Repository navigation
test(ios): load-tolerant budgets for two timing tests - #14470
Conversation
`nextScopeDoesNotWaitForOldSocketClose` failed both cases in the full iOS simulator suite on main (runs 36114121829 and 36114119494). Each failure happened after 6.8 s, even though its waits have a 1 s budget. Run alone on the same runner class (run 36117744354), the suite passes in 26 ms. Under the parallel full suite the cooperative pool stalls long enough that the 1 s timer loses the race against an event that does arrive. Both waits in the suite now share a 15 s budget. That does not weaken the check: in a regression the blocked socket close is released only after the waits return, so they still time out and fail. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`pathTimeoutPreservesCompletedProbeProofs` expects the gate deadline to fire after readiness and a successful probe, while it waits for a path that never arrives. Its deadline was 20 ms. On the owned mini (run 36114119494) the full suite delayed the probe past that deadline, so the report carried no proofs at all and nine assertions failed. The same commit passed on Blacksmith. Three seconds keeps the test bounded (it always waits out the deadline) and leaves room for the parallel suite. The sibling path tests already use one second. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Warning Review limit reachedNext included review available in 35 seconds. View limit detailsLimit details: You’ve used all 10 included reviews currently available. You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Review configuration: ⚙️ Run configurationConfiguration used: Repository: manaflow-ai/cmux/.coderabbit.yaml Review profile: ASSERTIVE Plan: Advanced Run ID: 📒 Files selected for processing (2)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
All contributors have signed the CLA ✍️ ✅ |
|
Merge receipt for |
1 participant
Two tests race a short timer against work that the parallel full suite delays:
nextScopeDoesNotWaitForOldSocketCloseuses 1 s waits. Both cases failed after 6.8 s in the full suite on both runners, but the suite passes in 26 ms alone (run 36117744354). The waits now share a 15 s budget. A regression still fails, because the blocked close is released only after the waits return.pathTimeoutPreservesCompletedProbeProofsuses a 20 ms gate deadline, which must outlast readiness and the probe. It failed on the owned mini (run 36114119494) with no proofs recorded. It now uses 3 s; the sibling path tests use 1 s.Part of getting the full iOS simulator suite (
test-ios.yml,ios-simulator (iphone)) green on main. It failed the same way on Blacksmith (run 36114121829) and the owned mini cmux10s (run 36114119494). These tests were hidden while the iOS test target did not compile (fixed in #14421), and PR CI runs only selected tests.Verified in the combined full-suite run 36119701161 (all fix branches stacked): every test this PR touches passes.
🤖 Generated with Claude Code
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.Summary by cubic
Makes two iOS simulator tests tolerant of the parallel full suite by giving their timeouts more headroom.
nextScopeDoesNotWaitForOldSocketClosewaits now share a 15 s budget (was 1 s); a regression still fails because the blocked close is released only after the waits return.pathTimeoutPreservesCompletedProbeProofsuses a 3 s gate deadline (was 20 ms) so the deadline no longer fires before readiness and the probe finish.Written for commit 76b90bb. Summary will update on new commits.