env-shim.mjs post-processes each generated <app>.js to
merge Module.ENV into ENV (idempotent; runtime no-op when unset), wired into
docker/build.sh's per-app host post-process next to the dyncall-shim inject.
Replaces a fragile manual glue edit re-applied after every build. Verified:
symbol-editor eager load now emits KI_TRACE_SYM_CHOOSER fatLoad timing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The native-EH migration pointed apply-finalize.sh at emsdk's bundled
wasm-emscripten-finalize (${EMSDK:-tools/emsdk}/upstream/bin), which exists on a
dev machine (tools/emsdk persists from a prior build) but never on the ephemeral
CI host — emsdk lives only inside the Docker image, and nothing on the host
post-process path provisions it. So the host post-process died at finalize for
every app once the binaryen-SHA cache re-key forced it to actually run. (Not the
missing-ninja theory — ninja was installed; the run never reached the wasm-opt
build, which is downstream of finalize.)
Finish the "Binaryen submodule everywhere" migration: build-wasm-opt.sh now also
builds wasm-emscripten-finalize, and apply-finalize.sh takes it from that build
(next to wasm-opt). The host post-process is now emsdk-free (dyncall=node,
finalize+asyncify=submodule v130) and finalize/wasm-opt share one Binaryen
version (previously finalize was emsdk's v121, wasm-opt the submodule's v130).
get-wasm-opt.sh stays for bench only; dropped it from the wasm cache key and
added build-wasm-opt.sh.
Also surface the from-source Binaryen build as a "Build Binaryen" stage in
build-monitor.sh (it had no marker), emitted at the pipelined pre-warm sites.
Validated: a clean from-scratch `docker/build.sh calculator --build-deps` (deps +
compile + finalize via the submodule binary + asyncify + -O2) succeeded
end-to-end; finalize resolved to build-wasm/tools/binaryen-hoist-build/bin.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the legacy Emscripten JS-exceptions model with native wasm-EH (legacy
encoding) across the whole build, keeping Asyncify coroutines working via a
from-source Binaryen --hoist-cpp-catches pre-pass. Net result: native-EH is the
only build mode, the 3D viewer is on by default, and pcbnew shrinks substantially.
Highlights:
- Binaryen submodule everywhere + --hoist-cpp-catches integration in apply-asyncify;
post-link Asyncify covers every app wasm (not just standalone test wasm).
- Build deps (incl. OpenCASCADE without OCC_CONVERT_SIGNALS) and all KiCad apps
with -fwasm-exceptions; emscripten_sleep added to the post-link asyncify-imports.
- libcontext fiber entry wired under native exceptions; while-loop main loop +
currData shim injected into all wx apps.
- Native-EH collab apply fixed: DEBUG-define the embind TU + match all out-of-CMake
C++ TUs' ABI flags to the core, fixing the vtable-layout skew / mis-dispatch.
- 3D viewer enabled by default (real raytracer linked, not the stub).
- Retire the EH-spike scaffolding; flip the asyncify-races ablation pins to
shim-redundancy pins (native-EH stays clean with the legacy shims ablated).
- Fix the asyncify-races quiescence check to not require Asyncify.currData==0:
under the native-EH per-frame-yield top loop the main stack is asyncify-suspended
every frame, so currData legitimately churns (a freed-but-not-yet-nulled buffer,
not a leak). Refresh the pcbnew toolbar screenshot baseline for the new kicad.
- CI: drop the obsolete binaryen_version input/env (the build uses the binaryen
submodule fork's wasm-opt, not a version download); key the wasm-output cache on
the binaryen submodule SHA instead.
Bumps the wxwidgets + binaryen submodules to their squashed feature commits.
