RomanNum3ral
/
gstack
mirror of https://github.com/garrytan/gstack.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix(test): tolerate skill-discovery race in PTY plan-mode smoke 

The e2e-pty-plan-smoke suite (office-hours / plan-mode-no-op) failed in CI with
`Unknown command: /office-hours` (claude exited ~10s) while passing locally. Root
cause: a cold CI container's overlay-FS scan of the symlinked ~/.claude/skills
registry finishes AFTER the runner's 8s boot grace, so the first `/skill` send
reaches claude before the skill is indexed and is rejected as unknown. The runner
gave up on the first "Unknown command:" line.

runPlanSkillObservation now re-sends the skill command up to 3x (6s apart),
re-marking the buffer each time so stale scrollback can't re-trip the check,
before concluding the skill is genuinely unregistered. A real dangling-symlink /
missing-skill still surfaces as 'exited' (after retries), preserving the original
diagnostic. Pure-helper contract unchanged: 95/95 unit tests pass.

This is a pre-existing harness bug (fails identically on #2077's own branch, which
introduced the suite) surfaced while shipping the activation feature.
This commit is contained in:
Garry Tan 2026-06-24 07:59:22 -07:00
parent 48254d7830
commit 7472d99031
No known key found for this signature in database
GPG Key ID: C1F69E85C74EFE1D
1 changed files with 18 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
test/helpers/claude-pty-runner.ts
View File

@ -1605,7 +1605,7 @@ export async function runPlanSkillObservation(opts: {
// command.
await Bun.sleep(3000);
}
const since = session.mark();
let since = session.mark();
session.send(`/${opts.skillName}\r`);
const budgetMs = opts.timeoutMs ?? 180_000;
@ -1619,6 +1619,7 @@ export async function runPlanSkillObservation(opts: {
// even if the current state is 'working'.
let proseAUQEverObserved = false;
let waitingEverObserved = false;
let unknownCmdRetries = 0;
const JUDGE_AFTER_MS = 60_000;
const JUDGE_INTERVAL_MS = 30_000;
while (Date.now() - start < budgetMs) {
@ -1634,9 +1635,24 @@ export async function runPlanSkillObservation(opts: {
};
}
if (visible.includes('Unknown command:')) {
// Skill-discovery race: in a cold CI container the overlay-FS scan of
// the symlinked ~/.claude/skills registry can finish AFTER the boot
// grace, so the first `/skill` send reaches claude before the skill is
// indexed and is rejected as unknown. (Reproduces only in CI — passes
// locally where claude is warm and discovery is instant.) Re-send the
// command a few times, re-marking first so the stale "Unknown command:"
// line in scrollback can't immediately re-trip this check, before
// concluding the skill is genuinely unregistered.
if (unknownCmdRetries < 3) {
unknownCmdRetries++;
await Bun.sleep(6000);
since = session.mark();
session.send(`/${opts.skillName}\r`);
continue;
}
return {
outcome: 'exited',
summary: `claude rejected /${opts.skillName} as unknown command (skill not registered in this cwd)`,
summary: `claude rejected /${opts.skillName} as unknown command after ${unknownCmdRetries} retries (skill not registered in this cwd)`,
evidence: visible.slice(-2000),
elapsedMs: Date.now() - startedAt,
};