Repository navigation
mcpx: internal/cli package test duration exceeds the 600s stress timeout on PR 2714 (stage 1 of #1980) #2859
Description
Problem
On PR #2714 (stage 1 of #1980, head 61a3d0f8), the stress check still fails on the
internal/cli package — even after the plumbing-source leak fix in that head
(scoped SetPlumbingSource per cli.Run invocation + regression test, A/B verified
on the laptop). The failure is aggregate package duration under the stress
configuration, not a single hanging test.
Evidence
- CI run 41140869569, job 124354518386 ("stress: PR changed tests (race/shuffle/repeat)"):
FAIL github.com/dezren39/mcpx/internal/cli 600.093s panic: test timed out after 10m0s; the only test still running at the alarm was
TestEverySettingFlagReachesTheResolvedValue (9s)— i.e. the preceding tests in the
race/shuffle/repeat run consumed the budget, the package total hit the 600s cap.- All other PR feat(mcpx): cli.Run in-process entrypoint, slim main (#1980 stage 1) #2714 checks pass (basic: lint, plugin tests, nix build .#mcpx-docs,
CI time budget). Only the stress leg is red, and it blocks the stage-1 merge. - Filed by queue worker lane A (task 3275) per the previous shift's recorded lead:
"if cli still times out with the plumbing fix, file the cli-package duration issue".
Acceptance
- Identify the top duration contributors in
pkgs/mcpx/internal/cliunder the stress
invocation (race + shuffle + repeat) — per-test timings, not guesses. internal/clipasses the stress check within the 600s package timeout on PR feat(mcpx): cli.Run in-process entrypoint, slim main (#1980 stage 1) #2714
(or on the follow-up PR that carries the duration fix), without weakening or
deleting tests to get there.- If the root cause is shared global state slowing tests under repetition (the
plumbing leak was one instance), the fix names the mechanism, not just the symptom.
Refs: #1980, PR #2714, queue task 3275.
Acceptance criteria
- The
internal/clipackage completes within the 600s stress timeout on the PR stress configuration (race/shuffle/repeat), or the package is split/throttled with the aggregate-duration budget documented. - The fix targets aggregate duration (the evidenced cause), not a single hanging test; before/after package timings are recorded on the PR.
- No test is deleted or skipped to achieve the pass without an equivalent replacement.
Metadata
Metadata
Assignees
Labels
Package/app: mcpx (pkgs/ or apps/ in this repo)Medium: expected under 1 day of workCI/CD pipelines, workflows, and build automationIntake review done: acceptance criteria, links, and normalized labels addedImportant and time-sensitiveQueue worker lifecycle stateFiled or worked by the queue/worker pilotQueue worker lifecycle stateQueue packet kind: verify, reproduce, and route this itemQueue worker lifecycle state
Activity
dezren39 commentedon Oct 31, 2026
Intake review
Related: PR #2714 (stage 1 of #1980); CI run 41140869569. Note the fleet context: task 3823/#2573 records the same stress leg red on main from #2859 — this issue is currently the named blocker for other merges' stress signal. Risks: sharding the package changes CI topology; measure first. Suggested approach: profile per-test durations in internal/cli under the stress config, split or budget the slowest group, verify with a full stress run.
— Fleet monitor intake enrichment (protocol §5): verified still live this run (no merged PR resolves it), labels normalized, acceptance criteria appended. Marked
meta/enriched.mcpx(cli): batch the flag-reach probe to one invocation per command (#…
dezren39 commentedon Nov 3, 2026
Mechanism found and fixed on PR #2714 (d775939, task 3275): the reach test's per-pair handler invocations each rebuilt the settings schema twice — O(commands x flags) builds, 14.45s of one no-race package run. Batched to ~3 invocations per command with per-flag fallback for refusals/collisions; 0.62s now, assertions unchanged. Stress leg re-running on the PR; closing this as fixed once it goes green, or follow up there if anything else dominates.
developing-today commentedon Nov 3, 2026
agent-note:v1
state: reviewed (coverage loop-back)
Coverage pass 2026-11-03, loop-back review: this issue was re-reviewed (drift / overlap / already-handled / stale / blockers / ordering) so every open issue in the four repos is reviewed and in the system. Verdict: still live; in the system as task 7366 (triage, queued, priority 50) in repo-health-code-pilot. Labels audited and synced. If it is already handled at current main, its triage closes it with evidence; if oversized, the triage files the split; blockers belong on this issue.
Found via: coverage loop-back 2026-11-03
dezren39 commentedon Nov 4, 2026
agent-note:v1
state: picked-up
queue: repo-health-code-pilot
task: 7366 (triage, worker repo-worker-code, lane 7b47536b-c162-444c-be8a-3a9538cc4cb5)
by: openchamber-scheduled-lane
at: 2026-11-04T03:16:40Z
Resuming the previous OpenCode handoff and checking current evidence. Issue #2859 remains OPEN, and its canonical follow-up PR #2714 is also OPEN. The issue records the suspected aggregate-duration mechanism and a large local improvement (14.45s to 0.62s for the identified reach test), but the current PR head 2787b48 has only CodeRabbit and production-line-guard checks; no stress run is attached to verify the 600-second acceptance criterion. I will record this blocker without opening a competing fix.
dezren39 commentedon Nov 4, 2026
agent-note:v1
state: triaged
queue: repo-health-code-pilot
task: 7366 (triage, worker repo-worker-code, lane 7b47536b-c162-444c-be8a-3a9538cc4cb5)
by: openchamber-scheduled-lane
at: 2026-11-04T03:17:39Z
Current disposition: issue #2859 remains OPEN; canonical fix path is the already-open PR #2714. The prior change reports the identified reach-test path dropping from 14.45s to 0.62s without weakening assertions. However, the current PR head 2787b48 has no stress-check run; only CodeRabbit and production-line-guard are visible and pass. Therefore the 600-second stress acceptance is still unverified and is the blocker to closing #2859. No duplicate branch, PR, or follow-up task was created.