jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// jspi-scheduler.js — the wx scheduler shim for the JSPI runtime.
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
//
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// Ships as a --pre-js. Provides the S4 token-wait registry
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// (beginWait/waitPromise/resolveWait/resolveTopWait/waitEarlyResolved/
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// takeWaitResult/pendingWaits/shutdown) that every C++ bridge and web
|
|
|
|
|
// caller relies on, plus the two things JSPI needs:
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
//
|
|
|
|
|
// 1. ACTIVATION TRACKING. Every promising export the app declares is wrapped
|
|
|
|
|
// so the shim always knows which activation is executing synchronously
|
|
|
|
|
// (an explicit stack; JS is single-threaded so this is exact).
|
|
|
|
|
//
|
|
|
|
|
// 2. SHADOW-STACK DISCIPLINE (emscripten #27364, red/green-proven by
|
|
|
|
|
// tests/apps/standalone/jspi-stack). JSPI switches the native stack per
|
|
|
|
|
// activation but NOT the C spill stack. Every wrapped activation runs on
|
|
|
|
|
// its own pooled spill-stack region with SP swapped at the window
|
|
|
|
|
// boundaries ("green-region", same as libcontext's JSPI backend — KiCad
|
|
|
|
|
// tool coroutines carry their own regions there and do NOT route through
|
|
|
|
|
// here). Green-copy (snapshot/restore of the suspended range) is
|
|
|
|
|
// deliberately NOT used: it rolls back writes other activations make
|
|
|
|
|
// into parked frames' locals (stack-allocated wxDialog members mutated
|
|
|
|
|
// by a cross-tick EndModal), resurrecting dead state at resume.
|
|
|
|
|
//
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// Observability: an event ring + live activation table via __wxWaitDump().
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
|
|
|
|
|
(function () {
|
|
|
|
|
"use strict";
|
|
|
|
|
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
if (globalThis.__wxSchedulerInstalled) {
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
return; // idempotent under double injection
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
var RING_CAP = 256;
|
|
|
|
|
|
|
|
|
|
var S = {
|
|
|
|
|
// --- mailbox lane (timers/wheel; ordering machinery, mechanism-free) ----
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// enqueueAfter queues a C callback; delivery happens through the dedicated
|
|
|
|
|
// _wxWasmMailboxTick export from a fresh task, in order. The tick is a
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// suspension inside a delivered handler parks the tick's own activation,
|
|
|
|
|
// and the rejection path carries the same containment (a throwing handler
|
|
|
|
|
// must not leave a parked quasi-modal unresolved).
|
|
|
|
|
mailbox: [],
|
|
|
|
|
enqueued: 0,
|
|
|
|
|
delivered: 0,
|
|
|
|
|
_tickArmed: false,
|
|
|
|
|
enqueueAfter: function (fn, arg, ms) {
|
|
|
|
|
var self = this;
|
|
|
|
|
setTimeout(function () {
|
|
|
|
|
if (self.dead) return; // never deliver into a torn-down app
|
|
|
|
|
self.mailbox.push({ fn: fn, arg: arg });
|
|
|
|
|
self.enqueued++;
|
|
|
|
|
self._armDeliveryTick();
|
|
|
|
|
}, ms);
|
|
|
|
|
},
|
|
|
|
|
pop: function () {
|
|
|
|
|
var m = this.mailbox.shift();
|
|
|
|
|
if (m) this.delivered++;
|
|
|
|
|
return m || null;
|
|
|
|
|
},
|
|
|
|
|
_tickErrorContainment: function (e) {
|
|
|
|
|
if (Module["_wx_dispatch_abandon"]) Module["_wx_dispatch_abandon"]();
|
|
|
|
|
this.resolveTopWait("nested", 0);
|
|
|
|
|
this.resolveTopWait("modal", 5101); // wxID_CANCEL
|
|
|
|
|
console.warn("[wx-scheduler] mailbox tick error: " + e);
|
|
|
|
|
},
|
|
|
|
|
_armDeliveryTick: function () {
|
|
|
|
|
if (this._tickArmed) return;
|
|
|
|
|
this._tickArmed = true;
|
|
|
|
|
var self = this;
|
|
|
|
|
setTimeout(function tick() {
|
|
|
|
|
if (self.dead) { self._tickArmed = false; return; }
|
|
|
|
|
var p;
|
|
|
|
|
try {
|
|
|
|
|
p = Module["_wxWasmMailboxTick"] ? Module["_wxWasmMailboxTick"]() : undefined;
|
|
|
|
|
} catch (e) {
|
|
|
|
|
self._tickArmed = false;
|
|
|
|
|
self._tickErrorContainment(e);
|
|
|
|
|
throw e;
|
|
|
|
|
}
|
|
|
|
|
Promise.resolve(p).catch(function (e) { self._tickErrorContainment(e); });
|
|
|
|
|
if (self.mailbox.length > 0) {
|
|
|
|
|
setTimeout(tick, 17);
|
|
|
|
|
} else {
|
|
|
|
|
self._tickArmed = false;
|
|
|
|
|
}
|
|
|
|
|
}, 0);
|
|
|
|
|
},
|
|
|
|
|
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// --- S1 embind lane --
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// Mutators (doc 18 classification) must not enter wasm while a load is in
|
|
|
|
|
// flight: the open activation is suspended mid-load and a collab-apply /
|
|
|
|
|
// save / theme flip entering between its parks would mutate the board
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// under it. The exclusion is semantic, independent of the suspension
|
|
|
|
|
// mechanism. The FIFO drains, in order, once kicadOpenFileBusy clears.
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
MUTATOR_NAMES: [
|
|
|
|
|
"kicadSetChrome", "kicadSetReadOnly",
|
|
|
|
|
"kicadCollabApply", "kicadCollabApplyItems",
|
|
|
|
|
"kicadCollabSnapshot", "kicadCollabSnapshotItems",
|
|
|
|
|
"kicadCollabPresenceStart", "kicadCollabSetRemote",
|
2026-08-28 16:14:44 +02:00
|
|
|
"kicadCollabSetRemoteCursors",
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
"kicadCollabSetPins", "kicadCollabSetStyle",
|
|
|
|
|
"kicadCollabSetViewport", "kicadCollabFitViewport",
|
|
|
|
|
"kicadCollabReleaseSelection", "kicadSetColorTheme",
|
|
|
|
|
"kicadSaveBoard", "kicadSaveSchematic", "kicadSaveDrawingSheet",
|
feat: viewer panels — layer selector + selection inspector (viewer-panels)
Canvas-only sessions (read-only viewers, hide-UI editors) get two
floating, draggable, closable panels standing in for the chrome-hidden
wx panes, opened from the overlay menu's View section:
- wasm/bindings: layer bridge — kicadLayersGetState/SetVisible/SetActive
(bodies mirror the compiled-in-but-unreachable IPC handlers, applies
on the coroutine lane, fresh state pushed to
window.kicadCollab.onLayersState); setters join the jspi-scheduler
mutator lane.
- standalone: LayerPanel + SelectionInspector on the comments-panel
shell conventions (useDraggablePanel, collapse, persisted state);
local-selection store fed from presence's onSelection in edit
sessions and bindLocalSelectionFeed for read-only viewers (with a
bounded post-gesture pull burst — clarify-menu selections produce no
canvas event); pure item-summary extraction + unit tests.
- kicad submodule: read-only selection unlock (selection live for
inspection; point editors + RMB context menus stay locked).
- tests/web: read-only spec updated — viewer click selects (or pops the
clarify list), RMB context menu suppressed with writer positive
control, Delete still swallowed; new viewer-panels test (eye toggle
round-trip, active layer, panel drag, inspector rows from a real
canvas click).
Record: docs/features/read-only-viewer/0002-viewer-panels.md (root).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011gJ3M1RpeZLeNUUj8jKC4h
2026-08-20 11:29:46 +02:00
|
|
|
"kicadLayersSetVisible", "kicadLayersSetActive",
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
],
|
|
|
|
|
mutatorQueue: [],
|
|
|
|
|
mutatorsWrapped: 0,
|
|
|
|
|
mutatorsDelivered: 0,
|
|
|
|
|
_mutatorPumpArmed: false,
|
|
|
|
|
_openBusy: function () {
|
|
|
|
|
var probe = Module["kicadOpenFileBusy"];
|
|
|
|
|
if (typeof probe !== "function") return false;
|
|
|
|
|
try { return !!probe(); } catch (e) { return true; }
|
|
|
|
|
},
|
|
|
|
|
_wrapMutators: function () {
|
|
|
|
|
var self = this;
|
|
|
|
|
this.MUTATOR_NAMES.forEach(function (name) {
|
|
|
|
|
var orig = Module[name];
|
|
|
|
|
if (typeof orig !== "function") return;
|
|
|
|
|
self.mutatorsWrapped++;
|
|
|
|
|
Module[name] = function () {
|
|
|
|
|
var args = arguments;
|
|
|
|
|
var call = function () { return orig.apply(Module, args); };
|
|
|
|
|
if (self.mutatorQueue.length === 0 && !self._openBusy()) {
|
|
|
|
|
self.mutatorsDelivered++;
|
|
|
|
|
return call();
|
|
|
|
|
}
|
|
|
|
|
return new Promise(function (resolve, reject) {
|
|
|
|
|
self.mutatorQueue.push({ name: name, call: call, resolve: resolve, reject: reject });
|
|
|
|
|
self._armMutatorPump();
|
|
|
|
|
});
|
|
|
|
|
};
|
|
|
|
|
});
|
|
|
|
|
if (this.mutatorsWrapped > 0)
|
|
|
|
|
this._note("wrapped", "mutators", this.mutatorsWrapped);
|
|
|
|
|
},
|
|
|
|
|
_armMutatorPump: function () {
|
|
|
|
|
if (this._mutatorPumpArmed) return;
|
|
|
|
|
this._mutatorPumpArmed = true;
|
|
|
|
|
var self = this;
|
|
|
|
|
var now = (typeof performance !== "undefined" && performance.now)
|
|
|
|
|
? function () { return performance.now(); }
|
|
|
|
|
: function () { return Date.now(); };
|
|
|
|
|
setTimeout(function pump() {
|
|
|
|
|
if (self.dead) { self._mutatorPumpArmed = false; return; }
|
|
|
|
|
// Unkillable: an exception escaping this body would end the setTimeout
|
|
|
|
|
// chain and wedge the queue forever (observed: 559 frozen messages).
