Skip to content

[STALE: REBASE FROM NEW FORK] test: run client-connect.test.ts in its own process (shard 4/4 batch 8 stall) - #42

Open
vadymhimself wants to merge 1 commit into
release/2.65.0-vadymhimself.15from
gauge/age3241-client-connect-serial
Open

vadymhimself wants to merge 1 commit into
release/2.65.0-vadymhimself.15from
gauge/age3241-client-connect-serial

Conversation

@vadymhimself

Copy link
Copy Markdown
Collaborator

What

Adds clients/client-connect.test.ts to SERIAL_FULL_SUITE_FILES in scripts/test.ts, so the batch runner gives it a fresh Bun process.

Why

Test shard 4/4 timed out at the 120 s batch limit in batch 8 on recent release-line PRs. The last passing case is client-connect.test.ts:554; the next one (:577, "transaction fixture stops a child retained after valid output") keeps its child alive on purpose and relies on a blocking spawnSync timeout to kill it. While spawnSync blocks, Bun's per-test timeout cannot fire. Every file in the batch passes alone, so the stall needs earlier files' process state; a fresh process removes that. Same quarantine reason as the existing serial entries (see tests/helpers/cold-spawn-warmup.ts:216-224).

Containment, not a rewrite: making that fixture async would touch about 30 callers in a size-ratcheted file.

Proof wanted

Shard 4/4 green on two runs in a row.

Eval added/updated: N/A test-runner manifest only, no product behaviour change

🤖 Generated with Claude Code

The transaction fixture bounds a child with a blocking spawnSync timeout; in a shared
Bun process after earlier files, shard 4/4 batch 8 stalled to the 120 s batch limit while
every file passes alone. Add it to SERIAL_FULL_SUITE_FILES so it runs as a one-file batch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@vadymhimself vadymhimself changed the title test: run client-connect.test.ts in its own process (shard 4/4 batch 8 stall) [STALE: REBASE FROM NEW FORK] test: run client-connect.test.ts in its own process (shard 4/4 batch 8 stall) Nov 18, 2026
@vadymhimself

Copy link
Copy Markdown
Collaborator Author

This PR is based on the superseded 2.65 gateway line. Rebuild the change from port/ar-on-upstream-2026-11-18 (upstream-derived 2.100 AR overlay, PR #46): #46 . Do not rebase onto or merge the old release line. Preserve the new gateway quota/cache gate and re-run the affected regressions. The candidate is not deployed; do not publish or replace a running gateway without explicit approval.

@vadymhimself

Copy link
Copy Markdown
Collaborator Author

main is now the stable upstream-derived AgentRadar gateway line: 2.100.0-vadymhimself.2, merged via #47 (merge commit ad594a0778). Its tree is identical to the tested candidate 6a14467972, and it shares history with upstream lidge-jun/opencodex dev at 6bae884cbc.

Trial evidence: before the merge the build ran 30.5 h on the laptop gateway and 24 h on the personal Mac Mini gateway, both with OPENCODEX_AGENTRADAR=1. Neither restarted and neither left unresolved spend. The Mini served 11,138 requests with 2 5xx; the old build had 303 502s over the prior day. The remaining non-2xx were exhausted provider quota, not gateway faults. The fork's GitHub Actions has not run since Nov 18, so there is no CI result. Run checks locally.

For new work:

  • Branch from main. Do not target release/2.65.0-vadymhimself.*.
  • Check FORK-DELTA.md first; many old-line fixes are already carried or upstream-equivalent.
  • If a behavior is still missing, re-cut it as a small, single-concern change on main.

Before opening a PR, use pinned Bun 1.4.4 with an isolated HOME/OPENCODEX_HOME, synthetic credentials and loopback providers (never live accounts):

  • bun run test:gateway-quota (29-file AR gateway gate)
  • affected suites: bun scripts/test.ts --changed=main
  • bun run typecheck, bun run structure:check, bun run privacy:scan, git diff --check

Hosted-only behavior needs OPENCODEX_AGENTRADAR=1 set inside the test; without it the branch behaves like upstream. Register new test files in scripts/test-layout/layout-agentradar.json.

For this PR: upstream test layout and sharding changed. Re-check whether the client-connect stall reproduces on main before re-cutting.

🤖 Generated with Claude Code

This branch has not been deployed

No deployments
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