AI_Assistant/pcbjam
Watch
1
0
Fork
You've already forked pcbjam
 0
Code Issues Pull requests Projects Releases 4 Packages Wiki Activity Actions 2

fix(asyncify): retire the deferral family — guard-layer road closed

Browse source 
v0.1.24 in prod, doubly convicted the same morning: the Leonardo open
crawled/hung (open:settled result=failed at the 60s escape, heap never past
256MB — every main-loop iteration runs INSIDE its yield-wake extent, so the
"root-owned wake" scope matched thousands of legitimate nested coroutine
Call/returns per open, each paying a deferred macrotask, throttled to ≥1s in
a background tab), AND the Nano crashed 22ms after deferrals=1 fired.
Harmful and insufficient: the fatal nested-rewind interleave and the benign
bulk are observationally identical at this layer — no discriminator exists.

Retired (second and final retraction, async/16 round 5). What stays shipped
and clean: consume-once root suspensions, the internally-parked quarantine +
laundering check, the flight recorder + beacons, the WSOD floor, the
pendingSleeps leak fix (confirmed by pendingSleeps=[] in the Nano dump). The
rare nested-rewind crash is ACCEPTED and fully observable until the
structural fix — the design-B fiber-first runtime (async/06,12,13), where
one scheduler owns every suspension and this interleave cannot exist.

.ci-cache-epoch 6→7.

Local: fiber 2/2 (one refusal beacon) + timer + firefox sweep 21 passed,
chromium scenarios 11 passed/4 quarantine-skips, web fatal+follow 2/2.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019SE4o46Lnq3hF574FFq8x4
This commit is contained in:
Gergő Törcsvári 2026-08-02 08:18:08 +02:00
commit f734d700a2
No known key found for this signature in database
GPG key ID: 8E75F2CDE64E5322
3 changed files with 37 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

24
docs/features/async/16-fiber-resume-guard.md
View file

@ -182,6 +182,30 @@ wake is live (`Asyncify.__wakingRoot`), with the
pushes sleep contexts (the v0.1.23 dump carried ~380 leaked zero-linked
entries).
## Round 5 (2026-08-02) — the scoped deferral hangs opens: RETIRED, road closed
v0.1.24 in prod: `open:settled result=failed` at the 60 s escape, heap never
past 256 MB, no traps, no beacons — the open crawled and (in a
timer-throttled background tab) effectively hung. Cause: the main loop's
every iteration runs INSIDE its yield-wake's synchronous extent (the v0.1.23
recording shows `sleep … w=1` on every tick), so "root re-entry during a
root-owned wake" matches every legitimate nested coroutine Call/return in a
board open — thousands per load, each paying a deferred macrotask.
**Verdict on the whole deferral family:** the fatal nested-rewind interleave
and the benign nested round-trips share the same observable signature at the
JS runtime layer. No discriminator exists here — this road is closed. What
remains shipped and sound: the consume-once validity + internally-parked
quarantine (never implicated in a regression), the flight recorder, the
beacons, and the WSOD floor. The rare nested-rewind crash is ACCEPTED, fully
observable, until the structural fix.
**The structural fix is design B** (06/12/13-design-b-*): one fiber-first
scheduler owning every suspension, so dual-protocol nesting cannot exist.
That is the next real investment; guard-layer iteration has hit its ceiling
— five variants, each defeated by a neighboring interleave or by taxing the
benign bulk.
## Flight recorder (round 3, targeting instrument)
The shim keeps a 96-entry ring of asyncify/fiber events (sleep entries with