pcbjam/tests/apps/standalone/asyncify-races/races_test.cpp

897 lines
33 KiB
C++
Raw Normal View History

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
// races_test.cpp - suspension race-condition red-green harness.
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +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
// Reproduces the KiCad-WASM suspension failure modes deterministically so the
// scheduler-shim fixes stay pinned by tests (see features/async/ research
// dossier). The scenarios were decoded on the retired asyncify runtime; the
// topologies they stage are runtime-agnostic and now pin the JSPI scheduler's
// turnstile/window contracts:
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +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
// - The app performs a coroutine swap during OnInit BEFORE the main loop
// parks. This is the load-bearing topology detail: the park then lands on
// a stack that already completed a suspension chain (the startup shape
// that historically wedged the asyncify fiber trampoline — the KiCad
// schematic/PCB tool hang).
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
// coroutine-nested/nested_test.cpp does NOT do a pre-park swap, which is
// why it never reproduced that hang.
//
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
// - EM_ASYNC_JS sleeps (modal dialogs, token waits) overlapping coroutine
// swaps — historically the single-slot Asyncify.currData clobber family
// (the KiCad clipboard "index out of bounds" crash), now the concurrent
// multi-suspension bookkeeping the scheduler's turnstile serializes.
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
//
// URL parameters:
// ?only=<scenario> run a single scenario instead of the default battery
// (used for scenarios that intentionally wedge/crash)
// ?mode=sleep-park make the LAST pre-park suspension a sleep instead of a
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
// coroutine swap, so the park arrives out of a sleep
// resume rather than a swap (scenario
// unwind_through_promise)
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +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
// Output protocol (polled by tests/jspi/suspend-races.spec.ts):
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
// [ASYNCIFY_RACES] CASE <name>
// [ASYNCIFY_RACES] PASS <name> / FAIL <name> :: <detail>
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
// [ASYNCIFY_RACES] WATCHDOG <name> windowLive=.. resumeReady=..
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
// [ASYNCIFY_RACES] SUMMARY total=N passed=N failed=N
#include "wx/wx.h"
#include "wx/dialog.h"
#include "wx/evtloop.h"
#include "wx/timer.h"
#include "kicad_coroutine_harness.h"
#ifdef __EMSCRIPTEN__
#include <emscripten/emscripten.h>
#include <emscripten/em_js.h>
#endif
#include <functional>
#include <memory>
#include <sstream>
#include <string>
#include <vector>
using coroutine_test::TestCoroutine;
namespace
{
constexpr int ID_SCENARIO_TIMER = wxID_HIGHEST + 700;
constexpr int ID_POLL_TIMER = wxID_HIGHEST + 701;
struct CaseContext
{
bool passed = true;
std::vector<std::string> failures;
void Expect( bool aCondition, const std::string& aMessage )
{
if( !aCondition )
{
passed = false;
failures.push_back( aMessage );
}
}
};
std::string JoinFailures( const std::vector<std::string>& aFailures )
{
std::ostringstream oss;
for( std::size_t i = 0; i < aFailures.size(); ++i )
{
if( i > 0 )
oss << " | ";
oss << aFailures[i];
}
return oss.str();
}
void LogLine( const std::string& aLine )
{
#ifdef __EMSCRIPTEN__
EM_ASM( { console.log( UTF8ToString( $0 ) ); }, aLine.c_str() );
#else
std::printf( "%s\n", aLine.c_str() );
#endif
}
#ifdef __EMSCRIPTEN__
// --- JS helpers ----------------------------------------------------------------
// Park the calling stack until JS resolves the token (races_resolve_token_after).
EM_ASYNC_JS( int, races_await_token, ( int aToken ), {
Module.__racesWaits = Module.__racesWaits || {};
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 p = new Promise( ( resolve ) => { Module.__racesWaits[aToken] = resolve; } );
// JSPI: EVERY suspension must route through the scheduler's turnstile —
// a raw await's engine-level resume bypasses the SP discipline and never
// ends the current window (windowLive wedges the pump).
var S = globalThis.__wxScheduler;
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( S )
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 await S.promiseYield( p, 'races-token' );
return await p;
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
} );
// Resolve a parked token after a JS-side delay (independent of the C++ world,
// so it fires even while every C++ stack is parked).
EM_JS( void, races_resolve_token_after, ( int aToken, int aValue, int aDelayMs ), {
setTimeout( function() {
var w = Module.__racesWaits && Module.__racesWaits[aToken];
if( w ) { delete Module.__racesWaits[aToken]; w( aValue ); }
else { console.log( '[ASYNCIFY_RACES] WARN resolve-token ' + aToken + ' had no waiter' ); }
}, aDelayMs );
} );
// Plain parked sleep.
