Repository navigation
test: skip the git hook scan in the slack bar budget tests - #39530
bolichen97 wants to merge 2 commits into
Conversation
Each git read in these tests ran three extra scan gits first. On a loaded Windows runner a six-refresh test passed --timeout=180 and took its xdist worker down. Stub the scan to the pins; the scan has its own tests.
UX Review (Opus 5.5) — ⏭️ skippedRevision |
First Principles Review (Opus 5.5) — 🔴 BLOCKPremise-level review of First-Principles-Verdict: BLOCK The description says the PR stubs the hook scan in What this change shipsInventory (1 items) — 0 justifiedIntent: stop The patch file's other hunks (approval-timeout note,
BlockersThe PR's framing and diff disagree; the fix is somewhere else — rides along [FIRST-PRINCIPLES-REVIEWED] 2ea3fae Description read[DESCRIPTION-READ] 420307c50f0d3cfe5f7cdfbf4b7fea43ef19ba402e9858d913f4929a0d1246da That sha256 names the PR description this verdict read. Recompute it with |
Design Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)Design-level review of Design-Verdict: PASS The diff does not fix the timeout its Goal names: it adds only a docs bullet. The fixture stub the description describes is not in it, and the bullet points to #39505 for that stub. BlockersPhantom description: the stated fix is missing from the diff Notes
[DESIGN-REVIEWED] 2ea3fae Description read[DESCRIPTION-READ] a8c0860bcee673c7f1201d6c5bb780450f44feeaefaa73938efbc62893aa6e19 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Review detailsNo findings. Notes:
[OPUS-REVIEWED] 2ea3fae Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 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: |
|
Closing as a duplicate of #39505, which makes the same test-side change and opened first. The testing-conventions rule moves to its own docs PR. |
Problem / Motivation
Goal:
test_the_bar_is_resent_only_when_its_items_changestops timing out on Windows CI, so it no longer turnsBackend Tests (Windows) (8)red on unrelated PRs.Since #39390 (six xdist workers per Windows backend shard),
test/test_slack_code_channel_bar_budget.py::test_the_bar_is_resent_only_when_its_items_changehitTimeout >180.0sand took its worker down (worker 'gwN' crashed,--max-worker-restart=0) on six PRs targeting main in seven days (#36573 #39005 #39500 #39504 #39506 #39507) and on main (run 44547665024). Linux stays green.Why it matters
Every PR whose Windows shard 8 lands this test on a busy worker goes red for nothing, and the dead worker drops the rest of its tests from the shard.
Not a goal
git_config_hooks' scan orworktree._run_gitcheaper in production.test/test_slack_code_channel_board.py,test/test_slack_code_channel_workspace.py,test/test_slack_code_channel_board_diff.py) are not on the Windows flaky guard's failure list; a timeout there is a follow-up with its own evidence.What changed (motivation → approach → change)
Symptom → cause: the timeout dump in all three logs read (main job 135658554621, PR #39500 job 135681376478, PR #39506 job 135707964255) shows the main thread waiting in the event loop at the test's last
refresh(line 610) and theasyncio_0thread inside a git spawn ofgit_config_hooks._submodule_hook_names. No lock or sleep is involved: the test is slow, not stuck. Every git read throughworktree._run_gitfirst runs the hook scan, three more gits (config,rev-parse --show-toplevel,ls-files -s), so each read is four Windows spawns. The same logs' progress records show this file's service tests at 58-107 s each on shard 8; this test makes six refreshes and crosses 180 s.Change: the file's autouse
_no_sandboxfixture, which already replacessandboxed_spawn_argvwith a passthrough, also replacesworktree.config_hook_disable_args_sandboxedwith a stub that returns the submodule-recursion pins only (git_config_hooks._disable_args_for_names([])). These throwaway repositories define no config hooks, so the real scan would return the same argv. No assertion in the file is about the scan; it is tested intest/test_git_config_hooks.py. Thetesting-conventions.mdbullet next to "Sandbox only the cases whose subject is the sandbox" records the rule and the measurement.Backwards compatibility
Compatible: test-only and docs change; no product code moves.
Removes nothing: only a test fixture stub and a doc bullet are added.
Tests
test/test_slack_code_channel_bar_budget.py: every test keeps its assertions; only the hook scan under its git reads is stubbed.Proof: not provable -- test-only change
Manual verification
N/A — the real-machine runs below are the verification.
Real-machine verification
Pending: filled in from the AWS Windows box run (base vs head, same command,
-n 6).Related Issues
Windows flaky test guard finding; regression window opened by #39390.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...); PR Hygiene does not count commitsContribution License Agreement
Contributed under the project's Apache-2.0 license.