Repository navigation
test: record the pid-reader site that #14510 added - #14883
Conversation
The ratchet counts 161 reader sites on main and records 160. The extra site is session_cleanup.py (2 -> 3), added when the sweep started asking the ownership record which sessions run on a pid, without the count being re-recorded in the same commit. Record it, so shard 4 stops failing on every open PR.
Design Review (Fable 5) —
|
GPT 5.6 Review — ✅ no blocking findingsGPT 5.6 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
Opus 5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
Problem / Motivation
Goal: make
test_pid_reader_sites_match_the_recorded_count_exactlypass onmainagain.The ratchet scans 161 pid-reader sites on
mainand records 160, soBackend Tests (3.12, 4)andBackend Tests (Windows) (4)fail on every open PR whose CI merges pasta452877dc5. The extra site issrc/kiro_crew/session_cleanup.py(2 -> 3), from #14510, which started asking the ownership record which sessions run on a pid without re-recording the count in the same commit. Tracked in #14776.Why it matters
Shard 4 is red for the whole repository, not for the PRs it appears on. Until the count matches the tree, every PR carries a red it cannot fix, and a second real regression on that shard would hide behind this one.
What changed
test/test_pid_reader_ratchet.py:_BASELINE_SITES160 -> 161.That is the step the ratchet's own doc prescribes for an added reader: record the new number. The site is an ordinary reader in a sweep, not the module that owns a runtime or the function that ends it, so
pid-owner-okdoes not apply to it.Not a goal
session_cleanup.py.Tests
test/test_pid_reader_ratchet.py: 6 passed on this branch; 1 failed on unmodifiedorigin/mainwithpid-reader sites: 161 in 31 files (recorded 160).black --check,flake8, comment-history gate (COMMENT_HISTORY_BASE_REF=origin/main): clean.Manual verification
Reproduced the failure on a fresh worktree at
origin/mainbefore the change; the same command passes after it.Security & Privacy
None. Test constant only.
Rollback
Revert the commit.
Pattern harvest
Rule candidate: a PR that adds a pid-reader site must move
_BASELINE_SITESin the same commit; the ratchet says so in its failure message, and a reviewer can check the two hunks travel together.Related