EM_ASYNC_JS( int, races_sleep_ms, ( int aMs ), {
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 S = globalThis.__wxScheduler;
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( S ) {
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
await S.sleepYield( aMs ); // turnstile-routed (see races_await_token)
return 1;
}
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
await new Promise( ( r ) => setTimeout( r, aMs ) );
return 1;
} );
// Schedule an async ccall into an exported C function on a FRESH JS/wasm stack.
// This is how the harness drives suspensions while every C++ stack is parked
// (mirrors KiCad's EndModal/clipboard work arriving on fresh event stacks).
EM_JS( void, races_schedule_ccall, ( const char* aFunc, int aDelayMs ), {
var fn = UTF8ToString( aFunc );
setTimeout( function() {
try {
var p = Module.ccall( fn, null, [], [], { async: true } );
if( p && p.catch )
p.catch( function( e ) { console.error( '[ASYNCIFY_RACES] ccall ' + fn + ' rejected: ' + e ); } );
} catch( e ) {
console.error( '[ASYNCIFY_RACES] ccall ' + fn + ' threw: ' + e );
}
}, aDelayMs );
} );
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
// Watchdog: if the scenario hasn't marked itself done in aMs, dump the
// scheduler state and emit a FAIL line. JS-side, so it fires even when C++ is
// wedged.
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
EM_JS( void, races_arm_watchdog, ( const char* aName, int aMs ), {
var name = UTF8ToString( aName );
Module.__racesDone = Module.__racesDone || {};
setTimeout( function() {
if( !Module.__racesDone[name] ) {
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
var S = globalThis.__wxScheduler;
var wl = S ? !!S._windowLive : 'n/a';
var rr = ( S && S._resumeReady ) ? S._resumeReady.length : 'n/a';
console.log( '[ASYNCIFY_RACES] WATCHDOG ' + name + ' windowLive=' + wl
+ ' resumeReady=' + rr );
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
console.log( '[ASYNCIFY_RACES] FAIL ' + name + ' :: watchdog timeout (suspension never completed)' );
}
}, aMs );
} );
EM_JS( void, races_mark_done, ( const char* aName ), {
Module.__racesDone = Module.__racesDone || {};
Module.__racesDone[UTF8ToString( aName )] = 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
// Quiescence invariant sampled from C++ between scenarios, keyed to the JSPI
// scheduler (globalThis.__wxScheduler): between scenarios no resume window may
// still be live (_windowLive) and no resume may sit queued (_resumeReady).
// The main loop's own per-frame suspension (wxWasmYieldToBrowser) does not
// count against either — its window closes when the frame yield's suspension
// completes, before the next C++ code runs. A genuinely stuck suspension is
// additionally caught by the scenario watchdogs. Without a scheduler (the
// raw-await harness builds) there is no shared state to wedge — quiescent by
// construction.
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
EM_JS( int, races_quiescent, (), {
try {
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
var S = globalThis.__wxScheduler;
if( !S )
return 1;
// NOTE: _windowLive is NOT part of quiescence here — this probe runs
// from INSIDE a tracked activation, whose own window is live by
// definition. A wedge manifests as backlog: queued-but-unarmed
// resumes or a stuck mutator FIFO (the 2s force-clear watchdog keys
// on the same signal).
var rrOk = !S._resumeReady || S._resumeReady.length === 0;
var mqOk = !S.mutatorQueue || S.mutatorQueue.length === 0;
return ( rrOk && mqOk ) ? 1 : 0;
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
} catch( e ) {
return 0;
}
} );
EM_JS( void, races_log_state, ( const char* aTag ), {
try {
var tag = UTF8ToString( aTag );
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
var S = globalThis.__wxScheduler;
var wl = S ? !!S._windowLive : 'n/a';
var rr = ( S && S._resumeReady ) ? S._resumeReady.length : 'n/a';
console.log( '[ASYNCIFY_RACES] STATE ' + tag + ' windowLive=' + wl
+ ' resumeReady=' + rr );
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
} catch( e ) {}
} );
// Throw a raw JS error out of the current wasm frame. Used inside the nested
// quasi-modal pump to force the pump's `await ccall('ProcessEvents')` to reject
// (the c27fe8bf silent-stall path).
EM_JS( void, races_throw_js_error, (), {
throw new Error( 'races forced pump error' );
} );
#endif // __EMSCRIPTEN__
} // namespace
// ---------------------------------------------------------------------------------
// Exported helpers driven from JS on fresh stacks (fire-and-forget async ccalls).
