Skip to content

v1.91.166.0 test: warm PowerShell 5.1 outside the setup-ps1 legacy cases; xvfb round trip on a private display range (GSTA-57) - #3324

Merged
capy-ai[bot] merged 3 commits into
mainfrom
flaky/2026-10-20-free-load
Oct 20, 2026
Merged

capy-ai[bot] merged 3 commits into
mainfrom
flaky/2026-10-20-free-load

Conversation

@garrytan

Copy link
Copy Markdown
Owner

Daily flaky-fix pass (GSTA-57), group 6: free and load-sensitive tests. These are two timing and isolation repairs. No test timeout, case or assertion was loosened.

test/setup-ps1.test.ts, PowerShell 5.1 legacy cases ("space in TMPDIR", "empty argument refused"; windows-setup on 2026-10-17 job 117591782811, and run 39351736921 attempt 1, job 118397262881). windows-setup-e2e.yml runs the file with plain bun test, so each case gets Bun's 5 s default. The first powershell.exe launch on a runner pays a cold start that later launches don't. Across 94 windows-setup jobs:

  • The first legacy launch took p50 1.46 s, p95 3.42 s and max 5.04 s, and two passing jobs took 4.88 s and 4.99 s.
  • The second launch had p95 1.41 s.
  • Both reds are the first launch killed at 5 s with exit null.

A beforeAll now runs the legacy interpreter once on the stub checkout, with a 60 s hook budget that matches the spawns' own 60 s bound. To reproduce it on Linux, I used pwsh 7.4.6 in Legacy mode under plain bun test:

  • A wrapper that adds 6 s to the first launch reproduces the CI symptom on main and passes with the warm-up.
  • A wrapper that adds 6 s to every launch still fails 3 cases (negative control: the warm-up hides only a cold start).

The file triggers the windows-setup workflow, so this PR's CI exercises real PS 5.1.

browse/test/xvfb.test.ts, spawn → cleanup round trip (Ubicloud full run, 2026-10-18). The test picked a display from :99-:110 and then spawned on it, inside the default range that headed daemons in parallel shards allocate from. With six default-range allocators running, main failed 9/20 and 5/20 with "X display :N was reserved by another X server during startup". The case now uses :25000-:25100, which no other allocator uses, the same convention as the file's other private-range cases. If no display is free it now fails instead of silently returning. Under the same load it passed 20/20, 20/20 and 10/10. The ownership, lock and refusal assertions are unchanged.

Diagnosed and left open, with no edits here:

Checks: test/setup-ps1.test.ts and browse/test/xvfb.test.ts pass (37 pass, 34 skip on Linux). typecheck, the typecheck:test ratchet, the module-size ratchet and the spawnSync tripwire pass. No paid cases are affected.

Open in Capy

…trip uses a private display range (GSTA-57)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
@trunk-io

trunk-io Bot commented Oct 20, 2026 •

Copy link
Copy Markdown

😎 This pull request was merged.

garrytan and others added 2 commits October 20, 2026 15:03
…-load

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

# Conflicts:
#	CHANGELOG.md
…ses; xvfb round trip on a private display range (GSTA-57)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
@github-actions github-actions Bot changed the title test: warm PowerShell 5.1 outside the setup-ps1 legacy cases; xvfb round trip on a private display range (GSTA-57) v1.91.166.0 test: warm PowerShell 5.1 outside the setup-ps1 legacy cases; xvfb round trip on a private display range (GSTA-57) Oct 20, 2026
@capy-ai
capy-ai Bot merged commit a3ed43c into main Oct 20, 2026
62 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant