`sendReply` ignored `proto.write`'s `false` return, so a program resolving
several large values in one `asyncio.gather` round encoded every reply in the
same turn and queued all of them in fd 3's writable buffer. Binding resolution
carries no seam-level byte cap to bound that, and the failure kills the host
process rather than failing the run: measured on a 64 KiB-highWaterMark pipe,
eight 4 MiB replies buffered 32.0 MiB at once against 0.0 MiB once paced.
Replies now go through a queue that encodes and writes one frame at a time,
awaiting `drain` when the pipe is full. The encode happens inside the loop, so a
queued reply the run no longer needs is dropped by the `settled` check without
ever being serialized.
This was previously deferred on the grounds that serializing would narrow the
seam's concurrency contract. That reasoning was wrong: the child matches each
reply to its `call` by id from a pump that reads fd 3 continuously, so arrival
order was never observable, and the bindings still run concurrently. Only the
host's peak memory and the flush timing change. The README entry recording the
deferral is removed and the Agent Note records the mechanism instead.
`sendReply` already refuses to write after the run settled, but only after
`snapshotJsonValue` walked and copied the resolution. Binding resolution carries
no seam-level byte cap, so a binding resolving a wide value after `maxWallMs`,
an abort, or dispose settled the run spent host heap building a frame that was
then discarded. The check moves ahead of the snapshot.
Also in this change:
- `readProcessStart` moved after `messageOf`. Inserting it between `messageOf`'s
JSDoc and its body left that function undocumented and the orphaned block
reading as a second doc for the reader; `verify-export-jsdoc` does not catch it
because `messageOf` is not exported.
- The README pair adds the disposed-runtime rejection to `run()`'s public
contract, which `src/index.ts` has enforced all along.
- Known Limitations records three deferred constraints that until now existed
only in review discussion: the combined log-and-value peak the load gate does
not model, the host-side per-member expansion of a wide binding reply (owned by
`packages/core/session`, and shared with the worker-thread backend), and the
absence of fd-3 backpressure for concurrent replies.
- The Agent Note's same-group section records the teardown identity guard and its
two rulings, including why an ABSENT start-time reading proceeds rather than
withholding the signal, and that reading it as a mismatch is what turned the
three same-group heartbeat cases red on Linux.
The directive carried its whole justification inline at 203 characters, past the
140 the @stylistic/max-len rule allows (imports and template-literal messages
are exempt; a line comment is not). The reasoning moves to the lines above and
the directive keeps a short pointer, since a v8 ignore must stay on one line.
The PID-reuse guard has two arms no single OS can execute: the non-Linux early
return in readProcessStart (the Linux coverage lane always takes the read path)
and the refusal arm, which needs a real pid recycled into a new group leader
between spawn and teardown -- no test can schedule that. The coverage lane
reported 99.53% statements / 99.14% branches on src/index.ts for exactly these
two.
Both carry a v8 ignore naming what cannot be reached and why, the convention
this file and subprocess-local already use for platform defenses. The reader
itself stays covered by the process-identity test rather than being exempted
wholesale.
The PID-reuse guard refused to signal whenever the current reading differed
from the one taken at spawn, including when it was ABSENT. On Linux a reaped
leader has no /proc/<pid>/stat, so every teardown after the leader exited
skipped SIGTERM/SIGKILL while the group it led still held survivors -- the
exact case the process-group teardown exists to reap. Three same-group survivor
tests went red on the coverage lane; they pass on Darwin because the reader
always returns undefined there, leaving the guard inert.
Only a present-and-different reading now blocks the signal. Verified on the
self-hosted Linux box: a reaped leader with live survivors allows the signal, a
pid whose start time differs still blocks it, and a live matching process is
signalled.
The README pair described `run()` as rejecting "a malformed binding namespace or
non-positive config", which understated and misplaced the configuration
failures: a non-Unix platform, a non-integer budget, a timer value setTimeout
would clamp, a budget larger than one fd-3 frame, and an incompatible
addressSpaceMb/output-budget pair all throw from the CONSTRUCTOR, so they fail
when the plugin loads rather than on a later run. Both sides now separate the
load-time platform/configuration errors from the run-result contract.
The Chinese README's Model Experience and KV Cache effect sections were still
untranslated English; the pairing record only tracks hashes, so it could not
show that. Both are now translated.
Adding a published Python backend and reordering `flush_line` left several
owning documents stating things that are no longer true.
`src/invariant.ts` justified its empty installer with "ships only the fd-3
wire-protocol codec", which the subprocess execution path contradicts. The
reason now states the actual one: every relation this backend maintains lives
in the CPython child or on the fd-3 wire, so no same-process event sequence is
observable from a listener -- the same shape the sibling worker-thread backend
uses.
The seam's `PORTABLE_RESERVED_WORDS` and `language` JSDoc, the code-runtime
README pair, and docs/subsystems/code-runtime both said only TypeScript has a
published backend. Corrected in all four, with the generated cordis catalog
regenerated for the `language` change.
The note attributed the 12x multiple to the settlement flush holding three
copies. That stopped being true when `flush_line` was reordered to drop the
pending chunks before its push: the binding worst case is the newline path's
single near-budget write. Corrected in the note (both sides) and in the test
comment that repeated it.
The note's Testing section now registers the cases this stack added, and the
Chinese side receives the O(depth) entry it never got plus the new ones -- it
had drifted from the English.
`INTERPRETER_BASELINE_BYTES` argued 64 MiB from a RESIDENT set while RLIMIT_AS
bounds address space. It now cites the bootstrap's own measurement (30.23 MiB
of mappings for `python3 -I`), making 64 MiB roughly twice the measured
baseline.
Also: a hardcoded `(:232-235)` comment reference becomes a reference by name,
a "which now walks in O(depth) too" change narrative becomes a current-state
statement, and a stray double blank line is removed.
Four independent corrections in the run lifecycle.
`killGroup` signalled `-child.pid` with a raw `process.kill`. Node keeps the
numeric `child.pid` after the leader is reaped and only clears its internal
handle, so `child.kill()` refuses while the raw call does not; `close` can
trail `exit` by seconds when a pipe-holding descendant keeps the streams open.