// Globals because ccall'd plain C functions have no frame pointer.
// ---------------------------------------------------------------------------------
static int g_token2Value = 0; // out_of_order: second parker's result
static bool g_token2Done = false;
static std::vector<std::string>* g_oooSeq = nullptr;
static int g_wdtBValue = 0; // wakeup_during_transition: B-side result
static bool g_wdtBDone = false;
static wxDialog* g_activeModal = nullptr;
extern "C" {
// A complete fiber swap cycle on a fresh stack (Call + Resume to completion).
// Mirrors KiCad's EndModal-driven tool teardown swaps that clobber a parked sleep.
EMSCRIPTEN_KEEPALIVE void races_swap_once()
{
TestCoroutine co( []( TestCoroutine& self ) { self.Yield( 7 ); } );
co.Call( 1 );
co.Resume( 2 );
LogLine( "[ASYNCIFY_RACES] SWAP-ONCE done" );
}
// Park a second, independent stack on token 2 (out_of_order scenario).
EMSCRIPTEN_KEEPALIVE void races_park_token2()
{
#ifdef __EMSCRIPTEN__
LogLine( "[ASYNCIFY_RACES] OOO second parker parking" );
g_token2Value = races_await_token( 2 );
g_token2Done = true;
if( g_oooSeq )
g_oooSeq->push_back( "t2" );
LogLine( "[ASYNCIFY_RACES] OOO second parker resumed" );
#endif
}
// Park a stack on token 11 (wakeup_during_transition B side).
EMSCRIPTEN_KEEPALIVE void races_wdt_park_b()
{
#ifdef __EMSCRIPTEN__
LogLine( "[ASYNCIFY_RACES] WDT B parking" );
g_wdtBValue = races_await_token( 11 );
g_wdtBDone = true;
LogLine( "[ASYNCIFY_RACES] WDT B resumed" );
#endif
}
// End the active modal from a fresh stack (mirrors KiCad's EndModal arriving
// while a clipboard sleep is parked).
EMSCRIPTEN_KEEPALIVE void races_end_active_modal()
{
if( g_activeModal )
{
LogLine( "[ASYNCIFY_RACES] ending active modal from fresh stack" );
g_activeModal->EndModal( wxID_OK );
}
}
} // extern "C"
// ---------------------------------------------------------------------------------
// The scenario-driver frame
// ---------------------------------------------------------------------------------
class RacesDialog : public wxDialog
{
public:
RacesDialog( wxWindow* aParent, const wxString& aTag ) :
wxDialog( aParent, wxID_ANY, aTag, wxDefaultPosition, wxSize( 260, 120 ) )
{
}
};
class RacesFrame : public wxFrame
{
public:
RacesFrame( const std::string& aOnly, bool aSleepParkMode ) :
wxFrame( nullptr, wxID_ANY, "Asyncify Races Test", wxDefaultPosition,
wxSize( 900, 600 ) ),
m_only( aOnly ),
m_sleepParkMode( aSleepParkMode ),
m_scenarioTimer( this, ID_SCENARIO_TIMER ),
m_pollTimer( this, ID_POLL_TIMER )
{
wxPanel* panel = new wxPanel( this );
wxBoxSizer* sizer = new wxBoxSizer( wxVERTICAL );
m_summary = new wxStaticText( panel, wxID_ANY, "Running asyncify race scenarios..." );
sizer->Add( m_summary, 0, wxEXPAND | wxALL, 8 );
panel->SetSizer( sizer );
CreateStatusBar();
Bind( wxEVT_TIMER, &RacesFrame::OnScenarioTimer, this, ID_SCENARIO_TIMER );
Bind( wxEVT_TIMER, &RacesFrame::OnPollTimer, this, ID_POLL_TIMER );
// Scenarios run AFTER the main loop parks (CallAfter fires on the first
// rAF ticks) - the same place KiCad tool interactions live.
CallAfter( [this]() { RunNext(); } );
}
private:
// ----- bookkeeping -----
bool ShouldRun( const std::string& aName ) const
{
if( m_sleepParkMode )
return aName == "unwind_through_promise";
if( !m_only.empty() )
return m_only == aName;
// Default battery: everything that is safe to chain in one page load.