|
|
|
|
|
try {
|
|
|
|
|
if (!self._openBusy()) {
|
|
|
|
|
// Time-boxed drain: ~8 ms of work per 16 ms tick keeps the page
|
|
|
|
|
// live while a long backlog drains in order.
|
|
|
|
|
var t0 = now();
|
|
|
|
|
while (self.mutatorQueue.length > 0 && now() - t0 < 8) {
|
|
|
|
|
if (self._openBusy()) break;
|
|
|
|
|
var m = self.mutatorQueue.shift();
|
|
|
|
|
self.mutatorsDelivered++;
|
|
|
|
|
try { m.resolve(m.call()); } catch (e) { m.reject(e); }
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
} catch (e) {
|
|
|
|
|
self._pumpErrors = (self._pumpErrors || 0) + 1;
|
|
|
|
|
if (self._pumpErrors <= 5)
|
|
|
|
|
console.warn("[wx-scheduler] mutator pump error (occurrence "
|
|
|
|
|
+ self._pumpErrors + "): " + e);
|
|
|
|
|
}
|
|
|
|
|
if (self.mutatorQueue.length > 0) setTimeout(pump, 16);
|
|
|
|
|
else self._mutatorPumpArmed = false;
|
|
|
|
|
}, 16);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// The embind PARKERs (kicadOpenFile / kicadOpenFiles / kicadLibsReload,
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// registered emscripten::async()): wrap them with the
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// same activation tracking as the raw promising exports, so their parks
|
|
|
|
|
// (wxWasmYieldUntil inside the load) find a tracked record and get the
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// green-region spill-stack discipline. Embind names live on Module WITHOUT
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// the underscore prefix, hence the separate installer.
|
2026-08-19 13:47:13 +02:00
|
|
|
PARKER_NAMES: [
|
|
|
|
|
"kicadOpenFile",
|
|
|
|
|
"kicadOpenFiles",
|
|
|
|
|
"kicadLibsReload",
|
|
|
|
|
"kicadLibsAddEntry",
|
|
|
|
|
],
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
_wrapParkers: function () {
|
|
|
|
|
var wrapped = 0;
|
|
|
|
|
for (var i = 0; i < this.PARKER_NAMES.length; i++) {
|
|
|
|
|
var name = this.PARKER_NAMES[i];
|
|
|
|
|
if (typeof Module[name] === "function") {
|
|
|
|
|
Module[name] = this._wrapPromising(name, Module[name], this.PARKER_REGION_BYTES);
|
|
|
|
|
wrapped++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
this._note("wrapped", "parkers", wrapped);
|
|
|
|
|
return wrapped;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// --- S4 wait registry (contract-compatible) ----------------------------
|
|
|
|
|
waits: new Map(), // token -> {kind, promise, resolve, resolved, result, awaited}
|
|
|
|
|
waitSeq: 0,
|
|
|
|
|
waitStacks: {}, // kind -> [unresolved tokens], LIFO
|
|
|
|
|
waitsBegun: 0,
|
|
|
|
|
waitsResolved: 0,
|
|
|
|
|
earlyWaitResolves: 0,
|
|
|
|
|
|
|
|
|
|
beginWait: function (kind) {
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
if (this.dead || this.terminal) {
|
|
|
|
|
// Refuse to mint a wait an unhealthy instance can never satisfy.
|
|
|
|
|
// Callers treat token 0 as "not started" (the C++ bridges bail);
|
|
|
|
|
// a stray waitPromise(0) settles immediately and warns.
|
|
|
|
|
this._note("beginWaitRefused", kind, 0);
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
var token = ++this.waitSeq;
|
|
|
|
|
var entry = { kind: kind, resolved: false, resolve: null, promise: null };
|
|
|
|
|
entry.promise = new Promise(function (resolve) { entry.resolve = resolve; });
|
|
|
|
|
this.waits.set(token, entry);
|
|
|
|
|
(this.waitStacks[kind] = this.waitStacks[kind] || []).push(token);
|
|
|
|
|
this.waitsBegun++;
|
|
|
|
|
this._note("beginWait", kind, token);
|
|
|
|
|
return token;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
waitPromise: function (token) {
|
|
|
|
|
var entry = this.waits.get(token);
|
|
|
|
|
if (!entry) {
|
|
|
|
|
console.warn("[wx-scheduler] waitPromise(" + token + "): unknown token");
|
|
|
|
|
return Promise.resolve(0);
|
|
|
|
|
}
|
|
|
|
|
if (entry.resolved) {
|
|
|
|
|
// resolved before the waiter parked (early-resolve window)
|
|
|
|
|
this.waits.delete(token);
|
|
|
|
|
return Promise.resolve(entry.result | 0);
|
|
|
|
|
}
|
|
|
|
|
entry.awaited = true;
|
|
|
|
|
this._note("park", entry.kind, token);
|
|
|
|
|
return this._suspendOn(entry.promise, entry.kind, token);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
waitEarlyResolved: function (token) {
|
|
|
|
|
var entry = this.waits.get(token);
|
|
|
|
|
return entry && entry.resolved ? 1 : 0;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
takeWaitResult: function (token) {
|
|
|
|
|
var entry = this.waits.get(token);
|
|
|
|
|
if (!entry || !entry.resolved) return 0;
|
|
|
|
|
this.waits.delete(token);
|
|
|
|
|
return entry.result | 0;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
resolveWait: function (token, result) {
|
|
|
|
|
var entry = this.waits.get(token);
|
|
|
|
|
if (!entry || entry.resolved) return false;
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
if (this.terminal) {
|
|
|
|
|
// Resolving would resume the parked frame INSIDE the trapped module
|
|
|
|
|
// (the runWaitCompletion invariant, which the bare finishers used to
|
|
|
|
|
// bypass). Refuse WITHOUT consuming the entry — the frame stays
|
|
|
|
|
// visibly parked in dump() and the ring says why.
|
|
|
|
|
this._note("resolveRefused", entry.kind, token);
|
|
|
|
|
console.warn("[wx-scheduler] resolveWait(" + token + ", " + entry.kind
|
|
|
|
|
+ ") refused: instance is terminal");
|
|
|
|
|
return false;
|
|
|
|
|
}
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
entry.resolved = true;
|
|
|
|
|
this.waitsResolved++;
|
|
|
|
|
var stack = this.waitStacks[entry.kind];
|
|
|
|
|
if (stack) {
|
|
|
|
|
var idx = stack.indexOf(token);
|
|
|
|
|
if (idx !== -1) stack.splice(idx, 1);
|
|
|
|
|
}
|
|
|
|
|
entry.result = result | 0;
|
|
|
|
|
entry.resolve(result | 0);
|
|
|
|
|
this._note("resolve", entry.kind, token);
|
|
|
|
|
if (entry.awaited) {
|
|
|
|
|
this.waits.delete(token);
|
|
|
|
|
} else {
|
|
|
|
|
// early resolve: keep the entry, result attached, for the late waiter
|
|
|
|
|
this.earlyWaitResolves++;
|
|
|
|
|
}
|
|
|
|
|
return true;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
resolveTopWait: function (kind, result) {
|
|
|
|
|
var stack = this.waitStacks[kind];
|
|
|
|
|
if (!stack || stack.length === 0) return false;
|
|
|
|
|
return this.resolveWait(stack[stack.length - 1], result);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
pendingWaits: function (kind) {
|
|
|
|
|
var stack = this.waitStacks[kind];
|
|
|
|
|
return stack ? stack.length : 0;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
dead: false,
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
// --- E-8: admission gate for delayed worker/MEMFS completions -----------
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
// `terminal` means the wasm instance TRAPPED (WebAssembly.RuntimeError,
|
|
|
|
|
// or emscripten's abort — which throws a RuntimeError itself and is also
|
|
|
|
|
// latched authoritatively via Module.onAbort → terminalize): the heap may
|
|
|
|
|
// be mid-mutation, so no further native work (malloc / heap stores / FS
|
|
|
|
|
// writes) may run and no parked frame may be resumed into it. Distinct
|
|
|
|
|
// from `dead` (orderly shutdown). One-way.
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
terminal: false,
|
|
|
|
|
canTouchNative: function () { return !this.dead && !this.terminal; },
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
// Public one-way latch (also wired from boot's Module.onAbort — the
|
|
|
|
|
// authoritative abort notification).
|
|
|
|
|
terminalize: function (site, e) {
|
|
|
|
|
if (this.terminal) return;
|
|
|
|
|
this.terminal = true;
|
|
|
|
|
this._note("terminal", site, 0);
|
|
|
|
|
console.error("[wx-scheduler] instance is terminal (" + site
|
|
|
|
|
+ ") — all further native completions are inert: " + (e || ""));
|
|
|
|
|
},
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
_terminalizeNativeTrap: function (site, e) {
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
// Structural signals only: a genuine engine trap in this same-realm
|
|
|
|
|
// prepare/entry IS a WebAssembly.RuntimeError instance; the duck-typed
|
|
|
|
|
// name fallback survives realm loss on a relayed error object. The old
|
|
|
|
|
// message-substring sniff ('Aborted(', 'index out of bounds', …) only
|
|
|
|
|
// added false positives — any plain JS error QUOTING such text bricked
|
|
|
|
|
// a healthy instance permanently.