Validated green: all 7 apps native-EH (real 3D in pcbnew); KiCad e2e 63/63
Firefox + Chromium (3D viewer renders); wx 336; coroutine 34/34 both engines;
asyncify 7/7 both engines.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Commit 7dc1bf2 renamed scripts/build-wxuniversal-wasm.sh to
build-wx-wasm.sh but missed the CI workflow, so the "Build wxWidgets
(wxUniversal WASM)" step ran a script that no longer exists. Update the
run command in ci-ubicloud.yml plus the two stale comments referencing
the old name (ci-ubicloud.yml:43, scripts/common/stages.sh:7).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The canvas (wxUniversal) mode is gone (wxwidgets submodule); remove every
piece of side-by-side plumbing so there is exactly one build and one test
flow:
- scripts/build-wxuniversal-wasm.sh -> scripts/build-wx-wasm.sh; no
--dom/--enable-universal; builds into build-wasm/wxwidgets
- build-wasm-test.sh: no DOM_BUILD / apps-dom rsync mirror / PORT=dom;
apps build straight into tests/apps (Makefile.wasm PORT conditionals
collapsed; wx.js + wx-dom.js always pre-js)
- docker/build.sh, build-kicad-target.sh, env.sh: WX_PORT / -dom /
-universal suffixes removed; kicad builds to kicad-<app>, outputs to
output/; wx.js/wx-dom.js copied from the real source path
(/workspace/wxwidgets/build/wasm — the old build-wasm path never
existed and silently failed)
- setup-kicad-wasm.sh: single target dir; the perl wx-dom.js injection is
gone — the 7 checked-in kicad pages now reference wx-dom.js directly
- playwright configs serve apps/; fixtures drop the test-results/dom and
logs/wxwidgets/dom namespacing; boot.spec asserts wxDomPort
unconditionally; pcbnew.spec uses one reference image;
appearance.spec assertions unconditional
- compare/update-baseline-screenshots.sh: --port removed
- tests/gal-regression/wasm/Makefile: links build-wasm/wxwidgets and
carries wx-dom.js as a second pre-js — the gal-webgl suite (30 specs)
now actually builds and runs here (it needed host-side boost+glm via
scripts/deps; the bundle had been missing, timing the whole spec out)
- tests: clickCanvas() dispatches via page.mouse (DOM widgets
legitimately cover the canvas; locator actionability refused the
click); the comprehensive spec drives wxChoice through its native
<select> (browser-owned popup cannot be coordinate-clicked)
- docs: README/CLAUDE.md/build.md script names and dirs;
features/wx-dom-port README reframed (DOM is THE port), visual-notes
bugs 26-28; FindwxWidgets.cmake config label drops 'wasmuniv'
- wxwidgets submodule -> 9dbacc9448 (DOM-only port, fork diff shrunk)
Gate: full wx e2e suite 292 passed / 1 skipped / 0 failed — first run
ever with the gal-webgl specs green (28 scenarios + load + sequential).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The asyncify single-slot work, executed red-green (full ledger:
docs/features/asyncify-arbiter/redgreen.md; decisions record:
docs/features/async/07-decisions-and-outcome.md):
- tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated
playwright config: 8 scenarios reproducing the KiCad asyncify failure
family with the kicad-faithful startup topology (pre-park fiber swap →
park throw through the live trampoline). Built in 3 variants; the
SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds
keep the historical hang and index-out-of-bounds crash reproducible
forever (mutation-style pins for the existing shims).
- scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel
in the wakeUp path — when main's last pre-park suspension was a sleep,
the main-loop park throw escaped through that sleep's promise reaction
as an uncaught rejection (the calculator/gerbview console errors).
- scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs.
- Spec tightening (the acceptance bar): 'uncaught exception: unwind'
tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard
clean-console gate over 5 asyncify corruption signatures.
- wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error,
sync clipboard IsSupported (014f67e6c1).
Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40
passed / 2 skipped with ZERO corruption signatures in any log across all
six apps.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
build-kicad-target.sh swaps emsdk's wasm-opt/wasm-emscripten-finalize for
no-op stubs (real binaries preserved as *.real) so emcc skips in-link
asyncify — intended for the CONTAINER emsdk, but a host-mode run left the
host tools/emsdk stubbed (since Jun 9). The stub fakes --version and exits 0,
so every local build's host-side finalize/asyncify/-O2 "succeeded" while
doing nothing: output wasm shipped non-asyncified and aborts at boot with
"asyncify_stop_unwind is not a function" (and stayed 122M vs the correct
187M). get-wasm-opt.sh and apply-finalize.sh now prefer the *.real binary
whenever it exists, making the resolution immune to a stubbed emsdk.