// modal_in_modal_in_modal, wakeup_during_transition and
// nested_quasi_modal_pump_error are ?only= singles - they intentionally
// wedge/crash while their bugs are unfixed and would kill the chain.
return aName == "post_park_fiber_swap"
|| aName == "sleep_inside_fiber_inside_modal"
|| aName == "out_of_order_sleep_resolution"
|| aName == "long_parked_sleep_clobbered_by_swap";
}
void Finalize( const std::string& aName, CaseContext&& aCtx )
{
#ifdef __EMSCRIPTEN__
races_mark_done( aName.c_str() );
#endif
if( aCtx.passed )
LogLine( "[ASYNCIFY_RACES] PASS " + aName );
else
LogLine( "[ASYNCIFY_RACES] FAIL " + aName + " :: " + JoinFailures( aCtx.failures ) );
m_total += 1;
m_passed += aCtx.passed ? 1 : 0;
CallAfter( [this]() { RunNext(); } );
}
void CheckQuiescent( CaseContext& aCtx, const std::string& aWhere )
{
#ifdef __EMSCRIPTEN__
aCtx.Expect( races_quiescent() == 1,
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
"scheduler not quiescent " + aWhere
+ " (windowLive/resumeReady - see STATE log)" );
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
if( races_quiescent() != 1 )
races_log_state( ( "non-quiescent-" + aWhere ).c_str() );
#endif
}
void RunNext()
{
static const std::vector<std::pair<std::string, void ( RacesFrame::* )()>> ALL = {
{ "post_park_fiber_swap", &RacesFrame::Scenario_PostParkFiberSwap },
{ "modal_in_modal_in_modal", &RacesFrame::Scenario_TripleModal },
{ "sleep_inside_fiber_inside_modal", &RacesFrame::Scenario_SleepInsideFiberInsideModal },
{ "out_of_order_sleep_resolution", &RacesFrame::Scenario_OutOfOrder },
{ "long_parked_sleep_clobbered_by_swap", &RacesFrame::Scenario_LongParkedSleep },
{ "wakeup_during_transition", &RacesFrame::Scenario_WakeupDuringTransition },
{ "nested_quasi_modal_pump_error", &RacesFrame::Scenario_NestedPumpError },
{ "unwind_through_promise", &RacesFrame::Scenario_UnwindThroughPromise },
};
while( m_nextIndex < ALL.size() )
{
const auto& entry = ALL[m_nextIndex];
m_nextIndex += 1;
if( ShouldRun( entry.first ) )
{
LogLine( "[ASYNCIFY_RACES] CASE " + entry.first );
( this->*( entry.second ) )();
return;
}
}
FinalizeSuite();
}
void FinalizeSuite()
{
std::ostringstream oss;
oss << "[ASYNCIFY_RACES] SUMMARY total=" << m_total << " passed=" << m_passed
<< " failed=" << ( m_total - m_passed );
LogLine( oss.str() );
m_summary->SetLabel( wxString::Format( "Done: %d/%d passed", m_passed, m_total ) );
}
// ----- scenario 1: post_park_fiber_swap -------------------------------------
// The KiCad hang topology. OnInit already did a fiber swap, so the park throw
// went through the live trampoline. With the self-heal shim the guard is
// reset and this swap works; historically (ablated heal, retired at doc 20
// D-1) the guard stuck true, Call() never returned, and the watchdog fired.
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
void Scenario_PostParkFiberSwap()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "post_park_fiber_swap", 2500 );
races_log_state( "S1-pre-swap" );
#endif
CaseContext ctx;
{
TestCoroutine co( []( TestCoroutine& self ) { self.Yield( 42 ); } );
bool running = co.Call( 1 );
ctx.Expect( running, "post-park fiber should yield" );
ctx.Expect( co.LastReturnValue() == 42, "yield value should be 42" );
running = co.Resume( 2 );
ctx.Expect( !running, "post-park fiber should finish" );
}
#ifdef __EMSCRIPTEN__
races_log_state( "S1-post-swap" );
#endif
CheckQuiescent( ctx, "after post-park swap" );
Finalize( "post_park_fiber_swap", std::move( ctx ) );
}
// ----- scenario 2: modal_in_modal_in_modal ----------------------------------
// Three nested ShowModal sleeps (LIFO park stack three deep), closed
// innermost-first, each from a timer firing inside the innermost pump.
void Scenario_TripleModal()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "modal_in_modal_in_modal", 6000 );
#endif
m_tripleCtx = std::make_unique<CaseContext>();
m_tripleSeq.clear();
m_pendingScenario = [this]() { TripleLevelB(); };
m_scenarioTimer.StartOnce( 40 );
RacesDialog dlgA( this, "tripleA" );
m_dlgA = &dlgA;
int ra = dlgA.ShowModal(); // parks this (scenario) stack
m_dlgA = nullptr;
// Resumes only after B and C closed.