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
var isTrap = (typeof WebAssembly !== "undefined"
|
|
|
|
|
&& WebAssembly.RuntimeError
|
|
|
|
|
&& e instanceof WebAssembly.RuntimeError)
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
|| !!(e && e.name === "RuntimeError");
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
if (!isTrap) return false;
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
this.terminalize(site, e);
|
findings(E-5,E-8,E-9): module-identity ngspice events + runWaitCompletion admission gate
E-8 (re-implemented for JSPI — the codex gate is entangled with the dropped
execution owner; under JSPI a fresh non-suspending JS→wasm entry while
another activation is suspended is structurally safe on its own stack, so
the admission boundary for worker completions is liveness + trap state, not
execution ownership):
- jspi-scheduler.js grows `terminal` (trapped instance; distinct from `dead`),
canTouchNative(), _terminalizeNativeTrap() (WebAssembly.RuntimeError +
cross-realm string classification), and runWaitCompletion(site, token,
prepare, inertResult): prepare runs immediately and owns ALL native work;
stale tokens and dead/terminal instances drop loudly without resolving
(resolving would resume the parked frame inside the damaged module); a
trap latches terminal; a plain JS bug resolves inertResult so the wait
fails instead of stranding. beginWait refuses (token 0) when dead/terminal.
- all four delayed completion sites route their native work through the
gate: 'OCC export completion' (exporter_step_stub), 'OCC model completion'
(oce_plugin_stub — the MEMFS cache write moves inside the gate too),
'ngspice request completion' and 'ngspice vector completion'
(sharedspice_client — every HEAP32/HEAPF64/malloc write inside prepare,
inertResult 1 = transport error). Every wxWasmBeginWait caller in the
stubs bails on token <= 0.
- deliberately NOT ported from codex: ownerModule, enqueueNativeCompletion,
executionBarrier, the byte-credit native-entry FIFO — completions are
one-shot per wait token and stream volume is bounded at the E-6 transport
credit window. Cross-refs logged for group M (M-2/M-6/M-8).
E-5 (re-implemented; codex shape kept, owner APIs replaced with the E-8
gate): js_ngspice_install_events binds the handler to the EXACT installing
module (handler.__pcbjamNgspiceOwnerModule stamp; presence is not identity),
re-installation replaces a foreign module's handler, a superseded handler
disarms itself, native entry goes through installingModule._malloc/
._pcbjam_ngspice_event (never lexical Module), each dispatch checks
canTouchNative() (loud drop on a dead/terminal module), and a trap on the
per-line entry latches the terminal gate.
Tests: scheduler-shim.test.ts +7 (gate happy/stale/dead/terminal/cross-realm/
js-bug/beginWait-refusal). e2e specs updated from the codex line: occ-export
decode-fault recovery (real onmessageerror transition via failDecode, J-4),
ngspice-probe direct-service coverage, eeschema-sim rewritten onto the E-7
applied-generation receipt (codex's executionBarrier await replaced with a
pendingWaits('ngspice') drain poll — the JSPI-line equivalent).
Also bumps the kicad submodule to the E-7/E-9 commit (dd5751038f7).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-19 16:03:16 +02:00
|
|
|
return true;
|
|
|
|
|
},
|
|
|
|
|
// The one admission boundary for delayed completions that both touch
|
|
|
|
|
// native state and wake a parked waiter (the four worker/MEMFS completion
|
|
|
|
|
// sites: OCC export, OCC model, ngspice request, ngspice vector).
|
|
|
|
|
// `prepare` runs IMMEDIATELY, never queued — it owns the parked waiter's
|
|
|
|
|
// output pointers, and queuing it behind anything can deadlock the very
|
|
|
|
|
// frame this completion wakes. Disposition (every drop is loud, never
|
|
|
|
|
// silent):
|
|
|
|
|
// stale/unknown token -> drop + warn (late frame from a retired
|
|
|
|
|
// worker generation)
|
|
|
|
|
// dead or terminal instance -> drop + warn, DO NOT resolve — resolving
|
|
|
|
|
// resumes the suspended frame INSIDE the
|
|
|
|
|
// damaged module
|
|
|
|
|
// prepare() traps -> latch terminal, DO NOT resolve
|
|
|
|
|
// prepare() throws plain JS -> resolve inertResult (fail the wait
|
|
|
|
|
// rather than strand its parked frame in
|
|
|
|
|
// a healthy instance)
|
|
|
|
|
runWaitCompletion: function (site, token, prepare, inertResult) {
|
|
|
|
|
var entry = this.waits.get(token);
|
|
|
|
|
if (!entry || entry.resolved) {
|
|
|
|
|
console.warn("[wx-scheduler] " + site + ": completion for stale wait "
|
|
|
|
|
+ token + " dropped");
|
|
|
|
|
this._note("staleCompletion", site, token);
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
if (!this.canTouchNative()) {
|
|
|
|
|
console.warn("[wx-scheduler] " + site + ": completion dropped ("
|
|
|
|
|
+ (this.terminal ? "terminal" : "dead") + " instance)");
|
|
|
|
|
this._note("inertCompletion", site, token);
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
var result;
|
|
|
|
|
try {
|
|
|
|
|
result = prepare();
|
|
|
|
|
} catch (e) {
|
|
|
|
|
if (this._terminalizeNativeTrap(site, e)) {
|
|
|
|
|
this._note("completionTrap", site, token);
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
console.error("[wx-scheduler] " + site + ": completion failed: " + e);
|
|
|
|
|
this._note("completionError", site, token);
|
|
|
|
|
this.resolveWait(token, inertResult == null ? 0 : inertResult | 0);
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
this.resolveWait(token, result | 0);
|
|
|
|
|
return true;
|
|
|
|
|
},
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
shutdown: function (why) {
|
|
|
|
|
this.dead = true;
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// S6 teardown contract: queued-but-
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// undelivered mutators FAIL LOUDLY instead of hanging their callers,
|
|
|
|
|
// and undelivered mailbox messages drop — the pumps stop themselves on
|
|
|
|
|
// the dead flag.
|
|
|
|
|
var q = this.mutatorQueue.splice(0, this.mutatorQueue.length);
|
|
|
|
|
for (var i = 0; i < q.length; i++) {
|
|
|
|
|
try { q[i].reject(new Error("wx scheduler shutdown: " + why)); } catch (e) { /* reject never throws */ }
|
|
|
|
|
}
|
|
|
|
|
this.mailbox.length = 0;
|
|
|
|
|
var stranded = this.waits.size;
|
|
|
|
|
if (stranded) {
|
|
|
|
|
console.warn("[wx-scheduler] shutdown (" + why + ") stranded:" + stranded);
|
jspi: fix the dead-tools ownership bug, emscripten-6 fallout, and green the full suite on Playwright 1.62
Live-app fix (Place Footprints / routing dead in Chrome): submodule
bumps carry the coroutine ownership fix (kicad 012d95ecb4) and the
handler-exception survival fix (wxwidgets 1b5f0e31f4).
Emscripten-6 fallout:
- occ/ngspice worker wrappers: mainScriptUrlOrBlob was removed
upstream; pthread children re-run the wrapper blob, so an em-pthread
realm now importScripts the glue and gets out of the way (before:
recursive service boots, pool never fills, silent 180s boot hangs —
every occ spec and ngspice bg_run).
- Makefile.wasm: -sASYNCIFY frankenlinks on the no-wx coroutine repro
targets ported to -sJSPI (the JSPI-only libcontext crashed at first
yield under them); mainloop/gl repro pages drive their tick through a
promising export (emscripten_set_main_loop callbacks cannot suspend);
retired inject-dyncall-shims lines removed (targets were unbuildable
since Phase 8); $stringToNewUTF8 force-included (the EM_ASM value
bridge aborted the runtime on the first decoded exception).
- fiber-park levers: neither embind shape can drive suspending levers
(plain throws on strict-JSPI Firefox; emscripten::async() re-executes
its invoker on settle) — kept sync for manual Chromium probing, spec
coverage moved to the jspi-coroutine harness (18 cases).
Suite work:
- Playwright 1.61.1 -> 1.62.1 (Firefox 153: JSPI on by default).
- fiber-resume-park.spec retired -> coroutine-lifecycle.spec: census
gate over boot / board load / chooser open / cancel (deterministically
red on the pre-fix build).
- Blind asyncify-era pins re-keyed: quasimodal-strand + wait-beacons
beacon regexes, footprint-chooser-close liveness -> wx parking-timer
heartbeat (scheduler counters idle flat on Firefox).
- occ/ngspice test providers: 60s boot timeout + worker error
surfacing (a worker death used to be a silent 180s timeout).
- Harness pages: stale 9.99 config dir -> 10.0 (library_manager wxCHECK
noise, chooser had no libraries).
- gal-webgl harness: missing artifacts rebuilt (boost/glm extracted to
the host sysroot), PgmOrNull stub added for the rebased GAL.
- jspi-scheduler: clean-shutdown console line restored (app-quit
contract), quarantine never yanks SP from a live window.
Gates: test:e2e 699 passed / 0 failed (wx-chromium, kicad-firefox,
kicad-chromium, jspi-firefox, coroutine-firefox); web ff/cr/mobile 71
passed; lint:ci-coverage 166, lint:determinism 163, screenshots
manifest 492 current, corpus 7/7, tools:contract green. Offline
screenshot baselines show expected mass drift from the engine bump —
re-baseline (screenshots:noise -> promote) is a follow-up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-13 17:41:28 +02:00
|
|
|
} else {
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// teardown-gate contract (e2e/app-quit.spec.ts): a clean exit must
|
|
|
|
|
// SAY so on the console
|
jspi: fix the dead-tools ownership bug, emscripten-6 fallout, and green the full suite on Playwright 1.62
Live-app fix (Place Footprints / routing dead in Chrome): submodule
bumps carry the coroutine ownership fix (kicad 012d95ecb4) and the
handler-exception survival fix (wxwidgets 1b5f0e31f4).
Emscripten-6 fallout:
- occ/ngspice worker wrappers: mainScriptUrlOrBlob was removed
upstream; pthread children re-run the wrapper blob, so an em-pthread
realm now importScripts the glue and gets out of the way (before:
recursive service boots, pool never fills, silent 180s boot hangs —
every occ spec and ngspice bg_run).
- Makefile.wasm: -sASYNCIFY frankenlinks on the no-wx coroutine repro
targets ported to -sJSPI (the JSPI-only libcontext crashed at first
yield under them); mainloop/gl repro pages drive their tick through a
promising export (emscripten_set_main_loop callbacks cannot suspend);
retired inject-dyncall-shims lines removed (targets were unbuildable
since Phase 8); $stringToNewUTF8 force-included (the EM_ASM value
bridge aborted the runtime on the first decoded exception).
