2026-06-29 19:50:18 +02:00
|
|
|
/*
|
|
|
|
|
* nanosleep_yield.c — make a MAIN-THREAD nanosleep() YIELD to the JS event loop via
|
|
|
|
|
* Asyncify instead of busy-spinning, so on-demand pthread-Worker creation can complete
|
|
|
|
|
* WITHOUT editing KiCad.
|
|
|
|
|
*
|
|
|
|
|
* THE PROBLEM (the "non-warm thread" deadlock): once KiCad's thread pool has consumed all
|
|
|
|
|
* the pre-warmed Workers (-sPTHREAD_POOL_SIZE), a later raw std::thread (the raytracer)
|
|
|
|
|
* must spawn a NEW Worker on demand. Finalizing it needs the main thread's event loop to
|
|
|
|
|
* run the new-Worker 'loaded' -> 'run' handshake. KiCad's join is a sleep_for() busy-wait
|
|
|
|
|
* -> nanosleep -> emscripten_thread_sleep, which busy-spins and NEVER returns to the JS
|
|
|
|
|
* event loop, so the new Worker never starts -> deadlock.
|
|
|
|
|
*
|
|
|
|
|
* THE FIX: the only main-thread primitive that returns to the event loop is an Asyncify
|
|
|
|
|
* unwind (emscripten_sleep). This provides a nanosleep that, ON THE MAIN THREAD, yields via
|
|
|
|
|
* an EM_ASYNC_JS await (= emscripten_sleep semantics; __asyncjs__* is already in the
|
|
|
|
|
* post-link asyncify-imports). The unmodified sleep_for busy-wait then pumps the loop, the
|
|
|
|
|
* Worker handshake completes, and on-demand creation works with no KiCad edit.
|
|
|
|
|
*
|
|
|
|
|
* MECHANISM: a STRONG definition of nanosleep here SHADOWS musl's archive member — the
|
|
|
|
|
* linker only pulls musl's nanosleep.o if the symbol is left undefined, and ours defines it.
|
|
|
|
|
* (-Wl,--wrap=nanosleep is not an option here — it crashes wasm-ld with a SIGSEGV in
|
|
|
|
|
* lld::wasm::ImportSection::addImport.) On a pthread worker we fall back to
|
|
|
|
|
* emscripten_thread_sleep (the real underlying blocking sleep — workers may block).
|
|
|
|
|
*
|
|
|
|
|
* SCOPE: only the main browser thread yields; only it must never block the event loop.
|
2026-08-13 09:04:43 +02:00
|
|
|
*
|
|
|
|
|
* ZERO-DURATION GUARD (ported from staging 29c61b8): a nanosleep of 0 ms
|
|
|
|
|
* returns immediately WITHOUT yielding. mimalloc's spin-wait hint
|
|
|
|
|
* mi_atomic_yield() is sleep(0) on wasm, reached from malloc's slow path
|
|
|
|
|
* under cross-thread delayed-free contention. Under JSPI a yield there
|
|
|
|
|
* SUSPENDS the activation inside the allocator — any other activation that
|
|
|
|
|
* then runs can re-enter mimalloc mid-operation. A zero-duration sleep never
|
|
|
|
|
* promised an event-loop turn — stock emscripten busy-waits and returns
|
|
|
|
|
* immediately. mimalloc is the module's only zero-duration sleeper; the
|
|
|
|
|
* worker-boot deadlock this shim fixes needs only the ms-scale yields.