A recycled pgid could therefore receive this run's SIGTERM and armed SIGKILL.
`groupEmpty()` does not cover it: it reports whether the group has members, not
whether they are ours, and it first runs after the signal. The leader's start
time is now read at spawn and re-checked before each signal, matching the
position packages/subprocess/subprocess-local already states
("ProcessIdentity ... preventing teardown escalation after PID reuse"). Kept
local rather than depending on that package, which would add an architectural
edge. Linux reads /proc; Darwin has no /proc, so the reader reports undefined
and the guard degrades to the previous behavior instead of forking `ps` on a
teardown path.
`_push_bounded_prefix` built `(*self._pending, extra)`, copying every pending
reference into a same-size tuple before the bounded loop. For a
single-character drip that is a second pointer array as large as the list:
measured +80 MiB of tuple over a 40 MiB list for 5.2M chunks, the allocation
the bounded prefix exists to avoid. It now iterates the list in place and
handles `extra` in the loop's `else`; 4000 randomized inputs produce byte-identical
prefixes.
The settlement `flush_out()`/`flush_err()` ran outside any guard while `done`
was already decided, so a flush raising under memory pressure skipped
`send_done` and downgraded a child-classified `exception` into a host-side
`worker-exit`. Both are now wrapped, swallowing only the log tail.
The boot re-check's `if effective_soft != RLIM_INFINITY` was dead: `_clamped`
is asked for a finite `addr_bytes` on both sides and each branch returns that
value or a `min` with an inherited bound, so RLIM_INFINITY is unreachable. The
guard could only ever have skipped the re-check it claimed to protect.
Three separate paths in the CPython child allocated state proportional to a
value's width or a string's length, so a legitimate input the byte budgets
admit could die as the program's own MemoryError.
`_lossless_json_violation` enqueued one traversal tuple per member while
running, in `dispatch`, over MODEL-CONSTRUCTED binding arguments that no
child-side byte budget bounds first. It now uses the same (kind, container,
iterator) cursor the other two walks already had, checking dict keys as the
cursor pulls each entry. Measured over `[0] * 6_000_000` (~17 MB of JSON):
459.1 MiB of traversal tuples before, 0.0 MiB after.
`_decode_json_plain` matched JSON strings with a `(?:[^"\\]|\\.)*` repetition,
which makes CPython's engine retain backtracking state proportional to the
string's width: 146 MiB for a 1 MiB string, 557.8 MiB for 4 MiB. A legitimate
multi-megabyte binding reply raised MemoryError inside `_pump_replies`, and
because that pump is the only settler of the call's future, the run stranded
until the wall clock reported `timeout`. Strings now scan chunk-to-chunk over a
character class, which the engine matches without backtracking state; the same
4 MiB decode peaks at the 4.0 MiB result.
`_check_done_value` charged strings and dict keys what
`_dump_string(...).encode()` returned, building the escaped copy plus its
encode to MEASURE it -- ~6x the original each for control-heavy text, so
metering a value the budget then rejects could itself breach RLIMIT_AS and
report `exception` where the seam promises `output-limit`. The new
`_json_str_cost` counts instead, reusing `_json_string_cost`'s C-level passes
and reproducing `_dump_string`'s exact surrogate rules (fold spelled-out pairs,
charge six ASCII bytes per lone surrogate). Identical values, 228.9 MiB -> 19.1
MiB of peak on a 20M-NUL string.
Each fix ships a regression test. The two RLIMIT_AS repros are Linux-only:
Darwin does not apply the limit, so the peaks above are measured directly and
recorded in the test comments.
The O(depth) wide-value regression test ran under `maxWallMs: 20_000`, but the
cursor pulls 6M elements one at a time through Python-level frames: ~11s on an
idle machine, and more under the coverage lane's V8 instrumentation with several
workers sharing a runner. CI reported `timeout` instead of the round-trip.
Raise the run's ceiling to 60s inside a 90s vitest timeout, so the runtime's own
wall clock still fires first on a genuine hang. The assertion is unchanged and
still discriminates: restoring the O(width) `stack.extend` enqueue fails the test
with a child-side MemoryError in ~2.6s.
`_check_done_value` and `_encode_json_plain` pushed one stack entry per child
(plus a separator marker, and `dict.items()` materialized as a list), so the
bookkeeping scaled with the value's WIDTH rather than its depth. A value the
byte meter admits could then die on the walk's own frames: a flat
`[0] * 2_000_000` serializes to 4.0 MB, but measured peaks were 145.2 MB in the
meter and 114.7 MB in the encoder — 28.7x the serialized size, far past the 12x
the load-time address-space gate reserves.
Each container now pushes ONE cursor frame that pulls its children one at a
time and writes into a shared `io.StringIO`, so the output string is the only
width-proportional allocation and the caller already metered its size. Measured
on the same value: 0.0 MB in the meter and 9.0 MB in the encoder (2.3x), with
identical verdicts.
The load gate bounds maxLogBytes and maxValueBytes independently against the
address space, but the child framed the completion value (materializing its
escaped form to meter it, then encoding the frame) while a newline-free log tail
still sat unflushed in _pending. Those two peaks added, so two budgets each
admitted alone could together breach RLIMIT_AS and die as worker-exit instead of
settling. The success path now flushes both log streams before _done_with_value
runs; the trailing flush stays for the exception path and is an idempotent no-op
after a successful settle. A combined-peak regression test (32 MiB each against
512 MiB) asserts the over-budget value reports output-limit rather than OOMing.
Also corrects the worst-case-multiple JSDoc and Agent Note: after 1088d6f03d
made flush_line drop pending before its push, the settlement-flush path holds
two copies, not three, so the newline path is the sole 12x worst case. The
reorder is recorded as a called-out untested fix (the 12x gate already admits
only configs safe under both flush orders).
The load-time output-budget/addressSpaceMb gate used a worst-case multiple of 8,
assuming two simultaneous ~4x astral copies (the built string and its encode).