- fiber-park levers: neither embind shape can drive suspending levers
(plain throws on strict-JSPI Firefox; emscripten::async() re-executes
its invoker on settle) — kept sync for manual Chromium probing, spec
coverage moved to the jspi-coroutine harness (18 cases).
Suite work:
- Playwright 1.61.1 -> 1.62.1 (Firefox 153: JSPI on by default).
- fiber-resume-park.spec retired -> coroutine-lifecycle.spec: census
gate over boot / board load / chooser open / cancel (deterministically
red on the pre-fix build).
- Blind asyncify-era pins re-keyed: quasimodal-strand + wait-beacons
beacon regexes, footprint-chooser-close liveness -> wx parking-timer
heartbeat (scheduler counters idle flat on Firefox).
- occ/ngspice test providers: 60s boot timeout + worker error
surfacing (a worker death used to be a silent 180s timeout).
- Harness pages: stale 9.99 config dir -> 10.0 (library_manager wxCHECK
noise, chooser had no libraries).
- gal-webgl harness: missing artifacts rebuilt (boost/glm extracted to
the host sysroot), PgmOrNull stub added for the rebased GAL.
- jspi-scheduler: clean-shutdown console line restored (app-quit
contract), quarantine never yanks SP from a live window.
Gates: test:e2e 699 passed / 0 failed (wx-chromium, kicad-firefox,
kicad-chromium, jspi-firefox, coroutine-firefox); web ff/cr/mobile 71
passed; lint:ci-coverage 166, lint:determinism 163, screenshots
manifest 492 current, corpus 7/7, tools:contract green. Offline
screenshot baselines show expected mass drift from the engine bump —
re-baseline (screenshots:noise -> promote) is a follow-up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-13 17:41:28 +02:00
|
|
|
console.log("[wx-scheduler] shutdown (" + why + ") clean");
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
}
|
|
|
|
|
this._note("shutdown", why, stranded);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// --- activation tracking + shadow-stack discipline ---------------------
|
|
|
|
|
//
|
|
|
|
|
// Window model (JS is single-threaded, so this is exact):
|
|
|
|
|
// * FIRST window — a promising export runs synchronously from its JS
|
|
|
|
|
// caller until first suspend or completion. It is a real JS call
|
|
|
|
|
// frame, so _actStack (push in the wrap, pop in its finally) mirrors
|
|
|
|
|
// the JS stack exactly, including exports entered synchronously from
|
|
|
|
|
// inside a resumed window.
|
|
|
|
|
// * RESUMED window — the engine re-enters a suspended activation from a
|
|
|
|
|
// promise reaction. There is NO JS frame of ours around it, so it is
|
|
|
|
|
// tracked by _windowLive instead.
|
|
|
|
|
// The wasm code executing at any suspension therefore belongs to
|
|
|
|
|
// _actStack's top when non-empty, else to _windowLive.
|
|
|
|
|
//
|
|
|
|
|
// Discipline: GREEN-REGION (per-activation spill-stack region + SP swap
|
|
|
|
|
// at the window boundaries), the same leg of the jspi-stack red/green
|
|
|
|
|
// bake-off the libcontext JSPI backend uses. NOT green-copy: a snapshot/
|
|
|
|
|
// restore of the suspended range rolls back writes that OTHER activations
|
|
|
|
|
// legitimately made into the parked frames' locals — a stack-allocated
|
|
|
|
|
// wxDialog whose EndModal (from another tick's window) cleared
|
|
|
|
|
// m_isShowingModal would have the flag restored to true at resume, and
|
|
|
|
|
// its destructor then fires EndModal(wxID_CANCEL) into some OUTER modal's
|
|
|
|
|
// wait (observed as the triple-modal LIFO failure). With a region per
|
|
|
|
|
// activation nothing else ever executes on a parked activation's stack,
|
|
|
|
|
// so cross-activation writes persist and nothing needs copying.
|
|
|
|
|
//
|
|
|
|
|
// Resume TURNSTILE: SP swaps must happen (a) only at microtask
|
|
|
|
|
// boundaries (never while wasm frames are live on the JS stack) and
|
|
|
|
|
// (b) for at most ONE activation between wasm re-entries — between our
|
|
|
|
|
// swap and the engine's actual re-entry other microtasks still run, and
|
|
|
|
|
// a second swap would redirect it. So ready resumes queue in
|
|
|
|
|
// _resumeReady and _pumpResume (microtask-scheduled only) arms exactly
|
|
|
|
|
// one and resolves its gate; the engine's re-entry is the only reaction
|
|
|
|
|
// on that gate. The next pump happens when that window ENDS — its next
|
|
|
|
|
// suspension or its completion — both of which we observe.
|
|
|
|
|
_actSeq: 0,
|
|
|
|
|
_actStack: [], // records of FIRST windows currently on the JS stack
|
|
|
|
|
_suspended: new Map(),// actId -> record, while suspended (dump/watchdog)
|
|
|
|
|
_windowLive: null, // record whose RESUMED window is executing (or armed)
|
|
|
|
|
_resumeReady: [], // FIFO of {rec, gate} whose wait promise resolved
|
|
|
|
|
|
|
|
|
|
_sp: function () { return Module["stackSave"](); },
|
|
|
|
|
_setSp: function (v) { Module["stackRestore"](v); },
|
|
|
|
|
|
|
|
|
|
_top: function () {
|
|
|
|
|
return this._actStack.length
|
|
|
|
|
? this._actStack[this._actStack.length - 1]
|
|
|
|
|
: null;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// Per-activation spill-stack regions, pooled (malloc'd from the wasm
|
|
|
|
|
// heap; wx dispatch chains are shallow compared to tool coroutines —
|
|
|
|
|
// the deep KiCad tool bodies run on libcontext's own 256K regions).
|
|
|
|
|
// PARKER_REGION_BYTES for the embind load chains (kicadOpenFile parses
|
|
|
|
|
// whole boards on this stack).
|
|
|
|
|
// KiCad dispatch chains and board loads run DEEP (a full board parse
|
|
|
|
|
// happens on the parker's region; a paint dispatch can recurse through
|
|
|
|
|
// tool handlers) — an overflowing region scribbles the heap below it and
|
|
|
|
|
// kills the renderer. Regions are pooled, so generous sizes cost little.
|
|
|
|
|
REGION_BYTES: 1024 * 1024,
|
|
|
|
|
PARKER_REGION_BYTES: 8 * 1024 * 1024,
|
|
|
|
|
_regionPool: {}, // size -> [regions]
|
|
|
|
|
// This file is a --pre-js, so the glue's bare _malloc/_free are in scope
|
|
|
|
|
// at call time; Module["_malloc"] is the fallback for glue shapes that
|
|
|
|
|
// attach them there instead.
|
|
|
|
|
_mallocFn: function () {
|
|
|
|
|
return (typeof _malloc === "function") ? _malloc : Module["_malloc"];
|
|
|
|
|
},
|
|
|
|
|
_freeFn: function () {
|
|
|
|
|
return (typeof _free === "function") ? _free : Module["_free"];
|
|
|
|
|
},
|
|
|
|
|
_regionAlloc: function (size) {
|
|
|
|
|
var pool = (this._regionPool[size] = this._regionPool[size] || []);
|
|
|
|
|
var r = pool.pop();
|
|
|
|
|
if (r) return r;
|
|
|
|
|
var base = this._mallocFn()(size);
|
|
|
|
|
if (!base) throw new Error("[wx-scheduler] region alloc failed (" + size + ")");
|
|
|
|
|
// A JS-initiated _malloc can GROW wasm memory, and glue code holding
|
|
|
|
|
// pre-growth views then writes into a detached buffer (observed: the
|
|
|
|
|
// fd_write out-param never landing, musl's __stdio_write retrying a
|
|
|
|
|
// 0-byte writev forever). Refresh the glue's views immediately; the
|
|
|
|
|
// install-time preallocation below makes this path rare to begin with.
|
|
|
|
|
try {
|
|
|
|
|
if (typeof updateMemoryViews === "function" && typeof wasmMemory !== "undefined"
|
|
|
|
|
&& typeof HEAPU8 !== "undefined" && HEAPU8.buffer !== wasmMemory.buffer) {
|
|
|
|
|
updateMemoryViews();
|
|
|
|
|
}
|
|
|
|
|
} catch (e) { /* non-glue host (unit tests) */ }
|
|
|
|
|
// The wasm C stack REQUIRES 16-byte alignment; wasm32 malloc only
|
|
|
|
|
// guarantees 8. A region top at base+size can be 8 (mod 16), and a
|
|
|
|
|
// misaligned SP skews every alignment-derived address in the
|
|
|
|
|
// activation by 8 (observed: EM_ASM's readEmAsmArgs assert, and musl
|
|
|
|
|
// __stdio_write passing an iov pointer 8 below the array it populated
|
|
|
|
|
// -> an infinite 0-byte writev retry loop wedging the main thread).
|
|
|
|
|
// Align the top DOWN; the lost <16 bytes are spare.
|
|
|
|
|
return { base: base, top: (base + size) & ~15, size: size };
|
|
|
|
|
},
|
|
|
|
|
// Fill the pools while nothing is suspended (runtime init): any memory
|
|
|
|
|
// growth this causes happens at a safe boundary instead of mid-window.
|
|
|
|
|
_preallocRegions: function (haveParkers) {
|
|
|
|
|
// Host without a wasm heap (the vitest fake runtime): nothing to fill.