|
2026-06-29 19:50:18 +02:00
|
|
|
*/
|
|
|
|
|
#include <emscripten/emscripten.h>
|
|
|
|
|
#include <emscripten/threading.h>
|
|
|
|
|
#include <time.h>
|
|
|
|
|
|
|
|
|
|
/* EM_ASYNC_JS integrates with Asyncify automatically (binaryen instruments every caller). */
|
|
|
|
|
EM_ASYNC_JS( void, __wasm_main_thread_yield_ms, ( double ms ), {
|
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: route through the scheduler's turnstile when it exists - a raw
|
|
|
|
|
await's engine-level resume bypasses the shim's SP discipline and
|
|
|
|
|
leaves its window marked live (the pump then refuses every later
|
|
|
|
|
resume). Asyncify builds (no jspi scheduler) keep the raw await. */
|
|
|
|
|
var S = globalThis.__wxScheduler;
|
|
|
|
|
if( S && S.backend === 'jspi' ) {
|
|
|
|
|
await S.sleepYield( ms );
|
|
|
|
|
return;
|
|
|
|
|
}
|
2026-06-29 19:50:18 +02:00
|
|
|
await new Promise( function( resolve ) { setTimeout( resolve, ms ); } );
|
|
|
|
|
} );
|
|
|
|
|
|
Phase B: main-thread sleep parks its context; the real D blocker is DOM entries
wasm/shims/context_sleep.cpp: a main-thread nanosleep whose frame stands on a
scheduler context that OWNS the stack arms a mailbox wake and yield_parks that
context instead of suspending the stack in place. It lives in the sleep
primitive rather than in tool_manager.cpp on purpose - KiCad and the wx core
stay untouched (CLAUDE.md's fork rule) and the whole K7 class moves at once,
not just TOOL_MANAGER::RunSynchronousAction's spin loop.
MEASURED AT D-ON, and it is NOT what unblocks Phase D. The four canvas-tool
specs still fail, but the trace now names a different cause: the fatal swap is
old=<libcontext ROOT> new=<tool coroutine> with mouseEventHandlerFunc above it
- a DOM mouse handler entering wasm DIRECTLY on the main stack, bypassing the
tick. So one coroutine is entered two ways: by the tick through the dispatch
context as a STAR TRANSFER, and by DOM handlers as a DIRECT SYMMETRIC SWAP. A
capture written by one path cannot be rewound by the other -> index out of
bounds in doRewind. That is section 7 rule 5 (partial migration is worse than
none) in its purest measured form, and it is why the harness stays green: its
coroutines are only ever entered from one place.
So the next increment is the DOM event entries (mouse/key/wheel/resize must
hand their events to the dispatch context as the tick does), not another park
site. It subsumes the one-root work too: with no dispatch on the main stack,
resolve_root_identity() always answers "the running context".
Landing state: STAR_DISPATCH=0, kicad 139 passed / 1 (pre-existing occ-probe)
= baseline, with the sleep shim in and inert.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LBjomQfKyRa3jBdeAKpmTw
2026-08-07 20:48:35 +02:00
|
|
|
/*
|
|
|
|
|
* Scheduler-aware sleep (context_sleep.cpp, docs/features/async/22 Phase B):
|
|
|
|
|
* when this frame stands on a scheduler context that owns the stack, the wait
|
|
|
|
|
* PARKS THAT CONTEXT instead of suspending the stack in place. Returns 0 when
|
|
|
|
|
* no context owns the stack, and then the Asyncify yield below is still right.
|
|
|
|
|
*
|
|
|
|
|
* This is what makes TOOL_MANAGER::RunSynchronousAction's spin loop safe under
|
|
|
|
|
* Phase D: an in-place park inside a tool body leaves a capture in flight for a
|
|
|
|
|
* star transfer to land on (doRewind -> "index out of bounds").
|
2026-08-08 16:18:00 +02:00
|
|
|
*
|
|
|
|
|
* extern "C" + WEAK, both load-bearing: some app builds compile this file as
|
|
|
|
|
* C++ (em++ keys language off the DRIVER, not the .c extension — a bare extern
|
|
|
|
|
* here mangles and the app aborts at "missing function" on first sleep), and
|
|
|
|
|
* the standalone wx test apps do not link context_sleep.cpp at all. A weak
|
|
|
|
|
* null resolves to "no context lane here — always yield in place", which is
|
|
|
|
|
* exactly the pre-context behaviour those apps pin.
|
Phase B: main-thread sleep parks its context; the real D blocker is DOM entries
wasm/shims/context_sleep.cpp: a main-thread nanosleep whose frame stands on a
scheduler context that OWNS the stack arms a mailbox wake and yield_parks that
context instead of suspending the stack in place. It lives in the sleep
primitive rather than in tool_manager.cpp on purpose - KiCad and the wx core
stay untouched (CLAUDE.md's fork rule) and the whole K7 class moves at once,
not just TOOL_MANAGER::RunSynchronousAction's spin loop.