Three are live at the peak: on the newline path a single write holds the caller's
text argument, the line slice handed to push, and push's encode copy; the
settlement flush_line path held the pending chunks, their join, and that encode
copy. A budget admitted at 8x (e.g. maxLogBytes 48 MiB against addressSpaceMb 512)
could still OOM the child. The multiple is now 12, the strict `>` is `>=` so a
budget whose peak exactly equals the room left after the interpreter baseline is
rejected (that peak plus the baseline is the whole address space), and flush_line
drops the pending chunks before its push to match the newline path's
join-clear-push order. The child re-check mirror and both note sides move in step;
config-catalog is regenerated from the updated field JSDoc.
The boot re-check raises inside bootstrap's setrlimit-phase handler, which
classifies every resource-limit-application failure as kind 'exception'. The
test asserted 'worker-exit'; align it to the actual class and keep the message
assertion so the case still discriminates a config rejection from a generic
setrlimit error. The Agent Note's two references to the reported kind are
corrected on both language sides and the pair re-recorded.
The output-budget/address-space gate's 8x multiple had no room for the
interpreter's own footprint, so a budget sized right at addressSpaceMb/8 was
admitted while its worst-case peak plus the interpreter overran RLIMIT_AS
(e.g. 15 MiB maxLogBytes against 128 MiB). Reserve a fixed
INTERPRETER_BASELINE_BYTES (64 MiB) before the multiple claims the rest, so each
budget times 8 must fit the room LEFT after the baseline.
The host gate validates against the CONFIGURED addressSpaceMb, but a launch
environment can inherit a stricter RLIMIT_AS (a ulimit -v wrapper below
addressSpaceMb) that _clamped lowers the effective limit to, leaving the budgets
sized for a ceiling the child never gets. bootstrap.py now re-checks both budgets
against the effective clamped soft limit after applying it, mirroring the host
gate's multiple and baseline, and raises at boot rather than letting a
near-budget output OOM mid-run.
Add regression tests for both (the load gate against a 256 MiB address space
covering both budgets, and a ulimit -v wrapper for the inherited-limit re-check);
register the tail-copy test in the note Testing section; sync the zh pair. Merges
origin/feat/code-runtime-python-protocol to resolve the DIRTY base.
The tail-copy regression built `"first\n" + "A" * 200 MiB`, whose construction
alone peaks near 400 MiB (the string plus the concat temporary) and OOMs under
the 384 MiB addressSpaceMb before the log path under test runs — a MemoryError in
the model, not the defect. Build the tail in a variable and concatenate only the
newline (peak ~2x150 MiB = 300 MiB, under the address space), so the model's own
allocation fits; the pre-fix code then buffered the whole 150 MiB tail again,
pushing past 384 MiB, while the sliced prefix does not.
The load-time addressSpaceMb gate used a 1/8 fraction derived for ASCII, but the
child ledgers trigger on character count against a serialized-byte budget: an
astral character is one character yet ~4 bytes stored and ~4 encoded, live at
once, so the true worst-case peak is ~8x the budget, not ~2x. Replace the
fraction with an explicit OUTPUT_BUDGET_WORST_CASE_ADDRESS_SPACE_MULTIPLE (8)
and a strict `>`, and gate maxValueBytes the same way as maxLogBytes — the value
path builds and encodes a near-budget completion under the same RLIMIT_AS, so
the incompatible pair was previously admitted there too.
Slice the newline branch's unterminated tail to a budget-sized prefix: it
buffered the whole text[pos:] before the flush trigger could bound it, so an
early newline plus a huge tail made a second full copy of the model's string —
an RLIMIT_AS death the config gate cannot cover since the tail can far exceed
maxLogBytes.
Disclose the cross-field constraint in the maxLogBytes/maxValueBytes/addressSpaceMb
JSDoc (regenerating config-catalog); refresh the note's stale
Buffer.byteLength(JSON.stringify) reference; reconcile the arrival-order rebuttal
with the seam's "in order" logs JSDoc (within-stream, cross-stream best-effort).
Extend the load-rejection test to both budgets and add a tail-copy regression;
sync the zh pair.
The child log ledger encodes an admitted entry to UTF-8 once to charge its
serialized cost, so a maxLogBytes approaching addressSpaceMb lets a legitimate
near-budget log entry breach RLIMIT_AS and die as worker-exit instead of
truncating. Two runtime fixes were tried and both traded one resource bound for
another: an exact serialized-cost check is either a full encode (the allocation
being avoided) or a per-character Python loop that burns the CPU budget (a 10 MB
write hits SIGXCPU under cpuSeconds:1). The breach is a property of the
maxLogBytes/addressSpaceMb pair, not any write, so reject the incompatible pair
at load — maxLogBytes must stay within one eighth of the addressSpaceMb byte
count — and revert _LogStream to its original character-count buffering, which
is memory-safe once the budget fits the address space. The check runs on every
platform since the incompatibility is a config-value property, not a runtime one.
Replace the child-flood regression tests (which asserted the reverted runtime
behavior) with a load-rejection test. The host-side accrueStrayCost UTF-8
per-lead validation and its tests are unaffected. Update the note and zh pair.
accrueStrayCost accepted any 0x80-0xBF continuation, so a CESU-8 surrogate
(ED A0 80) or overlong (E0 80 80) — structurally well-formed but illegal, and
as cheap to flood as 0xFF — was charged its structural width 3 while
toString('utf8') renders each byte as its own U+FFFD (cost 9). Validate each
lead's first-continuation range (WHATWG E0/ED/F0/F4 bounds) and charge 3 per
byte of any sequence outside it, folding a broken prefix to one U+FFFD.
The child _LogStream newline path had the same char-vs-serialized gap the
newline-free trigger had: its per-line fit checks (first reconstructed line and
each subsequent line) compared character count against the serialized-byte
budget, so a control-char line passed and _logs.push encoded it whole, breaching
RLIMIT_AS. Route every check through _fragment_cost_upto, which sums per-char
costs from _json_char_cost over a start/end sub-range without slicing or
encoding and stops at the budget.
