Stabilize watcher test readiness (#28194)

This commit is contained in:
Kit Langton
2026-05-18 12:51:46 -04:00
committed by GitHub
parent a2e6bd503b
commit 3ab67f3280
2 changed files with 54 additions and 22 deletions
+7 -5
View File
@@ -67,6 +67,7 @@ Repeated setup work, long sleeps/timeouts, serial integration tests, filesystem/
| Remaining prompt behavior tests mostly do not require repository state | Removed git setup from safe loop/reference/error fixtures; restored shell queue/cancel cases | 23.400s | 19.610s | keep | Safety review found shell runner readiness depends on git-backed setup in several tests; current single rerun passes. |
| Session processor effect tests do not require repository state | Removed git setup from all processor-effect temp server fixtures | 12.500s | 9.230s | keep | Two targeted reruns passed after the change: 9.61s, 9.23s. |
| HTTP listen PTY ticket tests restart the same listener topology twice | Folded directory-scoped ticket regression into the broader unsafe-ticket test | 7.051s | 6.170s | keep | Two targeted reruns passed after the change: 6.76s, 6.17s; still covers mint failure and successful same-directory upgrade. |
| File watcher readiness can write before async native subscriptions are active | Retried short readiness writes and accepted symlink-realpath HEAD events | failed | 4.62s | keep | Three sequential focused watcher runs passed: 4.62s, 4.57s, 4.64s; full suite no longer failed in `watcher.test.ts`. |
## Profiling Results
@@ -107,11 +108,12 @@ Targeted 3-run baselines:
Full-suite sanity checks:
| Command | Result | Notes |
| -------------------- | -------: | -------------------------------------------------------------------- |
| `bun run bench:test` | 225.069s | Before continuing prompt/session work. |
| `bun run bench:test` | 186.729s | After prompt, processor, and PTY wins before safety review restores. |
| `bun run bench:test` | 202.317s | After restoring prompt shell coverage and SDK VCS parity coverage. |
| Command | Result | Notes |
| -------------------- | -------: | ----------------------------------------------------------------------------------------------------------------------------------------------- |
| `bun run bench:test` | 225.069s | Before continuing prompt/session work. |
| `bun run bench:test` | 186.729s | After prompt, processor, and PTY wins before safety review restores. |
| `bun run bench:test` | 202.317s | After restoring prompt shell coverage and SDK VCS parity coverage. |
| `bun run bench:test` | failed | Watcher blocker cleared; current run later failed in focused-passing `tool/skill.test.ts` and prompt shell timeout cases under full-suite load. |
## Dead Ends