Repository navigation
ci: Nightly's TLA+ legs drop a deleted config and expect EndedRelay's open finding - #8580
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… expects it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
No actionable comments were generated in the recent review. WalkthroughThe Nightly test checks that each configured TLA+ job path exists, including paths with a leading exclusion marker. The workflow comment documents this check. The failover configuration comments now identify NoSkipWhileHeld as an open finding and describe the ended-track relay scenario. Merge Risk:
|
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes: the full 3-file diff, plus a check that the new test's regex matches the tla: job in nightly.yml.
- Nightly matrix: drops the
route behindleg and theBehindDear.cfgexclusion from therouteleg, matching #8541's deletion. EndedRelay.cfg: adds the\* Open finding: NoSkipWhileHeldmarker, sosh/tla.shexpects the break.test_every_tla_leg_path_exists: fails when a TLA+ leg or exclusion names a path that no longer exists. Thetla:job is followed by other jobs, so the^ tla:\n(.*?)^ \Smatch works, andassertIsNotNoneandassertTrue(paths)guard against the regex silently matching nothing.
claude-sonnet-5-5 | 𝕏
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes: the delta since the prior Pullfrog review, which is a merge of main into the branch.
- Merge of
main: the matrix comment innightly.ymlwas reworded by the merge, andsh/gh/nightly.test.pygained aself-hostedexemption in an unrelated test. The PR's own changes (droppingroute behind, theEndedRelay.cfgopen-finding marker,test_every_tla_leg_path_exists) are unchanged.
claude-sonnet-5-5 | 𝕏

Problem
Two TLA+ legs of Nightly run 43250182795 failed without any rule regressing:
error: no TLA+ config under: rs/moq-net/verify/route/BehindDear.cfg. test(net): model the picked route rule with a route per prefix #8541 deleted the config; the matrix still named it, and therouteleg still excluded it.EndedRelay.cfgbrokeNoSkipWhileHeldat depth 31 (exit 12). That is the open finding theverify-failoverskill documents, but the config had no\* Open finding:line, sosh/tla.shcounted the expected break as a failure. Every other config in that leg passed or broke as marked.Approach
route behindleg and its exclusion from therouteleg.sh/gh/nightly.test.pyfails when a TLA+ leg or exclusion names a path that no longer exists.just gh testruns on every diff, so the next deleted config fails at the gate instead of overnight.EndedRelay.cfg\* Open finding: NoSkipWhileHeld, as ci: model-check every TLA+ config nightly, and review resyncs a model #8483 did for the other open findings, so a fix that makes it pass is noticed.Impact
None on public API or wire. Nightly runs one fewer TLA+ job.
Validation
main's workflow (listsBehindDear.cfgtwice) and passes on this branch.sh/tla.sh rs/moq-net/verify/failover/EndedRelay.cfgwith a stubjustexiting 12, the exit CI's TLC gave: exit 1 onmain's config (error: ... exit 12), exit 0 with the marker. TLC was not rerun locally: the host's gate was saturated, and CI's run reproduces the skill's trace (depth 31,NoSkipWhileHeld).git ls-fileson therouteleg's pathspec lists Blind, Dead, Early, Lag, LagWide, Loop and Twice.Caused by
BehindDear.cfgwithout updating Nightly's TLA+ matrix.\* Open finding:marker but missedEndedRelay.cfg, which test(net): model failover along a relay chain, and pin two fetch bugs it found #8482 had added earlier that day.Alternatives
sh/tla.shskip a missing path instead of failing: a leg that silently checks nothing is worse.Follow-ups
None.
Models
Size: S
Implementer: Claude Code, claude-opus-5-5, unknown effort
(Written by Claude Opus 5.5)
🤖 Generated with Claude Code
Review
Rewritten by
quest reviewafter each round.Head: 66e5149
Base: origin/main at 9ea441e
Reviewer: Codex, gpt-6.1-sol, medium effort
Session:
01a1d1b4-012f-72e1-b44d-13c78d488cf1, round 1Score: 9/10,
The path check matches the current TLA+ matrix and the Open marker identifies a documented expected failure; source review found no actionable defects, with tests not run as instructed.Findings: none