m_tripleSeq.push_back( "A" );
m_tripleCtx->Expect( ra == 101, "modal A should return 101, got " + std::to_string( ra ) );
m_tripleCtx->Expect( m_tripleSeq.size() == 3 && m_tripleSeq[0] == "C" && m_tripleSeq[1] == "B"
&& m_tripleSeq[2] == "A",
"modals should resume LIFO (C,B,A)" );
CheckQuiescent( *m_tripleCtx, "after triple modal" );
Finalize( "modal_in_modal_in_modal", std::move( *m_tripleCtx ) );
m_tripleCtx.reset();
}
void TripleLevelB()
{
m_pendingScenario = [this]() { TripleLevelC(); };
m_scenarioTimer.StartOnce( 40 );
RacesDialog dlgB( this, "tripleB" );
m_dlgB = &dlgB;
int rb = dlgB.ShowModal(); // parks the A-pump tick stack
m_dlgB = nullptr;
m_tripleSeq.push_back( "B" );
m_tripleCtx->Expect( rb == 102, "modal B should return 102, got " + std::to_string( rb ) );
if( m_dlgA )
m_dlgA->EndModal( 101 );
}
void TripleLevelC()
{
m_pendingScenario = [this]() {
if( m_dlgC )
m_dlgC->EndModal( 103 );
};
m_scenarioTimer.StartOnce( 40 );
RacesDialog dlgC( this, "tripleC" );
m_dlgC = &dlgC;
int rc = dlgC.ShowModal(); // parks the B-pump tick stack
m_dlgC = nullptr;
m_tripleSeq.push_back( "C" );
m_tripleCtx->Expect( rc == 103, "modal C should return 103, got " + std::to_string( rc ) );
if( m_dlgB )
m_dlgB->EndModal( 102 );
}
// ----- scenario 3: sleep_inside_fiber_inside_modal ---------------------------
// Modal sleep parked -> fiber started inside its pump -> fiber body parks in
// ANOTHER sleep -> resolves -> fiber yields -> resumes -> modal closes.
// Three different buffers (modal malloc, fiber struct, sleep malloc) in flight.
void Scenario_SleepInsideFiberInsideModal()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "sleep_inside_fiber_inside_modal", 6000 );
#endif
m_sifimCtx = std::make_unique<CaseContext>();
m_pendingScenario = [this]() { RunSleepInsideFiber(); };
m_scenarioTimer.StartOnce( 40 );
RacesDialog dlg( this, "sifim" );
m_dlgA = &dlg;
int result = dlg.ShowModal();
m_dlgA = nullptr;
m_sifimCtx->Expect( result == wxID_OK, "sifim modal should return wxID_OK" );
CheckQuiescent( *m_sifimCtx, "after sleep-inside-fiber-inside-modal" );
Finalize( "sleep_inside_fiber_inside_modal", std::move( *m_sifimCtx ) );
m_sifimCtx.reset();
}
void RunSleepInsideFiber()
{
#ifdef __EMSCRIPTEN__
CaseContext* ctx = m_sifimCtx.get();
{
TestCoroutine co( [ctx]( TestCoroutine& self ) {
// Parks the FIBER stack in a malloc'd sleep buffer while the
// modal sleep is also parked.
int r = races_sleep_ms( 150 );
ctx->Expect( r == 1, "fiber-side sleep should return 1" );
self.Yield( 901 );
} );
bool running = co.Call( 1 );
ctx->Expect( running, "fiber should yield after its sleep" );
ctx->Expect( co.LastReturnValue() == 901, "fiber yield value should be 901" );
running = co.Resume( 2 );
ctx->Expect( !running, "fiber should finish" );
}
if( m_dlgA )
m_dlgA->EndModal( wxID_OK );
#endif
}
// ----- scenario 4: out_of_order_sleep_resolution -----------------------------
// Two sleeps parked on independent stacks, resolved FIFO (not LIFO).
void Scenario_OutOfOrder()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "out_of_order_sleep_resolution", 4000 );
m_oooCtx = std::make_unique<CaseContext>();
m_oooSeqStore.clear();
g_oooSeq = &m_oooSeqStore;
g_token2Done = false;
g_token2Value = 0;
// Second parker arrives on a fresh stack at +50ms; resolutions at
// +600 (token 1, parked FIRST) and +1000 (token 2) - FIFO order.
races_schedule_ccall( "races_park_token2", 50 );
races_resolve_token_after( 1, 11, 600 );
races_resolve_token_after( 2, 22, 1000 );
int v1 = races_await_token( 1 ); // parks THIS stack
// Resumed at +600 while token 2 still parked.