|
|
|
|
|
if (typeof this._mallocFn() !== "function") return;
|
|
|
|
|
var i, rs = [];
|
|
|
|
|
for (i = 0; i < 8; i++) rs.push(this._regionAlloc(this.REGION_BYTES));
|
|
|
|
|
if (haveParkers) for (i = 0; i < 2; i++) rs.push(this._regionAlloc(this.PARKER_REGION_BYTES));
|
|
|
|
|
for (i = 0; i < rs.length; i++) this._regionFree(rs[i]);
|
|
|
|
|
},
|
|
|
|
|
_regionFree: function (r) {
|
|
|
|
|
var pool = (this._regionPool[r.size] = this._regionPool[r.size] || []);
|
|
|
|
|
if (pool.length < 8) pool.push(r);
|
|
|
|
|
else this._freeFn()(r.base);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// Wrap one promising export so the shim tracks its windows and gives the
|
|
|
|
|
// activation its own spill region. The returned promise
|
|
|
|
|
// (WebAssembly.promising exports and embind async() invokers always
|
|
|
|
|
// return one) settles when the ACTIVATION completes — that frees the
|
|
|
|
|
// region and un-parks the turnstile.
|
|
|
|
|
_wrapPromising: function (name, fn, regionBytes) {
|
|
|
|
|
var S = this;
|
|
|
|
|
var bytes = regionBytes || S.REGION_BYTES;
|
|
|
|
|
return function () {
|
|
|
|
|
var enclosingSp = S._sp();
|
|
|
|
|
var region = S._regionAlloc(bytes);
|
|
|
|
|
var rec = {
|
|
|
|
|
id: ++S._actSeq,
|
|
|
|
|
kind: name,
|
|
|
|
|
region: region,
|
|
|
|
|
entrySp: region.top,
|
|
|
|
|
suspendedAt: 0
|
|
|
|
|
};
|
|
|
|
|
S._actStack.push(rec);
|
|
|
|
|
S._setSp(region.top);
|
|
|
|
|
var out;
|
|
|
|
|
try {
|
|
|
|
|
out = fn.apply(this, arguments);
|
|
|
|
|
} finally {
|
|
|
|
|
var popped = S._actStack.pop();
|
|
|
|
|
if (popped !== rec) {
|
|
|
|
|
console.warn("[wx-scheduler] activation stack imbalance at " + name);
|
|
|
|
|
}
|
|
|
|
|
// First window over (completed, suspended, or threw): the caller
|
|
|
|
|
// continues on the enclosing stack either way.
|
|
|
|
|
S._setSp(enclosingSp);
|
|
|
|
|
if (!(out && typeof out.then === "function")) {
|
|
|
|
|
// sync throw or non-promise return: the activation is over now
|
|
|
|
|
S._endActivation(rec);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (out && typeof out.then === "function") {
|
|
|
|
|
out.then(
|
|
|
|
|
function () { S._endActivation(rec); },
|
|
|
|
|
function () { S._endActivation(rec); }
|
|
|
|
|
);
|
|
|
|
|
}
|
|
|
|
|
return out;
|
|
|
|
|
};
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
_endActivation: function (rec) {
|
|
|
|
|
this._suspended.delete(rec.id);
|
|
|
|
|
if (this._windowLive === rec) {
|
|
|
|
|
// completed from a RESUMED window: wasm's epilogue left SP at the
|
|
|
|
|
// region top — put the enclosing stack back before anything else
|
|
|
|
|
// enters wasm.
|
|
|
|
|
this._windowLive = null;
|
|
|
|
|
if (rec.enclosingSp !== undefined) this._setSp(rec.enclosingSp);
|
|
|
|
|
}
|
|
|
|
|
if (rec.region) {
|
|
|
|
|
this._regionFree(rec.region);
|
|
|
|
|
rec.region = null;
|
|
|
|
|
}
|
|
|
|
|
var S = this;
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
_pumpResume: function () {
|
findings(E-10..E-22): fix the defects a code review found in the E-1..E-9 work
A review of the group-E fixes found 13 further defects; ten were introduced by
those fixes, two pre-existed and were merely relocated, one is deferred.
Services / transport
E-10 retireWorker synthesized no bg/exit frame, so sharedspice's s_bgRunning
mirror stayed latched true after a mid-run worker death: Run stayed
disabled and the promised fresh-worker restart was unreachable for the
whole session. Retirement now dispatches a synthetic controlled-exit
straight to the installed handler (never through dispatchEvt — a
fabricated frame must not touch the credit ledger). Driving the repro
exposed two further defects, both fixed here: a replacement worker
trapped on pre-init engine reads, and the rerun's cm_input_path/circ hit
that uninitialized engine before KiCad's validate() re-init (the native
flow assumes a crashed engine survives in-process — true for the dll,
false for a dead worker). Reads now answer their empty shapes pre-init,
writes lazy-init, and init is idempotent per worker engine.
E-19 dispatchEvt acked only AFTER handler(evt) returned, and the sharedspice
client deliberately rethrows non-trap errors — so each throw leaked one
unit of the 64-frame credit window until the stream died with a
misattributed "transport exceeded". The ack moves to a finally in both
service copies; the throw still propagates (the trap machinery needs it).
E-20 the oversize-line path promises to transfer the accepted prefix, but
with the window full that flush only DEFERS, and stopEventStream wiped
the deferred queue — losing the diagnostics that explain the failure.
The terminal notice now carries them as pendingEvents; both hosts
deliver them in order, unacked (the fatal frame is outside the credit
protocol).
E-21 the 30s prefetch deadline discarded every model already collected and
reported nothing. A caller-owned progress sink ships the partials and
the omission reaches the export report. (Awaiting the aborted collection
was rejected: an in-flight source fetch is not abortable — E-4's
original disease.) Plus a serving-candidate memo, so a .wrl ref served
by its .step fallback stops re-probing the miss on every export.
Scheduler
E-14 _terminalizeNativeTrap classified by message substring, so any plain JS
error QUOTING 'Aborted(' or 'out of bounds' permanently bricked a
healthy instance. Now structural only: instanceof RuntimeError plus a
duck-typed name check (verified in this build's glue that abort() throws
a genuine RuntimeError both pre- and post-runtime-init). Module.onAbort
now latches the gate — the authoritative notification, previously
ignored.
E-15 the shim half: _pumpResume gates on terminal (catching wakes already
queued at latch time) and resolveWait refuses on terminal WITHOUT
consuming the entry, so a frame stays visibly parked rather than
resuming inside a trapped module.
E-16 the E-5 handler read the realm-global scheduler at dispatch instead of
its installing module's; also frees the per-line buffer on the non-trap
rethrow path.
E-11 get_vec trusted the worker's res.length over the transferred arrays.
Observed death shape: a 4 GiB std::vector threw an unhandled
std::length_error that exited the editor's main loop. Now clamped, with
the buffers freed on every failure path.
Guardrails (replacing two deferred refactors: e2e→production-code injection and
collapsing the four copies of the worker-lifecycle machinery)
E-18 the source contract asserted comment-string counts — rewording failed
CI while moving a guard outside its #ifdef passed. It now parses the
#ifdef regions and asserts on code.
service-stub-parity.ts pins what the four lifecycle copies must share:
credit-window equality parsed from source, the finally-ack, boot
deadlines, terminal-notice consumption. The transport numbers are now
single-sourced from the worker.
CI actually runs the gates: the web/standalone vitest suites (which had
NEVER run in CI), the reducer, the source contract and the parity tool —
with a NON_PLAYWRIGHT_GATES check so deleting a step re-fails the lint.
E-22 the e2e occ stub's 60s boot watchdog, deleted in a66e109, is restored in
the ngspice-stub shape with a wedgeNextBoot() repro hook.
Every behavioral fix has red-then-green evidence (the reds were captured first).
E-17 (a stale RUNNING cross-stamping the next run's generation under E-6's
transport deferral) is DEFERRED with its analysis recorded — a real fix needs
run identity on the bg frames.
Test hygiene: the dwell lint now requires the mandated ": <why>" and all 47 bare
markers carry their reason; three export-report dwells became modal-lease polls;
exact-ledger assertions became relative deltas; the dead data-wx-dom-id branch,
an unused fault hook and unused receipt plumbing are gone; abort scans, wx
dialog drivers, the sim harness and the vitest FakeWorker are each one copy now.
Bumps kicad and wxwidgets to their findings-group-e tips.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-27 12:27:12 +02:00
|
|
|
// `terminal` too: a queued wake must never re-enter a trapped module —
|
|
|
|
|
// resuming swaps SP into (and runs wasm on) a heap that may be
|
|
|
|
|
// mid-mutation. Freezing the pump on a terminal instance is by design:
|
|
|
|
|
// the fatal overlay owns the page from here.
|
|
|
|
|
if (this.dead || this.terminal) return;
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
if (this._windowLive) {
|
|
|
|
|
// Self-heal: an activation that suspended RAW (bypassing the shim)
|
|
|
|
|
// or completed untracked never ends its window here; without this
|
|
|
|
|
// the pump would refuse resumes forever. Anything armed >2s while
|
|
|
|
|
// resumes queue is such a leak — clear it loudly. (A window whose
|
|
|
|
|
// wasm is genuinely executing can't be observed here at all: the
|
|
|
|
|
// pump only runs between JS jobs.)
|
|
|
|
|
var w = this._windowLive;
|
|
|
|
|
if (w.windowArmedAt && Date.now() - w.windowArmedAt > 2000
|
|
|
|
|
&& this._resumeReady.length) {
|
|
|
|
|
console.warn("[wx-scheduler] force-clearing stuck window act "
|
|
|
|
|
+ w.id + ":" + w.kind + " — some suspension bypassed the shim");
|
|
|
|
|
this._note("forceClearWindow", w.kind, w.id);
|
|
|
|
|
this._windowLive = null;
|
|
|
|
|
} else {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (this._resumeReady.length === 0) return;
|
|
|
|
|
var e = this._resumeReady.shift();
|
|
|
|
|
var rec = e.rec;
|
|
|
|
|
if (rec.dead) {
|
|
|
|
|
// Quarantined (coroutine released while parked): the body's C++ is
|
|
|
|
|
// freed — a late wake must never re-enter it. Drop the wake; the
|
|
|
|
|
// gate promise parks the leaked activation forever (censused C-side).
|
|
|
|
|
console.warn("[wx-scheduler] dropping wake for quarantined " + rec.id);
|
|
|
|
|
this._note("deadWakeDropped", rec.kind, rec.id);
|
|
|
|
|
var S = this;
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
this._suspended.delete(rec.id);
|
|
|
|
|
rec.suspendedAt = 0;
|
|
|
|
|
// Arm the window: remember the enclosing stack (to put back when this
|
|
|
|
|
// window ends) and point SP back into the activation's own region,
|
|
|
|
|
// exactly where it suspended.