Found while validating the CI e2e fix: pcbnew.wasm built this morning could
not boot in any browser. Local artifacts built since Jun 9 may need their
postprocess re-run (apply-finalize.sh + apply-asyncify.sh).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three orchestration fixes, validated end-to-end on the Hetzner ccx53
(run 27280051992, 1h14m41s vs 4h05m baseline, e2e identical):
- docker-compose CPU/memory caps were hardcoded to dev-Mac defaults
(10 CPUs / 32G); now env-tunable via KICAD_DOCKER_CPUS/KICAD_DOCKER_MEM
(CI exports nproc/110G — the 32-core runner was compiling on 10 cores).
- docker/build.sh: split build_app into compile_app (container) +
postprocess_app (host-side dyncall shims + finalize + asyncify + -O2)
and added KICAD_PIPELINE=1 mode that overlaps each tool's host-side
wasm-opt with the next tool's container compile (max 2 concurrent
postprocesses — pcbnew -O2 peaks ~34G RSS). Also: comma-separated app
lists for cheap pipeline repros.
- get-wasm-opt.sh: Binaryen default 121 -> 130 (fixes the v121 -O2 lock
convoy, ~9x) and BINARYEN_BUILD_FROM_SOURCE=1 support: the official
x86_64-linux release tarballs (Alpine/musl, no LTO, assertions on) run
asyncify 4x slower than a stock gcc -O3+LTO build with sha256-identical
output; self-build takes ~5 min and is instantly repaid.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two things, both verified in the real web app (two-tab eeschema collab).
1. dynCall crash fix (all apps) — scripts/common/shims/dyncall-binding.js.tmpl.
Programmatic editor edits trapped with 'indirect call signature mismatch': the
asyncify-instrumented wasmExports[dynCall_<sig>] trampoline does call_indirect with a
stale type for some table indices (post-asyncify+O2) even though the table entry is
valid. Proven by patching the built js: at the trap getWasmTableEntry(index) SUCCEEDS
where the trampoline fails. Fix: the shim now catches the 'signature mismatch'
RuntimeError and falls back to getWasmTableEntry; the Asyncify unwind sentinel and real
exceptions re-throw, so instrumentation/unwind is untouched for normal calls. This
unblocks ALL programmatic edits, not just collab (e.g. eeschema SCH_ITEM::Move).
2. eeschema collab apply converters (wasm/bindings/eeschema_embind.cpp).
doApply now handles added-item construction (build the SCH_ITEM with the delta's uuid
via const_cast — as the s-expr parser does — + commit.Add) and richer SCH_LINE
serialization (start/end/layer) so wire edits reconstruct on the peer. Implemented for
SCH_LINE (wires) + SCH_JUNCTION; other types log 'no converter for added type' and are
skipped (next batch). eeschema re-enabled in the web app collab gate.
Tests: eeschema-collab.spec snapshot (green); apply/two-tab skipped — they no-op headless
because the e2e harness's kicadOpenFile returns false (OpenProjectFiles bails before
building the connectivity graph), so SCH_COMMIT::Push doesn't persist. Verified in-app.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The emscripten fiber glue gates Fibers.trampoline() on Fibers.trampolineRunning
and resets it at the end of its loop. At startup emscripten_set_main_loop(...,1)
throws "unwind" to establish the main loop, and KiCad does so from inside a tool
coroutine, so the throw propagates THROUGH the trampoline and skips the reset —
leaving the flag stuck true. Every fiber swap after startup then becomes a silent
no-op, so opening a schematic (SetScreen -> RunAction(selectionClear) -> fiber
swap) hangs forever with the editor stuck on "untitled".