m_oooSeqStore.push_back( "t1" );
m_oooCtx->Expect( v1 == 11, "token 1 value should be 11" );
// Wait (event-driven, not blocking) for the second parker to finish.
m_pollPredicate = []() { return g_token2Done; };
m_pollBudgetMs = 3000;
m_onPollDone = [this]( bool aOk ) {
m_oooCtx->Expect( aOk, "second parker should resume within budget" );
m_oooCtx->Expect( g_token2Value == 22, "token 2 value should be 22" );
m_oooCtx->Expect( m_oooSeqStore.size() == 2 && m_oooSeqStore[0] == "t1"
&& m_oooSeqStore[1] == "t2",
"continuations should run in resolution order t1,t2" );
g_oooSeq = nullptr;
CheckQuiescent( *m_oooCtx, "after out-of-order resolution" );
Finalize( "out_of_order_sleep_resolution", std::move( *m_oooCtx ) );
m_oooCtx.reset();
};
m_pollTimer.Start( 50 );
#else
CaseContext ctx;
Finalize( "out_of_order_sleep_resolution", std::move( ctx ) );
#endif
}
// ----- scenario 5: long_parked_sleep_clobbered_by_swap ------------------------
// The KiCad clipboard crash shape: a long-parked sleep crossed by complete
// fiber-swap cycles on fresh stacks. The scheduler shim restores the sleep's
// buffer at wakeUp; historically (no save/restore, ablation retired at doc 20
// D-1) doRewind read a clobbered currData -> "index out of bounds".
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
void Scenario_LongParkedSleep()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "long_parked_sleep_clobbered_by_swap", 4000 );
CaseContext ctx;
races_schedule_ccall( "races_swap_once", 300 );
races_schedule_ccall( "races_swap_once", 600 );
races_resolve_token_after( 3, 33, 1200 );
int v = races_await_token( 3 ); // parked for 1.2s, swaps land mid-park
ctx.Expect( v == 33, "long-parked sleep should resume with 33" );
CheckQuiescent( ctx, "after long-parked sleep" );
Finalize( "long_parked_sleep_clobbered_by_swap", std::move( ctx ) );
#else
CaseContext ctx;
Finalize( "long_parked_sleep_clobbered_by_swap", std::move( ctx ) );
#endif
}
// ----- scenario 6 (?only= single): wakeup_during_transition -------------------
// The KiCad "ENTER at state=2" family: a modal teardown arrives on a fresh
// stack while a token sleep is parked inside the modal's own pump, then the
// token resolves into the half-torn-down world. Closest deterministic analog
// of the clipboard-poll + EndModal collision.
void Scenario_WakeupDuringTransition()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "wakeup_during_transition", 5000 );
#endif
m_wdtCtx = std::make_unique<CaseContext>();
g_wdtBDone = false;
g_wdtBValue = 0;
m_pendingScenario = [this]() { RunWdtInsidePump(); };
m_scenarioTimer.StartOnce( 40 );
RacesDialog dlg( this, "wdt" );
g_activeModal = &dlg;
int result = dlg.ShowModal();
g_activeModal = nullptr;
m_wdtCtx->Expect( result == wxID_OK, "wdt modal should return wxID_OK" );
// The B-side sleep resolves after the modal is gone.
m_pollPredicate = []() { return g_wdtBDone; };
m_pollBudgetMs = 3000;
m_onPollDone = [this]( bool aOk ) {
m_wdtCtx->Expect( aOk, "B-side sleep should resume after modal teardown" );
m_wdtCtx->Expect( g_wdtBValue == 2, "B-side value should be 2" );
CheckQuiescent( *m_wdtCtx, "after wakeup-during-transition" );
Finalize( "wakeup_during_transition", std::move( *m_wdtCtx ) );
m_wdtCtx.reset();
};
m_pollTimer.Start( 50 );
}
void RunWdtInsidePump()
{
#ifdef __EMSCRIPTEN__
// Park a fresh stack on token 11 (B side) at +0ms - it outlives the modal.
races_schedule_ccall( "races_wdt_park_b", 0 );
// Tear the modal down from a fresh stack at +200ms (while B is parked
// AND this pump-tick stack is parked on token 10 below).
races_schedule_ccall( "races_end_active_modal", 200 );
// Resolve THIS stack's token at +300ms (after the modal teardown began)
// and B's at +350ms - both land in the post-teardown turbulence.
races_resolve_token_after( 10, 1, 300 );
races_resolve_token_after( 11, 2, 350 );
int a = races_await_token( 10 ); // parks this pump-tick stack
m_wdtCtx->Expect( a == 1, "A-side token should resolve to 1" );
// Immediately extend the in-flight window with a fiber swap cycle.