|
|
|
|
|
rec.enclosingSp = this._sp();
|
|
|
|
|
rec.windowArmedAt = Date.now();
|
|
|
|
|
// Truthful attribution for the resumed slice's NEXT suspension: after
|
|
|
|
|
// a foreign wake a coroutine re-enters through plain C with nothing on
|
|
|
|
|
// the path to re-arm g_current (own-yield resumes re-set it C-side —
|
|
|
|
|
// idempotent with this). Before the SP swap: make_current has real
|
|
|
|
|
// frames (registry lookup) and must run on the enclosing stack.
|
|
|
|
|
this._libctxMakeCurrent(rec.lcid || 0);
|
|
|
|
|
this._setSp(rec.sp);
|
|
|
|
|
this._windowLive = rec;
|
|
|
|
|
// The engine's re-entry is the only reaction on the gate; microtasks
|
|
|
|
|
// queued before it can only enqueue further resumes (no SP swaps —
|
|
|
|
|
// _windowLive is set), so SP survives untouched until wasm runs.
|
|
|
|
|
if (e.rejected) e.gate.reject(e.value);
|
|
|
|
|
else e.gate.resolve(e.value);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// Untracked activations (main — the runtime calls it before the wraps
|
|
|
|
|
// exist — and libctx coroutine bodies doing FOREIGN yields like
|
|
|
|
|
// sleepYield) get a FRESH anonymous record per suspension. Not a shared
|
|
|
|
|
// singleton: main is eternally parked on its frame yield, and a
|
|
|
|
|
// singleton would let a coroutine's suspension overwrite main's saved
|
|
|
|
|
// SP (cross-wired resumes). An uncontended activation like main still
|
|
|
|
|
// keeps a stable identity naturally: its next suspension is attributed
|
|
|
|
|
// to its own live window (rec === _windowLive) and reuses the record.
|
|
|
|
|
_anonSeq: 0,
|
|
|
|
|
|
|
|
|
|
// Route a suspension through the turnstile. No byte copying: the
|
|
|
|
|
// activation's frames live in its own region (main: the central stack)
|
|
|
|
|
// and stay valid while parked; only the shared SP global is handed back
|
|
|
|
|
// and forth.
|
|
|
|
|
_suspendOn: function (p, kind, token) {
|
|
|
|
|
var S = this;
|
|
|
|
|
// A FOREIGN yield from inside a KiCad coroutine body belongs to the
|
|
|
|
|
// COROUTINE's activation — the wx entry that entered it is still on
|
|
|
|
|
// the JS stack and must not be double-booked (its own libctx-enter
|
|
|
|
|
// suspension already owns that record).
|
|
|
|
|
var lcid = 0;
|
|
|
|
|
try {
|
|
|
|
|
lcid = (typeof _pcbjam_libctx_current === "function") ? _pcbjam_libctx_current()
|
|
|
|
|
: (Module["_pcbjam_libctx_current"] ? Module["_pcbjam_libctx_current"]() : 0);
|
|
|
|
|
} catch (e) { /* pre-runtime */ }
|
|
|
|
|
var rec;
|
|
|
|
|
if (lcid) {
|
|
|
|
|
rec = S._libctxRecs[lcid] || (S._libctxRecs[lcid] = {
|
|
|
|
|
id: "lc" + lcid, kind: "libctx", region: null,
|
|
|
|
|
entrySp: S._sp(), suspendedAt: 0, libctx: true, lcid: lcid
|
|
|
|
|
});
|
|
|
|
|
} else {
|
|
|
|
|
rec = S._actStack.length ? S._top() : S._windowLive;
|
|
|
|
|
}
|
|
|
|
|
if (!rec) {
|
|
|
|
|
rec = {
|
|
|
|
|
id: --S._anonSeq, kind: "untracked", region: null,
|
|
|
|
|
entrySp: S._sp(), suspendedAt: 0, anon: true
|
|
|
|
|
};
|
|
|
|
|
}
|
|
|
|
|
rec.sp = S._sp();
|
|
|
|
|
rec.suspendedAt = Date.now();
|
|
|
|
|
rec.waitKind = kind;
|
|
|
|
|
rec.waitToken = token;
|
|
|
|
|
S._suspended.set(rec.id, rec);
|
|
|
|
|
// A coroutine that parks stops being the "current context": whatever
|
|
|
|
|
// the event loop runs next would otherwise attribute ITS suspensions
|
|
|
|
|
// to this parked coroutine (g_current dangles — no C code runs on the
|
|
|
|
|
// unwind path). The arm below re-establishes it on resume.
|
|
|
|
|
if (lcid) S._libctxMakeCurrent(0);
|
|
|
|
|
if (S._windowLive === rec) {
|
|
|
|
|
// a RESUMED window just suspended again — its window ends here; put
|
|
|
|
|
// the enclosing stack back for whatever runs next.
|
|
|
|
|
S._windowLive = null;
|
|
|
|
|
if (rec.enclosingSp !== undefined) S._setSp(rec.enclosingSp);
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
}
|
|
|
|
|
// (a FIRST window's end — including its SP hand-back — is the wrap's
|
|
|
|
|
// finally; the completion hook frees the region.)
|
|
|
|
|
var gate = {};
|
|
|
|
|
gate.promise = new Promise(function (res, rej) {
|
|
|
|
|
gate.resolve = res;
|
|
|
|
|
gate.reject = rej;
|
|
|
|
|
});
|
|
|
|
|
p.then(
|
|
|
|
|
function (v) {
|
|
|
|
|
S._resumeReady.push({ rec: rec, gate: gate, value: v, rejected: false });
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
},
|
|
|
|
|
function (err) {
|
|
|
|
|
S._resumeReady.push({ rec: rec, gate: gate, value: err, rejected: true });
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
}
|
|
|
|
|
);
|
|
|
|
|
return gate.promise;
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// --- libcontext (KiCad coroutine) turnstile integration ----------------
|
|
|
|
|
// The coroutine backend manages its own spill REGIONS and SP save/restore
|
|
|
|
|
// around its awaits, but its engine-level resumes must still be
|
|
|
|
|
// SERIALIZED with everyone else's: un-turnstiled, the microtask that
|
|
|
|
|
// restores the coroutine's SP can interleave with a turnstile arm, and
|
|
|
|
|
// whichever runs last wins — the resumed wasm then spills into another
|
|
|
|
|
// activation's region (observed: PCB_SELECTION_TOOL's first Wait()
|
|
|
|
|
// trapping "memory access out of bounds" in Chromium, engine-ordering
|
|
|
|
|
// dependent). The coroutine's ENTER/RESUME awaits suspend the CALLING wx
|
|
|
|
|
// activation and route through promiseYield; the YIELD suspends the
|
|
|
|
|
// COROUTINE'S OWN activation and uses these two hooks instead (explicit
|
|
|
|
|
// SP, no _actStack attribution — at yield time the stack top is the
|
|
|
|
|
// caller, not the coroutine).
|
|
|
|
|
_libctxRecs: {},
|
|
|
|
|
// Re-point the C-side g_current at the activation being armed (0 = root).
|
|
|
|
|
// Leaf export; absent on heapless hosts (vitest) and pre-runtime — no-op.
|
|
|
|
|
_libctxMakeCurrent: function (lcid) {
|
|
|
|
|
try {
|
|
|
|
|
if (typeof _pcbjam_libctx_make_current === "function") _pcbjam_libctx_make_current(lcid);
|
|
|
|
|
else if (typeof Module !== "undefined" && Module["_pcbjam_libctx_make_current"]) Module["_pcbjam_libctx_make_current"](lcid);
|
|
|
|
|
} catch (e) { /* pre-runtime */ }
|
|
|
|
|
},
|
|
|
|
|
libctxSuspend: function (id, p, sp) {
|
|
|
|
|
var S = this;
|
|
|
|
|
var rec = S._libctxRecs[id] || (S._libctxRecs[id] = {
|
|
|
|
|
id: "lc" + id, kind: "libctx", region: null,
|
|
|
|
|
entrySp: sp, suspendedAt: 0, libctx: true, lcid: id
|
|
|
|
|
});
|
|
|
|
|
rec.sp = sp;
|
|
|
|
|
rec.suspendedAt = Date.now();
|
|
|
|
|
rec.waitKind = "libctx";
|
|
|
|
|
rec.waitToken = 0;
|
|
|
|
|
S._suspended.set(rec.id, rec);
|
|
|
|
|
// Parked: clear the C-side current-context pointer (see _suspendOn).
|
|
|
|
|
S._libctxMakeCurrent(0);
|
|
|
|
|
if (S._windowLive === rec) {
|
|
|
|
|
S._windowLive = null;
|
|
|
|
|
if (rec.enclosingSp !== undefined) S._setSp(rec.enclosingSp);
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
}
|
|
|
|
|
var gate = {};
|
|
|
|
|
gate.promise = new Promise(function (res, rej) {
|
|
|
|
|
gate.resolve = res;
|
|
|
|
|
gate.reject = rej;
|
|
|
|
|
});
|
|
|
|
|
p.then(
|
|
|
|
|
function (v) {
|
|
|
|
|
S._resumeReady.push({ rec: rec, gate: gate, value: v, rejected: false });
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
},
|
|
|
|
|
function (err) {
|
|
|
|
|
S._resumeReady.push({ rec: rec, gate: gate, value: err, rejected: true });
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
}
|
|
|
|
|
);
|
|
|
|
|
return gate.promise;
|
|
|
|
|
},
|
|
|
|
|
// Coroutine released while parked (quarantine contract): mark the record
|
|
|
|
|
// dead so the pump drops any late wake (never re-enter the freed body),
|
|
|
|
|
// and forget it. The rec object itself stays reachable from pending
|
|
|
|
|
// p.then closures — the dead flag is what protects those paths.
|
|
|
|
|
libctxQuarantine: function (id) {
|
|
|
|
|
var rec = this._libctxRecs[id];
|
|
|
|
|
if (!rec) return;
|
|
|
|
|
rec.dead = true;
|
|
|
|
|
this._suspended.delete(rec.id);
|
|
|
|
|
if (this._windowLive === rec) {
|
jspi: fix the dead-tools ownership bug, emscripten-6 fallout, and green the full suite on Playwright 1.62
Live-app fix (Place Footprints / routing dead in Chrome): submodule
bumps carry the coroutine ownership fix (kicad 012d95ecb4) and the
handler-exception survival fix (wxwidgets 1b5f0e31f4).