Wrap the trampoline loop in try/finally (inject-dyncall-shims.sh section "3c") so
the flag is always reset. Add tests/kicad/eeschema-load.spec.ts, which opens a
small wires/junctions schematic via Module.kicadOpenFile and asserts the editor
title switches away from "untitled": it times out (RED) without the shim and
passes (GREEN) with it. Also add features/web-init/0002-url-regex-modal-followup.md
capturing the unrelated URL-detection wxRegEx modal surfaced once loading works.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- inject-dyncall-shims.sh: add a dynCallLegacy -> wasmExports fallback so embind's
generic dynCall path finds the DYNCALLS=1 trampolines. Without it, an Asyncify
unwind/rewind through an embind call (kicadOpenFile -> OpenProjectFiles) died
with "f is not a function".
- open-flow.ts: kicadOpenFile runs OpenProjectFiles under Asyncify, so its sync
return is a falsy placeholder. Invoke it and poll the frame title for the load
instead of trusting the return value, and never fall back to UI automation
while the hook is in flight (it would re-enter the suspended Asyncify call).
- .gitignore: ignore .playwright-mcp scratch.
- bump kicad + wxwidgets submodules to the wasm schematic-open fixes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two related fixes for Chrome-specific WASM-runtime issues reported when
running a manually-loaded session (line tool wouldn't even toggle on click;
log filled with 'Uncaught (in promise) unwind' and stderr-tagged
[WASM_FCONTEXT]/[DIAG_*] spam):
1) scripts/common/apply-asyncify.sh — run 'wasm-opt -O2' as a separate
pass after '--asyncify'. Without this, large asyncify-instrumented
coroutine-entry trampolines (notably libcontext's wasm_fcontext_entry
and COROUTINE<int,TOOL_EVENT const&>::callerStub) exceed V8's
per-function locals limit and silently stall on first fiber entry,
leaving the toolbar click dispatched in C++ but the tool never
activating its 'running=1'/[checked] state in the user's Chrome.
Firefox tolerates the unoptimised version, so tests on Firefox passed
while real Chrome stalled. The -O2 pass shrinks every instrumented
function back under the threshold, fixing the family of stalls
systemically (no more per-function removelist whack-a-mole).
The removelist still contains setupUIConditions() etc. as a safety
net — they're now redundant under -O2 but harmless.
Bundle: 338 MB -> 187 MB raw (~45% smaller); test runtime nearly
halves because parse is faster. See DEBUG.md §7 and
memory/bundle-size-asyncify-optimization.md.
2) wxwidgets submodule bump (d1d1627 -> a998a8d) — wasm/dialog.cpp:
startModal()'s setTimeout-based runEventLoop now awaits
ccall('ProcessEvents', ..., {async:true}) so the Promise rejection
from an asyncify-suspended ProcessEvents is caught by the existing
try/catch instead of escaping as an 'Uncaught (in promise) unwind'
page error.
Verified: npm run test:kicad:chrome and test:kicad:firefox both pass on
the rebuilt wasm; zero pageerror events; user-reported manual flow now
selects the Draw Lines tool and draws successfully.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- build-pcbnew.sh: add --diag=<gal,coroutine,ctor,all> -> -DKICAD_DIAG_*,
off by default (forwarded by docker/build.sh)
- diagnostics.js: emit at console.log level (no longer error/warn); still
gated by SHIM_DIAGNOSTICS=1
- apply-asyncify.sh: exclude PCB_EDIT_FRAME::setupUIConditions() from
asyncify instrumentation (V8 cannot run the instrumented huge function
on the rewound ctor stack -> Chrome startup stall; Firefox unaffected)
- DEBUG.md: reusable WASM/asyncify/browser debugging guide, diagnostic
flag docs, and a production-build (release + -O2 asyncify) recipe
- tests: standalone coroutine vcall/gl repro probes
- bump kicad + wxwidgets submodules (diagnostic gating / debug cleanup)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Investigation scaffolding for the Chrome-only KiCad coroutine renderer crash.
Adds isolated reproduction probes exercising the coroutine/Asyncify/fiber layer
under KiCad-like conditions, runnable in BOTH Firefox and system Chrome.