TestCoroutine co( []( TestCoroutine& self ) { self.Yield( 5 ); } );
co.Call( 1 );
bool running = co.Resume( 2 );
m_wdtCtx->Expect( !running, "post-wake fiber should finish" );
#endif
}
// ----- scenario 7 (?only= single): nested_quasi_modal_pump_error --------------
// c27fe8bf's wxWasmRunNestedLoop pump catches a ProcessEvents rejection and
// stops pumping WITHOUT resolving its promise: the nested DoRun stays parked
// forever (silent stall). Red until the wx-layer resolve-on-error fix.
void Scenario_NestedPumpError()
{
#ifdef __EMSCRIPTEN__
races_arm_watchdog( "nested_quasi_modal_pump_error", 3000 );
CaseContext ctx;
// Queue the bomb as a PENDING EVENT: the nested pump's ProcessEvents ->
// ProcessPendingEvents dispatches it, so the JS error propagates out of
// the pump's awaited ccall and rejects it. (A wx timer would NOT work:
// wasm timers fire via emscripten_async_call/callUserCallback and
// bypass the pump entirely.)
CallAfter( []() { races_throw_js_error(); } );
wxGUIEventLoop nestedLoop;
LogLine( "[ASYNCIFY_RACES] entering nested quasi-modal loop" );
nestedLoop.Run(); // wxWasmRunNestedLoop parks here
LogLine( "[ASYNCIFY_RACES] nested loop returned" );
ctx.Expect( true, "" ); // reaching this line at all is the fix
CheckQuiescent( ctx, "after nested pump error" );
Finalize( "nested_quasi_modal_pump_error", std::move( ctx ) );
#else
CaseContext ctx;
Finalize( "nested_quasi_modal_pump_error", std::move( ctx ) );
#endif
}
// ----- scenario 8 (mode=sleep-park): unwind_through_promise -------------------
// OnInit made the LAST pre-park suspension a sleep, so the park throw escaped
// through that sleep's wakeUp promise reaction. The spec asserts no "unwind"
// reaches pageerror/console; this C++ side just proves the app stayed alive.
void Scenario_UnwindThroughPromise()
{
CaseContext ctx;
// A post-park fiber swap doubles as a liveness check in this mode too.
TestCoroutine co( []( TestCoroutine& self ) { self.Yield( 77 ); } );
bool running = co.Call( 1 );
ctx.Expect( running && co.LastReturnValue() == 77, "post-park fiber should work" );
co.Resume( 2 );
CheckQuiescent( ctx, "after sleep-park startup" );
Finalize( "unwind_through_promise", std::move( ctx ) );
}
// ----- timers -----
void OnScenarioTimer( wxTimerEvent& )
{
if( m_pendingScenario )
{
auto scenario = std::move( m_pendingScenario );
m_pendingScenario = nullptr;
scenario();
}
}
void OnPollTimer( wxTimerEvent& )
{
if( !m_pollPredicate )
{
m_pollTimer.Stop();
return;
}
m_pollBudgetMs -= 50;
bool ok = m_pollPredicate();
if( ok || m_pollBudgetMs <= 0 )
{
m_pollTimer.Stop();
m_pollPredicate = nullptr;
auto done = std::move( m_onPollDone );
m_onPollDone = nullptr;
if( done )
done( ok );
}
}
private:
std::string m_only;
bool m_sleepParkMode;
std::size_t m_nextIndex = 0;
int m_total = 0;
int m_passed = 0;
wxTimer m_scenarioTimer;
std::function<void()> m_pendingScenario;
wxTimer m_pollTimer;
std::function<bool()> m_pollPredicate;
std::function<void( bool )> m_onPollDone;
int m_pollBudgetMs = 0;
wxDialog* m_dlgA = nullptr;
wxDialog* m_dlgB = nullptr;
wxDialog* m_dlgC = nullptr;
std::unique_ptr<CaseContext> m_tripleCtx;
std::vector<std::string> m_tripleSeq;
std::unique_ptr<CaseContext> m_sifimCtx;
std::unique_ptr<CaseContext> m_oooCtx;
std::vector<std::string> m_oooSeqStore;
std::unique_ptr<CaseContext> m_wdtCtx;
wxStaticText* m_summary = nullptr;
};
class RacesApp : public wxApp
{
public:
bool OnInit() override
{
std::string only;
bool sleepPark = false;
#ifdef __EMSCRIPTEN__
// Params travel in the URL HASH (#only=...&mode=...), not the query:
// `npx serve` cleanUrls-redirects *.html and drops the query string on
// the way. The hash never reaches the server. (Query kept as fallback.)