Emscripten-6 fallout:
- occ/ngspice worker wrappers: mainScriptUrlOrBlob was removed
upstream; pthread children re-run the wrapper blob, so an em-pthread
realm now importScripts the glue and gets out of the way (before:
recursive service boots, pool never fills, silent 180s boot hangs —
every occ spec and ngspice bg_run).
- Makefile.wasm: -sASYNCIFY frankenlinks on the no-wx coroutine repro
targets ported to -sJSPI (the JSPI-only libcontext crashed at first
yield under them); mainloop/gl repro pages drive their tick through a
promising export (emscripten_set_main_loop callbacks cannot suspend);
retired inject-dyncall-shims lines removed (targets were unbuildable
since Phase 8); $stringToNewUTF8 force-included (the EM_ASM value
bridge aborted the runtime on the first decoded exception).
- fiber-park levers: neither embind shape can drive suspending levers
(plain throws on strict-JSPI Firefox; emscripten::async() re-executes
its invoker on settle) — kept sync for manual Chromium probing, spec
coverage moved to the jspi-coroutine harness (18 cases).
Suite work:
- Playwright 1.61.1 -> 1.62.1 (Firefox 153: JSPI on by default).
- fiber-resume-park.spec retired -> coroutine-lifecycle.spec: census
gate over boot / board load / chooser open / cancel (deterministically
red on the pre-fix build).
- Blind asyncify-era pins re-keyed: quasimodal-strand + wait-beacons
beacon regexes, footprint-chooser-close liveness -> wx parking-timer
heartbeat (scheduler counters idle flat on Firefox).
- occ/ngspice test providers: 60s boot timeout + worker error
surfacing (a worker death used to be a silent 180s timeout).
- Harness pages: stale 9.99 config dir -> 10.0 (library_manager wxCHECK
noise, chooser had no libraries).
- gal-webgl harness: missing artifacts rebuilt (boost/glm extracted to
the host sysroot), PgmOrNull stub added for the rebased GAL.
- jspi-scheduler: clean-shutdown console line restored (app-quit
contract), quarantine never yanks SP from a live window.
Gates: test:e2e 699 passed / 0 failed (wx-chromium, kicad-firefox,
kicad-chromium, jspi-firefox, coroutine-firefox); web ff/cr/mobile 71
passed; lint:ci-coverage 166, lint:determinism 163, screenshots
manifest 492 current, corpus 7/7, tools:contract green. Offline
screenshot baselines show expected mass drift from the engine bump —
re-baseline (screenshots:noise -> promote) is a follow-up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-13 17:41:28 +02:00
|
|
|
// Released from within its own running slice: that slice's wasm is
|
|
|
|
|
// executing on this region RIGHT NOW — restoring the enclosing SP
|
|
|
|
|
// here would yank the stack out from under it. Just drop the window
|
|
|
|
|
// marker; the turnstile moves on when this job ends. (g_current is
|
|
|
|
|
// reset C-side by release_fcontext.)
|
|
|
|
|
console.warn("[wx-scheduler] quarantine of the LIVE window lc" + id
|
|
|
|
|
+ " — self-release mid-slice; skipping SP restore");
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
this._windowLive = null;
|
|
|
|
|
}
|
|
|
|
|
delete this._libctxRecs[id];
|
|
|
|
|
this._note("libctxQuarantine", "libctx", rec.id);
|
|
|
|
|
var S = this;
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
},
|
|
|
|
|
// Coroutine's entry activation completed (finished or trapped): end its
|
|
|
|
|
// window so the turnstile moves on.
|
|
|
|
|
libctxEnd: function (id) {
|
|
|
|
|
var S = this;
|
|
|
|
|
var rec = S._libctxRecs[id];
|
|
|
|
|
if (!rec) return;
|
|
|
|
|
S._suspended.delete(rec.id);
|
|
|
|
|
if (S._windowLive === rec) {
|
|
|
|
|
S._windowLive = null;
|
|
|
|
|
if (rec.enclosingSp !== undefined) S._setSp(rec.enclosingSp);
|
|
|
|
|
}
|
|
|
|
|
delete S._libctxRecs[id];
|
|
|
|
|
queueMicrotask(function () { S._pumpResume(); });
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// Suspension helpers the wx EM_ASYNC_JS bodies route through, so every
|
|
|
|
|
// park shares the one discipline implementation.
|
|
|
|
|
frameYield: function () {
|
|
|
|
|
return this._suspendOn(
|
|
|
|
|
new Promise(function (r) { requestAnimationFrame(function () { r(0); }); }),
|
|
|
|
|
"frame", 0);
|
|
|
|
|
},
|
|
|
|
|
sleepYield: function (ms) {
|
|
|
|
|
return this._suspendOn(
|
|
|
|
|
new Promise(function (r) { setTimeout(function () { r(0); }, ms); }),
|
|
|
|
|
"sleep", 0);
|
|
|
|
|
},
|
|
|
|
|
promiseYield: function (p, kind) {
|
|
|
|
|
return this._suspendOn(Promise.resolve(p), kind || "promise", 0);
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
// Called once the runtime is up: wrap the app's promising entry exports.
|
|
|
|
|
// The list mirrors -sJSPI_EXPORTS (main excluded: the runtime calls it
|
|
|
|
|
// before this hook can matter, and boot suspensions predate any tracked
|
|
|
|
|
// activation anyway — see _suspendOn's null-rec path).
|
|
|
|
|
installExportWraps: function (names) {
|
|
|
|
|
var wrapped = 0;
|
|
|
|
|
for (var i = 0; i < names.length; i++) {
|
|
|
|
|
var key = "_" + names[i];
|
|
|
|
|
if (typeof Module[key] === "function") {
|
|
|
|
|
Module[key] = this._wrapPromising(names[i], Module[key]);
|
|
|
|
|
wrapped++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
this._note("wrapped", "exports", wrapped);
|
|
|
|
|
return wrapped;
|
|
|
|
|
},
|
|
|
|
|
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// --- observability ------------------------------------------------------
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
_ring: [],
|
|
|
|
|
_note: function (ev, a, b) {
|
|
|
|
|
this._ring.push([Date.now(), ev, String(a), b | 0]);
|
|
|
|
|
if (this._ring.length > RING_CAP) this._ring.shift();
|
|
|
|
|
},
|
|
|
|
|
|
|
|
|
|
dump: function () {
|
|
|
|
|
var acts = [];
|
|
|
|
|
this._suspended.forEach(function (rec) {
|
|
|
|
|
acts.push({
|
|
|
|
|
id: rec.id, kind: rec.kind, waitKind: rec.waitKind || null,
|
|
|
|
|
token: rec.waitToken || 0,
|
|
|
|
|
suspendedMs: rec.suspendedAt ? Date.now() - rec.suspendedAt : 0
|
|
|
|
|
});
|
|
|
|
|
});
|
|
|
|
|
return {
|
|
|
|
|
dead: this.dead,
|
|
|
|
|
waitsBegun: this.waitsBegun,
|
|
|
|
|
waitsResolved: this.waitsResolved,
|
|
|
|
|
earlyWaitResolves: this.earlyWaitResolves,
|
|
|
|
|
pendingWaits: this.waits.size,
|
|
|
|
|
runningActivations: this._actStack.length,
|
|
|
|
|
suspendedActivations: acts,
|
|
|
|
|
mutatorsWrapped: this.mutatorsWrapped,
|
|
|
|
|
mutatorsDelivered: this.mutatorsDelivered,
|
|
|
|
|
mutatorQueueDepth: this.mutatorQueue.length,
|
|
|
|
|
ring: this._ring.slice(-64)
|
|
|
|
|
};
|
|
|
|
|
}
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
globalThis.__wxScheduler = S;
|
|
|
|
|
globalThis.__wxSchedulerInstalled = true; // wxWasmSchedulerAssertInstalled probe
|
|
|
|
|
|
jspi cleanup: remove the asyncify-era residue — dead code, conditionals, pipeline scaffolding, stale prose
The runtime is JSPI-only; this removes everything that still pretended
otherwise. Three exhaustive sweeps (C++/JS+build+CI/tests+docs) drove
the inventory; every deletion verified by grep closure + full gates.
Broken-right-now fixes:
- deploy-staging.yml passed the retired opt_level input — the workflow
could not even start. Removed.
- env.sh carried dead exports with a live -sASYNCIFY=1 inside
(WASM_LDFLAGS/PTHREAD_LDFLAGS, zero consumers). Removed; the
WASM_LEGACY_EXCEPTIONS rationale rewritten to the real reason.
- docker/build.sh exported PCBJAM_ASYNC_BACKEND (read nowhere). Gone.
Dead weight removed:
- binaryen submodule (nothing builds or invokes it), wasm-opt-bench
workflow + scripts/bench/, get-wasm-opt.sh, diagnostics.js (242 lines
of Asyncify-API-only code), the KICAD_PIPELINE background-postprocess
scaffolding (existed to parallelize the deleted wasm-opt phase; the
postprocess is a seconds-long node script and now runs inline),
build-monitor's dead asyncify rows, sched-context orphan build
output, dead .gitignore entries, the .jspi-assets spike dir (the two
wf-result research JSONs moved to docs/features/async/migration-evidence/).
- bindings: fiber_park.h + its 12 embind registrations (broken-if-
called under JSPI), the kicadOpenFileStart/OPEN_JOB starter route,
main_stack_runner.h + 5 includes, the always-null context-sleep weak
hook in nanosleep_yield.c.
