You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Math.sumPrecise: capture the iterator next method once #750
Math.sumPrecise obtains the iterator object once, but calls callMethod(iter, "next", ...) for every step. That re-reads the observable next property each iteration instead of retaining the [[NextMethod]] captured by GetIterator. A getter or mutation can therefore change the method mid-consumption, diverging from ECMA-262 and potentially repeating attacker-controlled lookup side effects.
Scope
Capture the iterator next method once before the first step.
Invoke that exact value with the iterator as this on every step.
Preserve IteratorClose precedence for a non-Number element and all existing exact-summation behavior.
Cover a next getter, mutation after the first step, non-callable captured methods, and abrupt completion.
No-workaround rules
No array-only fast path, source rewriting, repeated property lookup, swallowed primary exception, skipped IteratorClose, or coercion of non-Number elements.
No stack, queue, job, Promise, Worker, or Map/Set iterator implementation changes.
Acceptance
The next getter is observed exactly once and later mutation does not change the captured method.
Non-callable/throwing paths preserve the specified error and IteratorClose behavior.
Focused units and the complete Test262 Math.sumPrecise subtree pass with exact nonzero accounting.
Full integration gates pass with 0 unrelated Test262 flips.
Math.sumPrecise now captures the iterator record's [[NextMethod]] once and reuses it for every IteratorStepValue call.
Added regression coverage for an observable next getter, mutation of iterator.next after the first step, a non-callable captured method, and IteratorClose preserving the original TypeError when return() throws.
Focused unit evidence: exact regression 1/1 passed; the complete Math.sumPrecise filter 2/2 passed; zig build test262-bin passed.
The official test/built-ins/Math/sumPrecise 10-file run and wider gates are queued until an unrelated active Zig corpus/build job clears, per the repository's no-overlapping-corpus rule.
Landed on main as 28482211. Exact-parent Test262 evidence for test/built-ins/Math/sumPrecise: 10 pass -> 10 pass, 0 fail, 0 flips. Final focused regression: 1/1 passed, 0 leaked. Leaving this issue open until the main CI integration run completes.
Descendant CI 32804870572 completed 51 jobs successfully, including all unit/TSan unit shards, Test262 parallel, every TSan no-GIL corpus shard, Wasm smokes, docs, private ABI, and lifecycle fuzzing. Its sole functional failure is the separately tracked #722 mc-lock-stop-vs-park assertion; the benchmark harness was canceled only when the next direct-main #752 push superseded the run. The full-integration acceptance box therefore remains deliberately unchecked pending a descendant matrix without #722 red.
Full integration remains deliberately open. CI 32890815195 completed 52 required jobs successfully (including all unit, TSan, Test262, no-GIL corpus, threadfuzz, docs, Wasm, and ABI legs); the sole remaining benchmark-comparison-harness leg was automatically cancelled at the matrix's exact 60-minute limit, with no test failure reported. The causal CI-duration defect is tracked in #765. I am not treating a cancelled leg as a pass.
Full integration is now proven by main CI run 32923883777: all 53 required jobs passed, the sole nightly-only job was skipped as configured, and 0 jobs failed or were cancelled. This includes all unit/TSan, Test262, threading/no-GIL, fuzz, Wasm, docs, ABI, and benchmark harness gates. The implementation remains 28482211, with exact-parent Math.sumPrecise Test262 evidence 10/10 -> 10/10 and 0 flips.
Activity
chrisbbreuer commentedon Aug 25, 2026
Implementation progress:
Math.sumPrecisenow captures the iterator record's[[NextMethod]]once and reuses it for everyIteratorStepValuecall.nextgetter, mutation ofiterator.nextafter the first step, a non-callable captured method, and IteratorClose preserving the original TypeError whenreturn()throws.1/1passed; the completeMath.sumPrecisefilter2/2passed;zig build test262-binpassed.The official
test/built-ins/Math/sumPrecise10-file run and wider gates are queued until an unrelated active Zig corpus/build job clears, per the repository's no-overlapping-corpus rule.fix(math): capture sumPrecise next method (#750)
chrisbbreuer commentedon Aug 25, 2026
Landed on main as 28482211. Exact-parent Test262 evidence for
test/built-ins/Math/sumPrecise: 10 pass -> 10 pass, 0 fail, 0 flips. Final focused regression: 1/1 passed, 0 leaked. Leaving this issue open until the main CI integration run completes.chrisbbreuer commentedon Aug 25, 2026
Main integration run: CI 32804374678. It is queued; this supersedes the prior main run under the repository concurrency policy.
chrisbbreuer commentedon Aug 25, 2026
The #750-only run was superseded by the immediate #751 follow-up. Its full integration gate is now CI 32804791010, which contains both fixes.
chrisbbreuer commentedon Aug 25, 2026
Final descendant integration run containing #750, #751, and the proof-only follow-up: CI 32804870572.
chrisbbreuer commentedon Aug 25, 2026
Descendant CI 32804870572 completed 51 jobs successfully, including all unit/TSan unit shards, Test262 parallel, every TSan no-GIL corpus shard, Wasm smokes, docs, private ABI, and lifecycle fuzzing. Its sole functional failure is the separately tracked #722 mc-lock-stop-vs-park assertion; the benchmark harness was canceled only when the next direct-main #752 push superseded the run. The full-integration acceptance box therefore remains deliberately unchecked pending a descendant matrix without #722 red.
chrisbbreuer commentedon Aug 25, 2026
Full integration remains deliberately open. CI 32890815195 completed 52 required jobs successfully (including all unit, TSan, Test262, no-GIL corpus, threadfuzz, docs, Wasm, and ABI legs); the sole remaining
benchmark-comparison-harnessleg was automatically cancelled at the matrix's exact 60-minute limit, with no test failure reported. The causal CI-duration defect is tracked in #765. I am not treating a cancelled leg as a pass.chrisbbreuer commentedon Aug 26, 2026
Full integration is now proven by main CI run 32923883777: all 53 required jobs passed, the sole nightly-only job was skipped as configured, and 0 jobs failed or were cancelled. This includes all unit/TSan, Test262, threading/no-GIL, fuzz, Wasm, docs, ABI, and benchmark harness gates. The implementation remains 28482211, with exact-parent Math.sumPrecise Test262 evidence 10/10 -> 10/10 and 0 flips.