char onlyBuf[64] = { 0 };
EM_ASM( {
try {
var p = new URLSearchParams( ( location.hash || "" ).replace( /^#/, "" ) );
var v = p.get( 'only' ) || new URLSearchParams( location.search ).get( 'only' ) || "";
stringToUTF8( v.slice( 0, 63 ), $0, 64 );
} catch( e ) {}
}, onlyBuf );
only = onlyBuf;
sleepPark = EM_ASM_INT( {
try {
var p = new URLSearchParams( ( location.hash || "" ).replace( /^#/, "" ) );
var m = p.get( 'mode' ) || new URLSearchParams( location.search ).get( 'mode' );
return ( m === 'sleep-park' ) ? 1 : 0;
} catch( e ) { return 0; }
} ) == 1;
LogLine( "[ASYNCIFY_RACES] PARAMS only='" + only + "' sleepPark="
+ std::to_string( sleepPark ? 1 : 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
// JSPI: the ccall'd test levers are promising entries (JSPI_EXPORTS)
// that can PARK — the shim must track their activations like the wx
// entries, or their windows leak (untracked completion wedges the
// resume turnstile until the watchdog clears it).
EM_ASM( {
if( globalThis.__wxScheduler && globalThis.__wxScheduler.installExportWraps )
globalThis.__wxScheduler.installExportWraps(
[ 'races_swap_once', 'races_park_token2', 'races_wdt_park_b' ] );
} );
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
#endif
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
// THE LOAD-BEARING TOPOLOGY: complete a coroutine swap cycle during
// OnInit, before the main loop parks (exactly what KiCad's startup
// tool burst does). Historically this put main() inside the asyncify
// fiber trampoline when the park throw tore through it.
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
{
TestCoroutine co( []( TestCoroutine& self ) { self.Yield( 1 ); } );
co.Call( 1 );
co.Resume( 2 );
LogLine( "[ASYNCIFY_RACES] PRE-PARK-SWAP done" );
}
#ifdef __EMSCRIPTEN__
if( sleepPark )
{
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
// Make the LAST pre-park suspension a sleep: main then reaches the
// park out of a sleep resume rather than a swap (historically the
// park throw escaped through the sleep's wakeUp promise reaction).
test(asyncify): red-green race harness + ablation flags, unwind-catch shim, spec tightening, decisions docs The asyncify single-slot work, executed red-green (full ledger: docs/features/asyncify-arbiter/redgreen.md; decisions record: docs/features/async/07-decisions-and-outcome.md): - tests/apps/standalone/asyncify-races/ + tests/asyncify/ + dedicated playwright config: 8 scenarios reproducing the KiCad asyncify failure family with the kicad-faithful startup topology (pre-park fiber swap → park throw through the live trampoline). Built in 3 variants; the SHIM_DISABLE_TRAMPOLINE_HEAL / SHIM_DISABLE_HANDLESLEEP ablation builds keep the historical hang and index-out-of-bounds crash reproducible forever (mutation-style pins for the existing shims). - scripts/common/shims/handlesleep.js: catch the "unwind" park sentinel in the wakeUp path — when main's last pre-park suspension was a sleep, the main-loop park throw escaped through that sleep's promise reaction as an uncaught rejection (the calculator/gerbview console errors). - scripts/common/inject-dyncall-shims.sh: SHIM_DISABLE_* ablation knobs. - Spec tightening (the acceptance bar): 'uncaught exception: unwind' tolerance DELETED from pcbnew/eeschema specs; load-pcb gained a hard clean-console gate over 5 asyncify corruption signatures. - wxwidgets pointer bump: modal LIFO resolvers, pump resolve-on-error, sync clipboard IsSupported (014f67e6c1). Final state: asyncify suite 7/7, wx e2e 291/292 (1 skip), KiCad e2e 40 passed / 2 skipped with ZERO corruption signatures in any log across all six apps. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 16:59:07 +02:00
races_sleep_ms( 30 );
LogLine( "[ASYNCIFY_RACES] PRE-PARK-SLEEP done (sleep-park mode)" );
}
#endif
RacesFrame* frame = new RacesFrame( only, sleepPark );
frame->Show();
return true;
}
};
wxIMPLEMENT_APP( RacesApp );