Decline arrival-order stray flushing: the two pipes' data events interleave
nondeterministically and logs carries no cross-pipe ordering guarantee, so a
fixed drain order is as valid as any and an arrival-tick branch could not be
covered without a flaky test.
Add CESU-8/overlong, newline-path-flood, and all-lead-class reassembly
regression tests; fix the note's now-inaccurate CESU/illegal-byte claims and a
fixture byte-count comment; sync the zh pair.
The prior child-flush fix measured each fragment with chunk.encode('utf-8'),
which copies the whole write — under a tight addressSpaceMb a single 340 MiB
write died on that encode (the exact allocation _push_bounded_prefix exists to
avoid), and re-scanning the whole pending list per write was quadratic under a
daemon-thread flood (the concurrent-write test timed out at 28s). Compute each
fragment's serialized cost with _fragment_cost_upto, which walks the str via a
new _json_char_cost (code point to escaped width, no encode) and stops once the
running total passes the budget, and accumulate it into _pending_cost once per
write. The early-flush trigger reads that accumulator: still charges control
chars their full serialized width (a NUL is 6 bytes), but never encodes a whole
write and never re-scans the buffer, so the 340 MiB single-write and
daemon-thread tests pass alongside the NUL-flood one.
Rework the child NUL-flood regression to write in 1 MiB chunks under a 512 MiB
address space so its own argument construction is not the allocation under test.
The child-log-flood regression built one 30M-char argument string, which under
the 64 MB addressSpaceMb died on RLIMIT_AS during construction (exit 120) before
the flush trigger under test could run, so it failed on Linux CI. Write the
flood in 1 MiB chunks under a 512 MiB address space instead: the argument str is
never itself the allocation under test, the fixed serialized-cost trigger keeps
the pending tail bounded to a few MiB, and the run completes at the marker; the
pre-fix char-count trigger accumulates the whole ~200 MiB and its ~1.2 GiB
settlement encode breaches RLIMIT_AS. Mirrors the addressSpaceMb budget the
existing oversized-completion tests use.
The host stray-capture cost function charged illegal UTF-8 bytes (0x80-0xC1,
0xF5-0xFF, and orphaned multibyte leads) the raw 1, but toString('utf8')
renders each as U+FFFD (3 serialized bytes). A b"\xff" flood was undercounted
threefold, so the residual grew to a full budget's worth of raw bytes before
flushing and, near a large maxLogBytes, expanded toward a ~1 GiB peak in the
flush's concat plus toString. Replace serializedBufferCost with accrueStrayCost,
a cross-chunk UTF-8 walker that charges each byte its decoded serialized width;
carry its sequence state on each StrayBuffer.
The child _LogStream had the same-family bug: its early-flush trigger compared
_pending_chars (character count) against remaining (a serialized-byte budget),
so a 30M-NUL newline-free flood stayed under a 50 MB char trigger yet encoded to
~180 MB at settlement, breaching RLIMIT_AS as worker-exit. Track _pending_cost
via the _JSON_BYTE_COST table and trigger on it; keep _pending_chars for the
char-based slice bounds.
Correct the note's surrogate claim (only the string-walking jsonStringCostUpTo
charges a lone surrogate six bytes; the byte walker never sees one). Shrink the
post-truncation fixture below PIPE_BUF for a deterministic single callback. List
the shared stdout/stderr budget as a third honest fail-before exception
(cross-pipe arrival timing is nondeterministic). Add illegal-UTF-8,
broken-multibyte, and child-log-flood regression tests; sync the zh pair.
The stray-sealing regression test asserted copied < 2 MiB — about 4x the
defended sealed shape, so reverting the seal to a re-merge (or removing it)
left the test green. Measured both shapes as the fd-3 sibling does: the sealed
shape copies ~120 KB, the re-merge shape ~538 KB. Tighten the bound to 256 KiB,
which sits between them, and record the measurements in the comment and the
Agent Note so the fail-before claim holds.
stdout and stderr each checked their pending serialized cost against the full
logBudget independently, so both could retain nearly a budget's worth of
newline-free residual at once — double the intended peak, up to ~512 MiB near
the ceiling. The flush threshold now reads the COMBINED cost of both pipes and
flushes both when it crosses, since they share one ledger.
Remove the post-truncation admit() v8-ignore: captureStray's per-line loop
makes that branch deterministically reachable within one data callback (a chunk
whose first newline-terminated line exhausts the budget hits it on the second),
so it is measured by a new regression test rather than ignored.
Refresh two stray-output test comments that still named the removed
StringDecoder; the raw-chunk buffer reassembles a split multibyte sequence by
concatenating before it decodes, and the end flush renders a stranded partial
as U+FFFD via toString('utf8').
Three follow-ups the review caught in the stray-capture rewrite, plus a cost
undercount shared with the log ledger.
Seal the stray fragment list into blocks past MAX_PENDING_CHUNKS, mirroring the
fd-3 reader: a program pacing single-byte os.write(1, ...) calls otherwise
accumulates one live Buffer per write, and the per-object overhead no byte
count sees exhausts the host heap far below the budget.
Flush the residual by its running SERIALIZED cost (serializedBufferCost, a
per-byte lower bound) rather than raw byte count: a control-char-dense
newline-free flood serializes several-fold, so a raw-byte threshold let it grow
to a full budget's worth of raw bytes — up to ~6x what the ledger admits —
before flushStray concat/decoded the whole ~256 MiB residual at once.
Charge a lone surrogate its full six escaped bytes (\uXXXX under ES2019
well-formed JSON.stringify) in both jsonStringCostUpTo and serializedBufferCost,
not the three bytes Buffer.byteLength reports for U+FFFD: a forged log frame
flooding \ud800 escapes was undercharged by half and admitted ~2x maxLogBytes.
Key the sync-spawn leak assertion off the exact bootstrap path from the mocked
spawn's argv, immune to a sibling worker's concurrent staging. Refresh the
stale load-check comment that named the replaced JSON.stringify mechanism.
Add lone-surrogate, stray-sealing, and companion regression tests (per-file
100% coverage); update the Agent Note and zh pair.