- tests/playwright-coroutine.config.ts + test:coroutine:firefox|chrome npm
scripts: run the coroutine specs in Firefox AND system Chrome (the old e2e
config only used bundled Chromium, which never reproduced the crash).
- tests/apps/standalone/coroutine-pthread/: no-wx + pthreads reproduction probes
(fiber-in-main, nested invoke_/dynCall boundaries, RunMainStack, embind,
main-loop/rAF activation) + worker_dom_stub.js for wx+pthreads builds.
- tests/apps/Makefile.wasm: coroutine-pthread{,-main,-nested,-nested-ex,-wx,
-embind,-mainloop} targets.
- scripts/common/shims/diagnostics.js: add EM_ASYNC_JS handleSleep enter/wake
tracking (DIAG_SLEEP) to detect nested-async at the crash.
Findings (details in research notes): every isolated factor so far — direct /
nested / RunMainStack fiber, wx event loop + all 13 scenarios incl EM_ASYNC_JS,
pthreads, and main-loop/rAF activation — runs CLEAN in system Chrome. The
coroutine/Asyncify layer is exonerated; GL/WebGL is the remaining untested factor
(next). The reliable FF-pass/Chrome-fail repro is still the KiCad pcbnew e2e.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The shim bound bare dynCall_* names to JS getWasmTableEntry() calls, bypassing
the asyncify-instrumented dynCall_* wasm trampolines that -sDYNCALLS=1 provides.
That broke Asyncify unwind/rewind through indirect calls -> "indirect call
signature mismatch" (caught every frame in Firefox; fatal renderer crash in
Chrome). Bind the bare names to wasmExports["dynCall_<sig>"] instead.
Result: the PCBnew "select draw lines" e2e is green in Firefox (tool selects and
draws, zero page errors). Dropped the fiber-stabilization block, the shipped
diagnostic block, and the exportCallStack JS hack (all compensated for the wrong
binding); shim shrank 521 -> ~250 lines.
- scripts/common/inject-dyncall-shims.sh: orchestrator only; injected JS extracted
to scripts/common/shims/
- scripts/common/shims/dyncall-binding.js.tmpl: per-signature binding template
- scripts/common/shims/handlesleep.js: nested-Asyncify handleSleep fix (#9153)
- scripts/common/shims/diagnostics.js: logging-only, opt-in via SHIM_DIAGNOSTICS=1
- tests/package.json: add test:kicad:firefox / test:kicad:chrome scripts
Known issue (tracked separately): Chrome still renderer-crashes on the first
coroutine resume. Asyncify.doRewind replays the deep main-context call stack and
exceeds V8's execution-stack limit (Firefox tolerates the same wasm). Proper fix
is JSPI.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Main-repo side of a multi-part WIP covering the KiCad WASM tool-selection
and nested-Asyncify work. Submodule commits are in kicad@f6e9239aaa
(libcontext hygiene) and wxwidgets@bb80f91e8b (auibar registration +
dialog diagnostics).
## scripts/common/inject-dyncall-shims.sh
Wrap Asyncify.handleSleep / allocateData to save-and-restore Asyncify.currData
around each EM_ASYNC_JS sleep. This fixes the nested Asyncify collision where
a fiber swap that fired during a modal's event loop clobbered currData, and
the modal's later doRewind used the fiber's buffer and hit "RuntimeError:
index out of bounds". Root cause documented as Emscripten Issue #9153
(wontfix upstream).
Diagnostic-rewind logging (forcedBottomOfCallStack, callStack traces) is
retained to help future debugging of Asyncify state corruption.
## tests/
- tests/playwright-kicad.config.ts: add `channel: 'chrome'` for the
chromium project so --project=chromium --headed uses system Chrome
(real GPU) instead of SwiftShader on ARM Mac. Also switch trace to
retain-on-failure + screenshot on-failure for easier E2E debugging.
- tests/kicad/pcbnew.spec.ts: replace `tool.checked` assertions with a
label-suffix check (`[checked]`) since our auibar registration encodes
checked state in the label (no schema change to the registry).
