Skip to content

v1.93.34.0 test: docsync fault replays and the runShadow deadline test fit loaded remote VMs (GST-1103) - #3896

Merged
capy-ai[bot] merged 4 commits into
mainfrom
capy/gst1103-docsync-timing
Nov 3, 2026
Merged

capy-ai[bot] merged 4 commits into
mainfrom
capy/gst1103-docsync-timing

Conversation

@garrytan

@garrytan garrytan commented Nov 2, 2026 •

Copy link
Copy Markdown
Owner
{
  "schema_version": 1,
  "pr": 3896,
  "head": "766510b8d69edb8373ff6028c10d80d1658bdad9",
  "version": "1.93.24.0",
  "base": "main",
  "gate": {
    "focused": {
      "pass": 51,
      "fail": 0,
      "skip": 0
    }
  },
  "session_kind": "unattended",
  "head_tree": "9e6496a231df69b8209308f90785e31bdb344a4e",
  "product_patch_id": "none",
  "gstack_version": "1.93.24.0",
  "review_depth": {
    "depth": "light",
    "class": "tests",
    "runtime_lines": 0,
    "rules": "b2272efd321ce7cb",
    "base": "e85005482addbd902d2557a3deb964b29b0335bb",
    "registry": "e85005482addbd902d2557a3deb964b29b0335bb",
    "gate_scan": {
      "tagged": 0,
      "head": "766510b8d69edb8373ff6028c10d80d1658bdad9"
    },
    "reasons": [
      "runtime=0"
    ],
    "risk": []
  },
  "pregate": "requires-remote cleared @9e6496a231df (1 checks, current for 766510b8d69e)",
  "requires_remote": [
    "free-tests.yml/macos-named-regressions: https://github.com/garrytan/gstack/actions/runs/41549880187/attempts/1",
    "pregate.json/remote_gate: remote-gate tree=9e6496a231df tests=3491"
  ],
  "issues": {
    "fixes": [],
    "refs": [],
    "flakes": []
  },
  "docs": {
    "status": "current",
    "scope": "docs/, README.md, CONTRIBUTING.md, AGENTS.md (read-only)"
  },
  "review_loop": {
    "passes": 1,
    "blocking_found": 0,
    "deferred": 0,
    "noted": 0,
    "fixed": 0,
    "scope": "frozen",
    "rounds": 1
  }
}

What changed

Docsync fault wrappers now take a budget derived from the replay count, and the runShadow deadline tests leave room for a loaded VM

Docsync fault wrappers now take a budget derived from the replay count, and the runShadow deadline tests leave room for a loaded VM.

Before/after as text
-const r = await shadowed('deadline-retry', [{ mode: 'missing', delay_ms: 1200 }, { mode: 'valid', delay_ms: 1200 }], undefined, 1500);
-expect(r.ms).toBeLessThan(2400);
-const slow = await shadowed('deadline-call', [{ mode: 'valid', delay_ms: 3000 }], undefined, 800);
-expect(slow.ms).toBeLessThan(2500);
-stdout: 'pipe', stderr: 'pipe', timeout: 120_000,
-}, 120_000);
+const r = await shadowed('deadline-retry', [{ mode: 'missing', delay_ms: 2500 }, { mode: 'valid', delay_ms: 3400 }], undefined, 3500);
+expect(r.ms).toBeLessThan(5900);
+const slow = await shadowed('deadline-call', [{ mode: 'valid', delay_ms: 6000 }], undefined, 800);
+expect(slow.ms).toBeLessThan(5000);
+}, 15_000);
+stdout: 'pipe', stderr: 'pipe', timeout: callbackBudgetMs(count),
+}, callbackBudgetMs(count) + 10_000);
  • Each generated docsync fault-callback child test now has its own 30 second limit (CALLBACK_TEST_MS) instead of Bun's 5 second default.
  • The four wrapper tests that spawn those children get a budget of (replays + 1) × 30 s from callbackBudgetMs(), so the outer limit always covers the inner ones.
  • The two runShadow deadline tests in test/autoplan-shadow.test.ts keep the same assertions (a retry stops at the deadline; a slow call is cut off) with wider gaps between the delays, the deadline and the bound.