The line-aggregating stray capture from the previous round regressed three
ways the review caught. Rewrite it on the fd-3 reader's raw-Buffer-chunk
shape: accumulate chunks with a byte counter and split on the raw 0x0a byte,
so a large newline-free write no longer re-copies the residual and re-scans
from index 0 per chunk (both O(N^2)). Meter each admitted entry by serialized
cost through a new jsonStringCostUpTo that walks to the cap and stops, so a
near-budget control-char-dense line never allocates the sixfold-inflated
JSON.stringify result the old ledger did (the critical: ~1.6 GiB transient
under a large maxLogBytes). Flush the residual explicitly in the closeDeadline
handler before it destroys the streams, so a setsid escapee's path (which
fires no end) does not drop a leader's final newline-free diagnostic.
Harden the sync-spawn leak assertion to a set difference against a pre-run
snapshot, immune to a parallel worker's concurrent tmpdir create/delete.
Decline the round-2 request to enforce the fd-3 ceiling per-frame: the counter
check must precede Buffer.concat to prevent ~2x memory doubling (two
regression tests assert this), and the batch-edge false reject it would fix is
reachable only at a maxLogBytes/maxValueBytes configured within one pipe read
of the 256 MiB ceiling, far past the defaults. Documented at the check and in
the note Alternatives.
Add flood, NUL-flood, short-escape, and closeDeadline-flush regression tests
(restoring per-file 100% coverage); update the Agent Note and zh pair.
The line-aggregating stray capture added two branches — the post-truncation
early return and the residual-overflow admit — that the aggregation and
split tests did not exercise, so per-file coverage dropped below 100%. A
2 MB newline-free native write under a 4 KiB maxLogBytes drives the residual
across the budget (admit-and-truncate) and then short-circuits later chunks,
asserting the captured output ends at the truncation marker and stays under
budget rather than buffering the whole flood.
Wrap spawn and the fd-3 narrowing so a synchronous throw (ENAMETOOLONG on
an over-PATH_MAX pythonBin, EMFILE) removes the run's staging directory and
resolves the same worker-exit class as the async error event, instead of
rejecting run() and leaking the directory.
Aggregate native stdout/stderr by real newline rather than by Node data
chunk: logs entries are joined with "\n" downstream, so a newline-free
write larger than one pipe read no longer reads back with spurious breaks.
The ledger still bounds a newline-free flood.
Track a running scan offset in both frame readers so a large frame
accumulated across chunks is scanned once, not re-scanned from 0 per chunk.
Reword the deadline hard-bound v8-ignore to state its real environment
dependence (PID-1-doesn't-reap container, zombie survivor) and cross-ref
the note's rejected signal-0 alternative; fix settle comments that quoted
the pre-qualification teardown contract; document the capMessage vs
_cap_message billing split on both sides; guard the dispose-after-resolve
heartbeat assertion against a vacuous 0===0 pass; reuse
_TRUNCATION_MARKER_BYTES; note the abandoned-call pending-entry bound.
Update the Agent Note Decision/Testing/Alternatives/Consequences for the
above and record the confirmed-empty finalize as a second honest
fail-before exception; sync the zh pair.
The reap-poll deadline arm sent SIGKILL then finalized immediately, declaring
quiescence on mere signal delivery while the group was still dying. It now keeps
polling for the group to actually empty (bounded by one more reap margin) after
its self-sent SIGKILL, so `finished` resolves only on a confirmed-empty group.
ProtocolChannel.read_frame read the boot/run handshake frames through
FileIO.readline() on the unbuffered fd — one os.read(1) per byte, so a
multi-megabyte program burned CPU (RLIMIT_CPU already in force for the run frame)
in millions of syscalls before ast.parse. It now reads in chunks into the same
_pending buffer the async reader uses; the wrapping os.fdopen is gone. read_frame
is this PR's own code (e7f22ed3), not the protocol layer. The chunked read is a
syscall-count improvement with no cross-platform-deterministic failure to assert,
noted as such in the Agent Note.
The closed-loop reply-pump guard now ships with a deterministic regression test:
a worker thread abandons a binding so its loop closes, the host answers that call
before a later binding, and the pump must survive the closed-loop
call_soon_threadsafe to deliver the later reply (host-gated ordering makes it
deterministic; unguarding the pump hangs the later binding to the wall clock).
Align the quiescence self-description with the shipped setsid limitation:
teardown()'s JSDoc and the Agent Note's Problem line now qualify "no subprocess
outlives the fiber" to subprocesses that stay in the child's process group, with
a setsid()-escape exception pointing at the README. Tighten the setsid-orphan
fixture's self-timeout to 5s and its upper-bound assertion to <4000ms so a failed
deadline backstop is a sharper red. Register the new regression tests in the note.
The child reads these byte budgets through int(...), which silently floors a
float, so maxLogBytes: 3.5 would truncate at 3 bytes child-side while the host
meters and marks at 3.5 — the two sides enforcing different public config. Gate
them to integers at load, as the worker backend does; correct the stale comment
that claimed the int()-truncated caps needed no gate. Adds a regression test.
The SIGXCPU timeout message changed from "CPU budget (Ns) exhausted" to name the
configured value as a ceiling; two existing timeout tests asserted the old text.
Assert "CPU time exhausted" to match.
Raising maxValueBytes' load bound to ceiling-envelope assumed both budgets are
metered in serialized (JSON-escaped) bytes, which held for completion values and
logs but not the diagnostic: _cap_message capped by raw UTF-8, so a control-heavy
message near maxValueBytes could serialize sixfold and breach the fd-3 frame
ceiling — the silent worker-exit inversion the load check prevents. _cap_message
now accumulates per-byte serialized cost (new _JSON_BYTE_COST table) and cuts the
prefix that fits. Also reword the host SIGXCPU timeout message to name cpuSeconds
as the configured ceiling rather than a budget a stricter inherited RLIMIT_CPU
soft may undercut. Adds a control-heavy-diagnostic regression test.