- tests/apps/Makefile.wasm: add `coroutine-nested` build target + include
it in the all: list.
- tests/apps/standalone/coroutine/: kicad_coroutine_harness.h + test app
reproducing KiCad COROUTINE semantics against real libcontext.
- tests/apps/standalone/coroutine-nested/: nested_test.cpp reproduces the
EM_ASYNC_JS-modal + fiber-swap nesting bug in isolation. 8 scenarios
from baseline_modal_alone through nested_fibers_inside_modal.
- tests/e2e/coroutine.spec.ts + coroutine-nested.spec.ts: Playwright specs
that load the standalone apps and assert all case cases pass via
[COROUTINE_TEST] SUMMARY log parsing.
## research/ and features/browser-tools/
Three background docs capturing the investigation trajectory:
- features/browser-tools/0001-kicad-wasm-tool-activation-investigation.md
Early investigation: why tools don't activate; initial dynCall-empty-
callback hypothesis.
- features/browser-tools/0002-wasm-coroutine-deep-dive.md
Deep dive on Asyncify internals, fiber API, QEMU's coroutine-wasm
reference implementation.
- features/browser-tools/0003-wxauitoolbar-registration-fix.md
The narrow fix: why wxAuiToolBar needs a registration block, where to
add it, what the fallback plan is.
- research/threading_1.md: corrected root-cause analysis after reading
runtime logs — nested-Asyncify currData collision, Emscripten #9153.
- research/threading_2.md: extended research on alternative approaches
(JSPI/WasmFX/state-machines) and why they don't help here.
## Submodule pointer updates
kicad: f6e9239aaa (wip: libcontext WASM hygiene cleanup)
wxwidgets: bb80f91e8b (wip: wxAuiToolBar element-registry registration +
dialog diagnostics)
## Open threads not yet in scope
- Firefox/Chrome divergent behavior: "indirect call signature mismatch"
traps in Firefox vs renderer crash in system Chrome (tracked in
plans/peaceful-hugging-pnueli.md and the research docs).
- E2E pixel-diff for Draw Lines fails because the test's diff region does
not cover where the line is actually drawn; tool activation works, the
line is visible in test-results/pcbnew-draw-lines-02-after-drawing.png.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Docker build used emscripten/emsdk:4.0.2-arm64 as base image but
env.sh couldn't find emsdk there, installing a second copy. The build
then applied wasm-opt/finalize stubs to the wrong emsdk (hardcoded
/emsdk/), so the real wasm-emscripten-finalize ran in Docker and got
OOM-killed.
- Use ubuntu:22.04 base with emsdk installed from source at /emsdk/
- Make stub paths dynamic via $EMSDK instead of hardcoded /emsdk/
- Skip local emsdk install in env.sh when $EMSDK is already active
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Emscripten 4.x removed dynCall_* WASM exports, breaking asyncify
rewind through indirect calls (modal dialogs, event handlers).
Generate JS shims that track Asyncify.exportCallStack and register
in wasmExports so doRewind can find them.
Also fixes empty callback functions ((() => {})) generated by
Emscripten 4.x + pthreads for HTML5 events, pthread entry,
sighandler, async timer, and main loop callbacks.
Build pipeline improvements:
- Stub wasm-opt/finalize in Docker (RAM limits), run on host
- Add setup-emsdk.sh for reproducible Emscripten setup
- Simplify env.sh and version management
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Emscripten with pthreads generates empty arrow functions `{}` for callback
paths it assumes won't be used. However, when registering HTML5 events from
the main browser thread, targetThread is 0 and the direct call path IS taken.