MEASURED AT D-ON, and it is NOT what unblocks Phase D. The four canvas-tool
specs still fail, but the trace now names a different cause: the fatal swap is
old=<libcontext ROOT> new=<tool coroutine> with mouseEventHandlerFunc above it
- a DOM mouse handler entering wasm DIRECTLY on the main stack, bypassing the
tick. So one coroutine is entered two ways: by the tick through the dispatch
context as a STAR TRANSFER, and by DOM handlers as a DIRECT SYMMETRIC SWAP. A
capture written by one path cannot be rewound by the other -> index out of
bounds in doRewind. That is section 7 rule 5 (partial migration is worse than
none) in its purest measured form, and it is why the harness stays green: its
coroutines are only ever entered from one place.
So the next increment is the DOM event entries (mouse/key/wheel/resize must
hand their events to the dispatch context as the tick does), not another park
site. It subsumes the one-root work too: with no dispatch on the main stack,
resolve_root_identity() always answers "the running context".
Landing state: STAR_DISPATCH=0, kicad 139 passed / 1 (pre-existing occ-probe)
= baseline, with the sleep shim in and inert.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LBjomQfKyRa3jBdeAKpmTw
2026-08-07 20:48:35 +02:00
|
|
|
*/
|
2026-08-08 16:18:00 +02:00
|
|
|
#ifdef __cplusplus
|
|
|
|
|
extern "C"
|
|
|
|
|
#endif
|
|
|
|
|
int pcbjam_context_sleep_ms( double ms ) __attribute__(( weak ));
|
Phase B: main-thread sleep parks its context; the real D blocker is DOM entries
wasm/shims/context_sleep.cpp: a main-thread nanosleep whose frame stands on a
scheduler context that OWNS the stack arms a mailbox wake and yield_parks that
context instead of suspending the stack in place. It lives in the sleep
primitive rather than in tool_manager.cpp on purpose - KiCad and the wx core
stay untouched (CLAUDE.md's fork rule) and the whole K7 class moves at once,
not just TOOL_MANAGER::RunSynchronousAction's spin loop.
MEASURED AT D-ON, and it is NOT what unblocks Phase D. The four canvas-tool
specs still fail, but the trace now names a different cause: the fatal swap is
old=<libcontext ROOT> new=<tool coroutine> with mouseEventHandlerFunc above it
- a DOM mouse handler entering wasm DIRECTLY on the main stack, bypassing the
tick. So one coroutine is entered two ways: by the tick through the dispatch
context as a STAR TRANSFER, and by DOM handlers as a DIRECT SYMMETRIC SWAP. A
capture written by one path cannot be rewound by the other -> index out of
bounds in doRewind. That is section 7 rule 5 (partial migration is worse than
none) in its purest measured form, and it is why the harness stays green: its
coroutines are only ever entered from one place.
So the next increment is the DOM event entries (mouse/key/wheel/resize must
hand their events to the dispatch context as the tick does), not another park
site. It subsumes the one-root work too: with no dispatch on the main stack,
resolve_root_identity() always answers "the running context".
Landing state: STAR_DISPATCH=0, kicad 139 passed / 1 (pre-existing occ-probe)
= baseline, with the sleep shim in and inert.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LBjomQfKyRa3jBdeAKpmTw
2026-08-07 20:48:35 +02:00
|
|
|
|
2026-06-29 19:50:18 +02:00
|
|
|
int nanosleep( const struct timespec* req, struct timespec* rem )
|
|
|
|
|
{
|
|
|
|
|
if( req )
|
|
|
|
|
{
|
|
|
|
|
double ms = (double) req->tv_sec * 1000.0 + (double) req->tv_nsec / 1.0e6;
|
2026-08-13 09:04:43 +02:00
|
|
|
if( ms > 0.0 )
|
Phase B: main-thread sleep parks its context; the real D blocker is DOM entries
wasm/shims/context_sleep.cpp: a main-thread nanosleep whose frame stands on a
scheduler context that OWNS the stack arms a mailbox wake and yield_parks that
context instead of suspending the stack in place. It lives in the sleep
primitive rather than in tool_manager.cpp on purpose - KiCad and the wx core
stay untouched (CLAUDE.md's fork rule) and the whole K7 class moves at once,
not just TOOL_MANAGER::RunSynchronousAction's spin loop.