The group-reap poll folded its deadline arm into the empty-group arm, so a host
event loop blocked past graceMs + CLOSE_REAP_MARGIN_MS would run the overdue
poll before the grace SIGKILL timer: the group is still non-empty, the deadline
has passed, and the shared arm cancelled the never-fired SIGKILL and finalized —
releasing a SIGTERM-ignoring same-group survivor for good. Split the arms: empty
group cancels the moot timer and finalizes; deadline-with-non-empty-group sends
SIGKILL itself (idempotent if the timer already ran) before finalizing. Adds a
regression test that busy-blocks the loop past both timers and asserts the
survivor's heartbeat freezes.
The group-reap poll's deadline arm (Date.now() >= deadline) is a backstop that
SIGKILL emptying the reachable group never reaches, leaving one uncovered branch
under the per-file 100% gate. Mark it v8-ignore with the reason and drop the
always-true graceTimer-defined guard inside pollGroup (it runs only when killing
is set, so kill() has armed the timer).
settle() dropped the run from `live` eagerly, before the grace-window SIGKILL
reaped a same-group survivor. A dispose() racing a just-resolved run() then
snapshotted an empty `live` and returned while the descendant was still alive,
so teardown's "no subprocess outlives the fiber" (and its JSDoc) was false for
that window. The run now stays in `live` until the process-group poll confirms
the group empty, at which point it is both dropped from `live` and its finished
promise resolved. Adds a regression test asserting dispose() of a completed run
with a same-group survivor returns only after the survivor stops executing.
The class docstring still credited the GIL plus per-frame PIPE_BUF atomicity for
serializing writes, which _write_lock's full-write loop already superseded. State
the current contract (writers serialized by _write_lock around a full-write loop)
and drop the double blank line under the binding-replies note heading.
Record that die_if_cpu_exhausted compares against the effective clamped cpu_soft
in the rlimit section, and add the recheck-timeout test to Testing; re-record pair.
The settlement-time CPU recheck compared spent CPU against the configured
cpuSeconds, but _clamped may have lowered the effective soft limit to a stricter
inherited value. A program that traps SIGXCPU, burns past the inherited soft,
and returns inside the soft-to-hard gap was checked against the configured value
and falsely reported successful, bypassing the inherited limit. The recheck now
uses the clamped cpu_soft. Adds a regression test that inherits a 1s soft CPU
limit and asserts a SIGXCPU-trapping over-burn is a timeout, not a success.
Record the call_soon_threadsafe-onto-a-closed-loop guard in the binding-reply
section of the settlement-fixes Agent Note; re-record the bilingual pair.
A binding called from a worker thread records that thread's loop for its reply.
If the thread finished and closed its loop before the host reply arrived,
_pump_replies' call_soon_threadsafe onto the closed loop raises RuntimeError;
unguarded, that ends the pump task and strands every later reply. Wrap the
schedule in a try/except that drops the moot reply (nothing awaits it) and keeps
the pump serving.
A descendant that calls setsid()/start_new_session leaves the child's process
group, so kill(-pid) teardown cannot reach it; if it also releases the inherited
pipes the run still settles and the fiber goes quiescent while the orphan runs.
This is the containment boundary (model code has bash-equivalent trust), not a
guarantee; reaching such an orphan needs descendant-pid tracking and is deferred.
Two further review findings on the CPython backend:
- The grace-window SIGKILL timer was left armed after settlement, so on a
normal completion a kill(-pid) could fire up to graceMs later and strike a
recycled pgid once the kernel reused the leader's pid. settle() now clears
the timer the moment the process group is confirmed empty (the normal path
and when the poll sees the survivor gone), bounding the reuse window to the
genuine-survivor case where the group cannot be empty to reuse.
- _clamped bounded rlimits by the inherited hard limit only, silently raising
an inherited soft limit stricter than the request (loosening RLIMIT_AS or
deferring RLIMIT_CPU SIGXCPU). It now clamps each side against its own
inherited counterpart and pins soft under hard, keeping the strictest of
configured and inherited. Adds an inherited-soft-limit regression test.
Agent Note expanded to seven fixes with the two new rejected alternatives;
zh pair re-recorded.
Two review findings on the CPython backend:
- Disposal could return while a same-group descendant that ignores SIGTERM
but releases the inherited pipes was still alive: the leader's close fired
and the previous fix relied on an unref'd SIGKILL timer that a short-lived
host never fires, reparenting the survivor to init. settle() now withholds
the run's finished promise on a ref'd process-group poll until the SIGKILL
has emptied the group (bounded by graceMs + margin, zero-cost when already
empty), so teardown's "await each child's exit" holds.
- A binding called from a model worker thread via asyncio.run created its
reply Future on that thread's loop, but _pump_replies completed it directly
from the main loop; asyncio.Future is not thread-safe across loops, so the
call hung to the wall clock. Replies now complete via the owning loop's
call_soon_threadsafe, and a lock serializes the id claim/write/advance.
Tests: the same-group reap case now asserts a heartbeat file stops (robust
whether the killed descendant is reaped or a zombie, so it holds where PID 1
does not wait() orphans); a cross-loop case runs a binding from a worker
thread and asserts the reply round-trips instead of timing out. Agent Note
expanded to all six fixes with rejected alternatives; zh pair re-recorded.
The frame-ceiling cap test asserted the old (ceiling-envelope)/6 bound and
its 44739232 message. The load bound is now ceiling-envelope because both
budgets are metered in already-escaped bytes; assert 268435392.
Address review findings on the CPython backend:
- CRITICAL: a model program could leave a descendant in the child's own
process group that ignores SIGTERM but releases the inherited pipes, so
the leader's `close` fired and settle() cancelled the pending SIGKILL
before it escalated — run()/dispose() returned while that child lived.
kill() now unrefs the grace timer and settle() no longer clears it, so
the SIGKILL reaches the whole group; killGroup swallows ESRCH when the
group is already gone (the normal case). Adds a real-subprocess
regression test.
- WARNING: the maxLogBytes/maxValueBytes load bound divided the frame
ceiling by 6 for escape expansion, but both budgets are metered in
already-escaped serialized bytes, so a payload occupies at most
cap+envelope on the wire. Bound is now ceiling-envelope; drop the unused
escape constant.