This fix post-processes the generated JS to replace empty callbacks with
actual dynCall invocations for:
- HTML5 event callbacks (dynCall_iiii) - 7 instances
- pthread entry points (dynCall_ii) - 1 instance
- Signal handlers (dynCall_vi) - 1 instance
- Async timers (dynCall_vi) - 1 instance
Also improves build logging to show completion status and exit code.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add complete test infrastructure for WebGL GAL visual regression testing:
- scripts/test-gal-regression.sh: Master script that builds both backends,
runs tests, and performs two-level comparison (native vs baseline,
webgl vs native)
- scripts/build-gal-webgl-test.sh: WASM build using Makefile with em++
- tests/gal-regression/wasm/: WebGL test harness (stub WEBGL_GAL)
- tests/e2e/gal-webgl.spec.ts: Playwright test for screenshot capture
Fix Homebrew Emscripten environment in scripts/common/env.sh:
- Set EMSDK_PYTHON for Python 3.10+ (em++ reads this, not $PYTHON)
- Add bundled LLVM to PATH (Emscripten needs its clang with WASM backend)
Verified: Native vs Baseline passes (28/28), WebGL generates blank
screenshots as expected (WEBGL_GAL implementation is Phase 2).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
wxwidgets submodule:
- Add m_isCreated flag to skip UpdateElementRegistry() during construction
- Prevents virtual method calls on partially constructed objects
scripts/common/logging.sh:
- Change -d to -e for .git check to support git worktrees
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Restore missing elementtracker.h in wxWidgets (required for E2E tests)
- Add automatic log file redirection for build scripts
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add autoreconf step for bundled PCRE in wxWidgets build script
(fixes automake version mismatch between 1.16.1 and 1.16.5)
- Fix apply-finalize.sh to use get-wasm-opt.sh for consistent tool paths
(tools moved to build-wasm/tools/ in previous commit)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Groups cached build tools (binaryen) with other build artifacts.
The build-wasm/ directory is already gitignored via build-*/ pattern.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add rsync-based source sync for Docker builds to fix macOS VirtioFS
timestamp inconsistencies that caused autoconf sanity checks to fail
- Add config.sub wrapper (SHELL/CONFIG_SHELL) to support emscripten/wasm32
targets without modifying submodule files
- Fix protobuf build to use native gcc/g++ for protoc instead of Emscripten
- Add python PATH fix for macOS (Homebrew's python3 symlink) in wxWidgets build
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Build improvements:
- Change -O0 to -O1 to fix "local count too large" asyncify error
- Add wasm-emscripten-finalize to host (Docker OOMs on large WASM)
- Use -gseparate-dwarf for smaller main binary with debug info
- Build native protoc for code generation
- Add more functions to asyncify removelist
Runtime fixes:
- Add inject-dyncall-shims.sh to fix "dynCall_* is not defined" in Emscripten 4.x
- Update wxwidgets with browserInfo.name fix
New stubs and bindings:
- Add Embind bindings for JavaScript interop
- Add stubs for scripting, API plugin, PCB frame, navlib
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add source hashing to stamp system (functions.sh) for detecting
when dependencies need rebuild based on source file changes
- Update build-pcbnew.sh to use source stamps for wxWidgets
- Add global-setup.ts to clean logs before test runs
- Add maximize_test standalone test to verify wxFrame::Maximize()
works correctly in WASM (it does - window is 1280x720)
- Update Makefile.wasm with maximize test build rules
The maximize_test proves wxWidgets display detection works fine.
KiCad's 20x20 window bug is KiCad-specific, not a wxWidgets issue.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Move asyncify from Docker to host to avoid memory issues
- Auto-download Binaryen v121 (v125 has regression bug)
- Use -O1 for debug builds (V8 local count limit)
- Remove asyncify flags from linker (handled by wasm-opt)
- Document two-phase build in build.md
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add build.md documentation for the WASM build system
- Default to debug builds with -gsource-map for WASM debugging
- Add --release flag to disable debug symbols
- Add --clean flag for full rebuild (deps + wxWidgets + KiCad)
- Add -j N flag for parallel compilation (default: 1 for sequential)
- Update Docker memory limit to 32GB
- Propagate DEBUG_BUILD and JOBS to all dependency scripts
Build modes:
- ./docker/build.sh --clean -j8 (full rebuild, parallel)
- ./docker/build.sh (rebuild KiCad only)
- ./docker/build.sh --no-clean (incremental)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
