Repository navigation
[WRONG BRANCH] test(ci): reap the cold-spawn descendant before the test ends - #17
Conversation
The descendant-holding-pipes test left a detached grandchild alive for about 3 s into the next file of the same Bun process. In run 42898913860 shard 1/4 batch 4 hung in that next file's Bun.spawnSync git call on both attempts until the 120 s batch timeout. The test now parses the grandchild pid, kills it, and asserts it is gone before the test ends. Co-Authored-By: Claude Code <noreply@anthropic.com>
|
✅ Deterministic PR hygiene checks passed. |
github-actions
Bot
changed the title
⏳ DRAFT
What to do
Its title has been prefixed with |
|
QA (qa-low-6e) code-read GREEN at cba399e. Pid parse is safe: it reads the child's own stdout line, NaN or 0 never reaches process.kill (guarded by > 0) and the test asserts pid > 0, so it fails loudly. SIGKILL targets the detached descendant only, within about 5 s of its spawn (8 s lifetime). Assertion discriminates: owner mutant log shows fixed 1 pass 0 fail, mutant 0 pass 1 fail. The PR body states the honest limit (the 120 s hang did not reproduce in 11+ runs; this removes the leaked state, it does not prove the Bun race). Nit: the mutant fails only while the 2 s wait is shorter than the descendant's remaining life, so on a very slow box the guard weakens (never the fixed run). Receipt: ~/.agentradar/qa-low-6e/review-ocx17-cba399e24.json |
Co-Authored-By: Claude Code <noreply@anthropic.com>
|
QA (qa-low-6e) re-stamp GREEN at 023c62d. Delta from the reviewed cba399e is one commit touching 4 files, one line each: package.json, desktop/src-tauri/Cargo.toml, Cargo.lock, tauri.conf.json, 2.65.0-vadymhimself.18 to .19 (base is on .18; same four files as #15's bump). The test change is unchanged. Receipt: ~/.agentradar/qa-low-6e/review-ocx17-023c62da6.json |
aef6e32
into
release/2.65.0-vadymhimself.15
What
tests/ci-workflows/cold-spawn-warmup.test.ts: the "descendant holding the child's pipes" test now kills the detached grandchild it spawns and asserts it is gone before the test ends.Why
Run 42898913860, test 1/4: shard 1/4 batch 4/40 hit the 120 s batch timeout (exit 124) on both attempts with 0 test failures. In both attempts the hang was in
build-release-changelog.test.ts("parses actual NUL-delimited git-log output", aBun.spawnSyncgit call), about 0.45 s after the cold-spawn descendant test ended.That test's grandchild runs
setTimeout(() => process.exit(0), 8_000)detached with inherited stdio. The test settles after about 5 s, so the grandchild stayed alive about 3 s into the next file of the same Bun process. That is the only cross-file state this batch leaves behind.Honest limits
Check
The new
expect(alive()).toBe(false)fails on today's code. Lane rundevops-r-ocxb4-check(bun 1.4.0):1 pass 0 fail0 pass 1 fail(Expected: false, Received: true)Eval added/updated: N/A deterministic (unit/integration test: tests/ci-workflows/cold-spawn-warmup.test.ts)
🤖 Generated with Claude Code