- Narrow the runtime.spec.ts header to "no subprocess mocks" (it mocks
node:fs.copyFileSync for staging-failure cases).
- Use full-width punctuation in the README.zh.md prose per translation
rules; re-record the pair.
The package README (both languages) still described this layer as
protocol-only with the PythonCodeRuntime implementation deferred to a
later PR, contradicting the shipped code. Rewrite the intro to describe
the registered runtime, add a Configuration section for every Config cap,
and drop the "implementation not in this layer" limitation. Also pin the
residual-detach fixture's size invariant: the byteLength assertion only
holds above Node's Buffer pool threshold.
The manifest omitted @deepseek-ai/dsh-code-runtime although src/index.ts
imports CodeRuntime and the portable-identifier constants from it and the
tsconfig references ../code-runtime. A three-way package.json merge over
the protocol-layer stub dropped the entry; restore it in peer and dev
dependencies so the declaration matches the import.
Unwrap the English note to one physical line per paragraph (verify-md-wrap)
and retarget the backend link to the fd-3 protocol architecture note that
this stack actually ships (verify-md-links); re-record the bilingual pair.
Land the PythonCodeRuntime implementation on top of the fd-3 protocol
seam: python3 -I per run, binding namespace over fd 3, RLIMIT_CPU/AS,
wall-clock timer, and SIGTERM->grace->SIGKILL process-group teardown,
with the real-subprocess integration suite.
Fixes three defects surfaced on the source PR's review before they ship:
- boot-write failure resolved a worker-exit through finish()/settle()
that read wallTimer/onAbort/live in their TDZ, rejecting run() instead;
the boot write now runs after those bindings and the v8-ignore that hid
the branch is removed.
- log capture serialized against settlement with no lock while model
daemon threads keep writing; LogBuffer now owns one shared re-entrant
lock taken by write/flush_line/push.
- the fd-3 line residual was a subarray view pinning the whole joined
frame; it is copied into a right-sized Buffer via detachResidual so
pendingBytes measures what is retained.
pnpm hardlinks node_modules files to the store on the same volume, and
TypeScript's native realpath resolves those links back to store paths
(F:/.pnpm-store/v11/files/...), producing TS6231 during tsc -b and vite
resolution. ReFS block cloning (package-import-method=clone) gives each
file an independent path while sharing physical blocks, avoiding the
leak without the copy cost. Clone mode needs the @reflink/reflink native
module, which the system corepack pnpm carries but pnpm/action-setup's
dest build omits, so installs run through corepack pnpm.
The install steps branch on the workspace filesystem: clone only on
ReFS, plain install on hosted NTFS (which rejects copy-on-write). The
serial-windows store points at F:\.pnpm-store to share the ReFS volume.
Agent Note 2026-08-30-windows-refs-store-block-clone-install records the
rationale; ci-workflow.spec asserts the branch.
Every case in the suite drives real git invocations against a scratch
repository, so it is bound by process creation rather than by its assertions.
The describe-level 15 s capped all 23 cases below the 90 s the Windows coverage
lane passes as --testTimeout, and the suite has been observed timing out at
15000ms on a branch that did not touch the file.
Refs #2677.
chmod 0 can only deny directory reads on POSIX to a non-root owner:
Windows exposes no directory permission bits for readdir, and root
bypasses them. Where the fixture stays readable the sealed candidate is
indexed, so the unreadable-branch behavior is pinned on POSIX non-root.
An injected readdir failure keeps that branch covered on every platform.
Replace Vitest's hash-based --shard with a coordinator-side
longest-processing-time assignment. The coordinator collects the
instrumented inventory from a vitest list run (dropping the exempt
heavy suites that list does not exclude), reads per-file durations from
the Vitest results cache, and seeds heavy subprocess-bound suites into
different partitions. A weight-aware test fails when assignment ignores
recorded weights, verified by injecting a file-count-only rule.
Windows coverage measured partition spread of 442s (275-717s) under
hash sharding; a simulation with the same file durations and the new
assignment balances partitions to within 21s, cutting the critical
partition to roughly half.
Replace the debounce-timer polling with Include.stop(), which flushes the
file-backed write queue deterministically. This avoids Windows coverage
flakes where the self-dispose write could land after the expect.poll
timeout.
Stacked on #3115: keep its windows-* setup-pnpm-js-<run_id>-<run_attempt>-<job>
destination, and extend the same isolation to non-Windows jobs in ci.yml
and ci-master.yml with setup-pnpm-<run_id>-<run_attempt>. This prevents
sequential self-hosted Windows jobs from tripping over a stale locked
pnpm.exe/reflink native module.
The anchors-a-relative-add-spec case serializes two subprocesses (plugin
add + remove) under a hardcoded 90s budget, which the 2x60s worst case
exhausts; derive it from SPAWN_TIMEOUT_MS * 2 + 30s like the other
dual-call cases. The pnpm setup isolation note now also records the python
SDK exe build's suffixed destination and its regression-test coverage.
The plugin add and dump-default-config cases serialize two runBuiltBin
calls, each with a 60s execa cap; the 90s outer budget could be exhausted
before the second call. Raise them to SPAWN_TIMEOUT_MS * 2 + 30s, matching
the multi-call treatment.
The requires-profile and routes-help cases serialize 4-6 runBuiltBin calls,
each with a 60s execa cap; under the loaded pool the 90s outer budget was
exhausted before the last call and vitest truncated the run without the
execa diagnostics. Raise both to SPAWN_TIMEOUT_MS * 3 + 30s.
* ci(build-exe): drop pull_request label trigger to avoid skipped checks
* docs(build-exe): sync agent note and pin event set in workflow spec
* test(ci): type-safe event key assertion for build-exe workflow
* ci(build-exe): use present-tense trigger comment and drop label-run note
The built-bin help/usage case still used a win32-conditional 60/30s outer
budget while serializing six runBuiltBin calls, and startProfileLifecycle
lacked the execa timeout/killSignal the sibling helper has; the tool-ralph
cases pinned 20-30s explicit timeouts that the 90s lane default cannot
override. Align all of them to the SPAWN_TIMEOUT_MS + 30s (or 90s) pattern.