MEASURED AT D-ON, and it is NOT what unblocks Phase D. The four canvas-tool
specs still fail, but the trace now names a different cause: the fatal swap is
old=<libcontext ROOT> new=<tool coroutine> with mouseEventHandlerFunc above it
- a DOM mouse handler entering wasm DIRECTLY on the main stack, bypassing the
tick. So one coroutine is entered two ways: by the tick through the dispatch
context as a STAR TRANSFER, and by DOM handlers as a DIRECT SYMMETRIC SWAP. A
capture written by one path cannot be rewound by the other -> index out of
bounds in doRewind. That is section 7 rule 5 (partial migration is worse than
none) in its purest measured form, and it is why the harness stays green: its
coroutines are only ever entered from one place.
So the next increment is the DOM event entries (mouse/key/wheel/resize must
hand their events to the dispatch context as the tick does), not another park
site. It subsumes the one-root work too: with no dispatch on the main stack,
resolve_root_identity() always answers "the running context".
Landing state: STAR_DISPATCH=0, kicad 139 passed / 1 (pre-existing occ-probe)
= baseline, with the sleep shim in and inert.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LBjomQfKyRa3jBdeAKpmTw
2026-08-07 20:48:35 +02:00
|
|
|
{
|
2026-08-13 09:04:43 +02:00
|
|
|
if( emscripten_is_main_runtime_thread() )
|
|
|
|
|
{
|
|
|
|
|
if( !pcbjam_context_sleep_ms || !pcbjam_context_sleep_ms( ms ) )
|
|
|
|
|
__wasm_main_thread_yield_ms( ms ); /* yield -> event loop runs -> Worker boots */
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
emscripten_thread_sleep( ms ); /* worker: real blocking sleep */
|
Phase B: main-thread sleep parks its context; the real D blocker is DOM entries
wasm/shims/context_sleep.cpp: a main-thread nanosleep whose frame stands on a
scheduler context that OWNS the stack arms a mailbox wake and yield_parks that
context instead of suspending the stack in place. It lives in the sleep
primitive rather than in tool_manager.cpp on purpose - KiCad and the wx core
stay untouched (CLAUDE.md's fork rule) and the whole K7 class moves at once,
not just TOOL_MANAGER::RunSynchronousAction's spin loop.
MEASURED AT D-ON, and it is NOT what unblocks Phase D. The four canvas-tool
specs still fail, but the trace now names a different cause: the fatal swap is
old=<libcontext ROOT> new=<tool coroutine> with mouseEventHandlerFunc above it
- a DOM mouse handler entering wasm DIRECTLY on the main stack, bypassing the
tick. So one coroutine is entered two ways: by the tick through the dispatch
context as a STAR TRANSFER, and by DOM handlers as a DIRECT SYMMETRIC SWAP. A
capture written by one path cannot be rewound by the other -> index out of
bounds in doRewind. That is section 7 rule 5 (partial migration is worse than
none) in its purest measured form, and it is why the harness stays green: its
coroutines are only ever entered from one place.
So the next increment is the DOM event entries (mouse/key/wheel/resize must
hand their events to the dispatch context as the tick does), not another park
site. It subsumes the one-root work too: with no dispatch on the main stack,
resolve_root_identity() always answers "the running context".
Landing state: STAR_DISPATCH=0, kicad 139 passed / 1 (pre-existing occ-probe)
= baseline, with the sleep shim in and inert.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LBjomQfKyRa3jBdeAKpmTw
2026-08-07 20:48:35 +02:00
|
|
|
}
|
2026-06-29 19:50:18 +02:00
|
|
|
}
|
|
|
|
|
if( rem )
|
|
|
|
|
{
|
|
|
|
|
rem->tv_sec = 0;
|
|
|
|
|
rem->tv_nsec = 0;
|
|
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
}
|