- shim: the backend field (installed-flag idempotency instead),
noteContextWait (dead both sides), the __wxAsyncifyDump alias (+ the
WasmTool fallback and string-dump normalize branch).
- web: the emscripten-6-ignored mainScriptUrlOrBlob option in boot.ts
(gerber-demo keeps it: it loads the deployed CDN release, which
predates emscripten 6 — noted inline).
Conditionals: all 'backend === jspi' checks reduced to scheduler-
presence checks; races_quiescent re-keyed from Asyncify.state (vacuous)
to real backlog quiescence (resumeReady/mutatorQueue — NOT _windowLive,
which is the probing activation's own window by definition).
Renames (identifiers only, no file renames): ASYNC_LINK_FLAGS→
JSPI_LINK_FLAGS and Makefile ASYNC_LDFLAGS→JSPI_LDFLAGS,
kicadCollabFiberBusy→kicadCollabBusy (embind + web + tests),
collab_common.h fiber*→apply*/coroutine naming, asyncifySignatures→
wasmTrapSignatures (lists byte-identical).
Tests: the two remaining vacuous [wx-asyncify]/fiber-resume-refused
asserts re-keyed to live JSPI beacons; eeschema-load's failure message
no longer sends the developer to a deleted script; wait-beacons' dead
families/parser deleted; lane-0 legacy-glue guards removed (lane 0 is
unconstructible); the embind test.fail re-gated with the JSPI reason
(plain embind invokers cannot suspend — verified still failing);
lint-determinism now scans tests/jspi (166 files clean);
eeschema-collab local-move gated to chromium (~50% flaky on FF even
solo; pcbnew twin covers both engines).
Docs: DEBUG.md rewritten as the JSPI debugging guide; build.md
describes the single-phase build; docs/features/async/README.md
banner-marked historical and repointed at the NEW
23-jspi-runtime.md (current architecture: export census, turnstile,
libcontext ownership + refusal contract, embind call shapes, the
em-pthread service-wrapper trick, exception policy, known gaps).
Gates on the cleaned tree: test:e2e 725 passed / 0 failed (after the
quiescence-probe fix; the 3 other reds were verified contention flakes
solo-green or the documented FF gate), web 76/0, jspi 18/18 both
engines, vitest 295/295 + 17/17, all lints green, live-app census
clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016X9eh1s5sTx1o9Em9KBuwR
2026-08-14 09:25:32 +02:00
|
|
|
// --- diagnostic signals ---------------------------------------------------
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
// SuspendError attributor: a SuspendError means a PLAIN (non-promising)
|
|
|
|
|
// wasm entry tried to park — a missed -sJSPI_EXPORTS/installExportWraps
|
|
|
|
|
// entry. The engine cannot say WHICH export, but the live dump (what was
|
|
|
|
|
// wrapped, what was executing) is exactly the targeting data needed.
|
|
|
|
|
if (typeof addEventListener === "function") {
|
|
|
|
|
var suspendErr = function (m) {
|
|
|
|
|
if (!m || !/suspend/i.test(String(m))) return;
|
|
|
|
|
console.error("[wx-scheduler] SuspendError: a NON-promising wasm entry "
|
|
|
|
|
+ "tried to park — add the missing entry export to -sJSPI_EXPORTS and "
|
|
|
|
|
+ "installExportWraps. dump=" + JSON.stringify(S.dump()));
|
|
|
|
|
};
|
|
|
|
|
addEventListener("unhandledrejection", function (ev) {
|
|
|
|
|
suspendErr(ev && ev.reason && (ev.reason.message || ev.reason));
|
|
|
|
|
});
|
|
|
|
|
addEventListener("error", function (ev) {
|
|
|
|
|
suspendErr(ev && (ev.message || (ev.error && ev.error.message)));
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
// Lost-wake watchdog: an activation parked on a TOKEN wait whose registry
|
|
|
|
|
// entry is GONE (resolved+consumed or never registered) can never be
|
|
|
|
|
// resumed — a lost wake. Frame/sleep parks are excluded (a hidden tab
|
|
|
|
|
// legitimately parks the frame yield for minutes).
|
|
|
|
|
setInterval(function () {
|
|
|
|
|
if (S.dead) return;
|
|
|
|
|
S._suspended.forEach(function (rec) {
|
|
|
|
|
if (!rec.waitToken || rec.waitKind === "frame" || rec.waitKind === "sleep") return;
|
|
|
|
|
if (rec.suspendedAt && Date.now() - rec.suspendedAt > 30000
|
|
|
|
|
&& !S.waits.has(rec.waitToken) && !rec._lostWakeWarned) {
|
|
|
|
|
rec._lostWakeWarned = true;
|
|
|
|
|
console.warn("[wx-scheduler] LOST WAKE: act " + rec.id + ":" + rec.kind
|
|
|
|
|
+ " parked " + Math.round((Date.now() - rec.suspendedAt) / 1000)
|
|
|
|
|
+ "s on " + rec.waitKind + "/" + rec.waitToken
|
|
|
|
|
+ " but the wait is no longer registered");
|
|
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
}, 10000);
|
|
|
|
|
globalThis.__wxWaitDump = function () { return S.dump(); };
|
|
|
|
|
|
|
|
|
|
// Self-install the activation wraps once the runtime is up (this file ships
|
|
|
|
|
// as a --pre-js, so Module exists here). The name set mirrors the
|
|
|
|
|
// suspension-capable half of -sJSPI_EXPORTS; absent names are skipped.
|
|
|
|
|
if (typeof Module !== "undefined") {
|
|
|
|
|
var prevInit = Module["onRuntimeInitialized"];
|
|
|
|
|
Module["onRuntimeInitialized"] = function () {
|
|
|
|
|
if (prevInit) prevInit();
|
|
|
|
|
S.installExportWraps([
|
|
|
|
|
"wx_dom_event", "wx_dom_mouse", "wx_window_close", "wx_window_move",
|
|
|
|
|
"wx_window_resize", "ProcessEvents", "wxWasmMailboxTick",
|
jspi: retire the asyncify pipeline — knob, post-link tail, binaryen hooks
Phase 8 in the parent repo. Deleted: asyncify-scheduler.js, apply-asyncify.sh,
apply-finalize.sh, inject-dyncall-shims.sh, asyncify-imports/removelist.txt,
the wasm-opt/finalize stub pair, scripts/binaryen-hoist-pass/ (the fork stays
a dormant submodule; removal is a follow-up), bench/wasm-opt-bench.sh (README
marked historical), wasm/shims/context_sleep.cpp, and the sched-context
harness app + Makefile targets.
PCBJAM_ASYNC_BACKEND is gone: build-wx-wasm.sh hardcodes the jspi stamp
(still force-cleans pre-migration trees), build-kicad-target.sh gives editors
the JSPI link surface and the CLIs nothing (they pin ASYNCIFY=0), the stub
dance is replaced by an unconditional .real-restore, build-wasm-test.sh lost
its whole post-link loop, docker/build.sh's postprocess is the ENV shim only,
and Makefile.wasm links every app JSPI with the scheduler shim as a tracked
prerequisite. pcbjam_async_policy.h keys on __EMSCRIPTEN__.
jspi-scheduler.js: wxWasmMainLoopPump dropped from the wrap census (the
export died with the D5 detach); inert [TRACE] instrumentation removed.
CI: wasm-build.yml rewritten for the single-cache pipeline (one output cache
keyed on compile inputs; post-processed bytes cached after the shim);
opt_level input removed from both callers. wasm-cache-hash.mjs inputs now
cover patch-env-shim.mjs + jspi-scheduler.js + jspi-exports.txt.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 08:39:12 +02:00
|
|
|
"wxWasmTopLevelTick", "wxWasmJobTick"
|
jspi: migration phases 0-7 — build knob, scheduler shim, test successor suite
Toolchain: emsdk 6.0.6 (versions.sh; cache-hash keys on it). Build knob
PCBJAM_ASYNC_BACKEND=jspi|asyncify: build-kicad-target.sh links editors with
-sJSPI + -sJSPI_EXPORTS=@scripts/common/jspi-exports.txt + --pre-js
jspi-scheduler.js (no DYNCALLS, no post-link asyncify pipeline); wx build
stamps the backend and forces clean on flip or unknown provenance;
docker/build.sh passes the knob, seeds the emscripten ports cache from the
volume every launch, jspi postprocess = patch-env-shim only.
scripts/common/shims/jspi-scheduler.js: the JSPI successor scheduler —
token-wait registry, resume turnstile (one armed resume between engine
re-entries, SP swaps only at microtask boundaries), green-region spill
stacks (16-aligned tops), S1 embind mutator FIFO lane + parker wraps, S6
shutdown, libctx integration hooks (suspend/end/quarantine + g_current
arm/clear), SuspendError attributor, lost-wake + stuck-window watchdogs,
__wxWaitDump observability.
Embind: PARKER registrations get emscripten::async() under PCBJAM_JSPI
(wasm/bindings/pcbjam_async_policy.h). nanosleep yields route via the shim.
Tests: tests/asyncify -> tests/jspi successor suite (jspi-stack red/green
shadow-stack battery, jspi-coroutine MiniCoro harness, suspend-races
semantic scenarios + __wxWaitDump books coherence); projects jspi-firefox/
jspi-chrome (asyncify-webkit retired — no JSPI in WebKit); unconditional
Firefox JSPI pref; guard-beacons -> wait-beacons (+wxScheduler/libctxJspi
families); Makefile.wasm links test apps against JSPI with the shim as a
tracked link prerequisite.
Web: WasmTool setRo await + __wxWaitDump forensics, open-flow contained
promise, scheduler-shim.test.ts retargeted (8 green).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
2026-08-13 07:06:24 +02:00
|
|
|
]);
|
|
|
|
|
// KiCad-only surfaces; both installers skip absent names, so the wx
|
|
|
|
|
// test apps (no embind) pass through here untouched.
|
|
|
|
|
var parkers = S._wrapParkers();
|
|
|
|
|
S._wrapMutators();
|
|
|
|
|
S._preallocRegions(parkers > 0);
|
|
|
|
|
};
|
|
|
|
|
}
|
|
|
|
|
})();
|