The per-case 15-30s budgets on the Windows native and coverage lanes fire
before oxlint, workflow-worker-thread, and other subprocess-spawning cases
finish under the loaded self-hosted pool; the failures rotate across cases
as load shifts, so per-case widening only moved the flake. Raise the lane
defaults (DSH_COVERAGE_TEST_TIMEOUT_MS and the native --testTimeout) to 90s,
align the oxlint and workflow-worker-thread case budgets, and keep the
built-bin SPAWN_TIMEOUT_MS at 60s under a 90s outer budget.
The windows-* jobs now install pnpm under a run/attempt/job-suffixed
destination, so the pre-install step that cleared the old fixed
setup-pnpm-js path no longer touches the actual destination and its
comment claims stale state. The suffix already gives every job a fresh
directory, so remove the four cleanup steps.
The execa timeout was widened to 60s but the outer vitest case budgets stayed
at 30s, so a cold-starting built bin would trip the vitest budget first and
the execa SIGKILL cleanup could not run inside it. Extract SPAWN_TIMEOUT_MS,
share it across the execa deadline, its error text, waitForFile, and the
outer case budgets (60s spawn + 30s headroom), so the widening is coherent.
Both suites spawn real subprocesses (oxlint probes; the dsh built bin) that
cold-start slowly on the contended self-hosted Windows pool, so their 20-25s
timeouts fire before the child finishes. Raise the oxlint contract case
timeouts to 60s and the built-bin execa timeouts to 60s, matching the
tool-ralph budget treatment.
The windows-* jobs keep a separate standalone pnpm executable under
runner.temp/setup-pnpm-js. A previous job on the same self-hosted runner
can leave a locked @reflink native module there, so the next job's
pnpm/action-setup fails with EPERM during unlink before any test runs.
Suffix the destination with run_id, run_attempt, and job so every job
gets a fresh directory even when sequential jobs land on the same
runner; apply the same to the python SDK exe build. Update the pnpm
setup isolation note to record the Windows-specific destination.
The full-corpus transform gate spawns one child that transforms and imports
every built bundle, so it runs for 8-25 minutes as a single case and
dominates one native Windows coverage partition, blowing its 900s budget
under load. Move it to the coverage-exempt heavy gate, which runs it with
its own worker budget instead of competing with the instrumented
partitions. The package's src is threshold-excluded in vitest.config.ts, so
the exemption carries no coverage; the exempt-heavy roster note records the
entry.
The self-hosted Windows coverage pool (16 shards x 12 workers on 192
threads) pushes real subprocess boots past their vitest deadlines: the
tool-pwsh Loader smoke reaches ~40s against a 30s process cap, and the
tool-ralph worker-thread cases exceed the 5s default. Give the pwsh
smoke a 90s process deadline (the subprocess keeps the assembled boot,
the vitest deadline stays at 120s), and give the two un-budgeted ralph
cases 30s each, matching the existing 20s quiescence case.
On Linux, pwsh may trap SIGTERM and exit cleanly when the subprocess service
is disposed, so the handle status is 'completed' rather than 'killed'. The
test already proved the process tree is gone via kill(pid,0); both statuses
satisfy the contract.
Other PRs are blocked by Windows ACP half-close tests timing out. Keep the
coverage job running for signal, but remove it from all-checks-passed.needs
until the Windows skip fix is validated.
Windows anonymous pipes do not surface a child stdout EOF while the child
process stays alive. The three tests that simulate 'child closes protocol but
stays alive' therefore cannot be reproduced on Windows and hang until the
test timeout. Skip them on win32.
These tests spawn real ACP child subprocesses. On contended self-hosted
Windows runners the default 30s budget times out. Instead of raising the
global coverage timeout, give this file 2x the configured default
(DSH_COVERAGE_TEST_TIMEOUT_MS) so it follows future default changes.
After the 4-partition split, other PRs' windows coverage now fails on
process-bound subagent-acp tests timing out at 30s under self-hosted
concurrency. Give the coverage lane the same 60s per-test budget that the
earlier failover profile used.
The PR changes native Windows coverage partitions from 8 to 4 to reduce
vitest worker startup pressure under high self-hosted concurrency. Sync the
implemented Agent Note (EN/ZH) so the decision record no longer says Windows
is fixed at 8, and revise the same-partition-count alternative accordingly.
Keep the 4-partition coverage profile, split the monolithic windows-native
job into smaller required jobs (build, coverage, native-tests) plus a
non-blocking observational job. Update ci-workflow.spec for the new topology.
Under high self-hosted concurrency, 8 partitions per Windows native job
triggered vitest fork worker startup timeouts. This branch lowers Windows
coverage to the same 4 partitions Linux uses, trading some single-job
coverage wall time for lower process-creation pressure.
The Windows native coverage lane was hitting Vitest's 30s per-test
ceiling on slow subprocess/ACP fixtures. Doubling the per-test budget
absorbs cold-start and process-teardown jitter without changing the
assertions or the job-level 120-minute cap.
python-release.yml only triggers on workflow_dispatch, so build.if:
github.event_name == 'workflow_dispatch' is always true and redundant; remove it
(the exact event set is already pinned in the spec). Update the spec assertion
accordingly.
Address PR #2875 review:
- Re-record python/development.i18n.yaml (corpus verify-translation-pairing was
out of sync after editing development.md/zh.md) and the 2026-08-11
python-publication-workflow pair after the dry-run wording tweak.
- Tighten the python-release spec assertion to the exact event set
(['workflow_dispatch']) instead of not.toHaveProperty('pull_request').
- Fix the 'dry-run run' wording in development.md and the note.
Corpus-wide verify-translation-pairing (1001 pairs) and note-format (594) pass;
ci-workflow.spec.ts 14/14.
The static gate (translation pairing) failed on a pre-existing master note:
2026-08-20-composer-edit-range-from-selection.zh.md:17 linked the zh target with
the en .md path. Point it at the .zh.md target and re-record the i18n hash. This
unblocks the required node 24 / static gate (it is not part of the python-release
gray-check change but sits on the same PR's CI path).
Chinesezjc