Why

Remote pregates on loaded VMs went red in these files while the code under test was fine (GST-1103). The GSTA-234 remote log shows the deadline-retry case finishing in 2,558 ms against a 2,400 ms bound, and the fault-callback replays hit Bun's 5 second default under load. With 8 busy loops on 4 cores, main failed all 12 legacy-control replays; this branch passed them.

Blast radius

Test files, one test helper and a changelog fragment only; no product, gate or workflow code changes. The assertions still check the same behavior, only the time each test may take changed, and four targeted remote runs plus one selected-tier remote gate passed with 0 failures.

Approvals

No approvals recorded for branch capy/gst1103-docsync-timing.

Summary

Docsync fault-callback tests get per-replay limits and wrapper budgets derived from the replay count, docsync-report-interface gets the same limits, and the runShadow deadline tests get margins that hold on a loaded VM.

Documentation

Status: current. The change only raises test timeouts, and no authored doc goes stale.

Scope: base f1cbdbb..HEAD 766510b on capy/gst1103-docsync-timing: 7 committed files (6 tests plus changelog.d/capy-gst1103-docsync-timing.md). Nothing staged, unstaged or untracked. The changed tests were cross-referenced against the candidate's doc inventory, including docs/TEST_PORTFOLIO.md and docs/TESTING_INTERNALS.md.

Documentation health:

  • docs/TEST_PORTFOLIO.md: Current (lists docsync-report-interface by name only)
  • docs/TESTING_INTERNALS.md: Current (does not describe the changed timeouts)
  • changelog.d/capy-gst1103-docsync-timing.md: Reviewed (the parent owns it; it matches the diff)

Coverage debt: none. The only new export is the test helper CALLBACK_TEST_MS/callbackBudgetMs, which is not public surface.

Diagram drift: none.

Test Coverage

The changed tests are the coverage: test/docsync-fault-interface{,-legacy,-faults}.test.ts, test/docsync-report-interface.test.ts and test/autoplan-shadow.test.ts, with the shared limits in test/helpers/docsync-fault-callbacks.ts.

Pre-Landing Review

REVIEW_LOOP: passes=1 blocking_found=0 deferred=0 noted=0 fixed=0 scope=frozen

  • Pass 1 (light, one reader, class tests): no actionable findings.
  • Outside voice (openai/gpt-6.1-sol): clean.

Eval Results

None; this change touches no prompt or skill text.

Scope Drift

None.

Issues

ISSUE_LINKS: none. Tracked as GST-1103.

Verification Results

  • Load repro (8 busy loops on 4 cores): main failed all 12 legacy-control replays at 5 s; this branch passed them.
  • Targeted remote runs (scripts/ubicloud/pregate-remote.sh, us-east-a2) on tree ed76b500: three runs of 51 tests, 0 failed each.
  • Remote pregate on tree 9e6496a2: 3,491 tests, 0 failed; macos-named-regressions cleared with CI run 41549880187.
  • CI at 766510b: 27 passed, 0 failed.

TODOS

None.

Test plan

  • Focused lane (bun test --timeout=30000 over 5 files): 51 tests ran, exit 0.
  • Remote pregate (gstack-pregate --stage tests --remote): 3,491 tests, 0 failed on tree 9e6496a2.

Open in Capy

…d remote VMs (GST-1103)

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

trunk-io Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

😎 This pull request was merged.

garrytan and others added 2 commits November 2, 2026 23:50
…s (GST-1103)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
@capy-ai
capy-ai Bot marked this pull request as ready for review November 3, 2026 00:17
@github-actions github-actions Bot changed the title test: docsync fault replays and the runShadow deadline test fit loaded remote VMs (GST-1103) v1.93.34.0 test: docsync fault replays and the runShadow deadline test fit loaded remote VMs (GST-1103) Nov 3, 2026
@capy-ai
capy-ai Bot merged commit 54606c5 into main Nov 3, 2026
45 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