fix(test): reduce and widen the agent-sdk-driver real-git test timeout - #6871
Merged
Conversation
…out budget The real-git-subprocess test in test/unit/agent-sdk-driver.test.ts (init/config/config/ add/commit against a temp repo, then the driver's own diff enumeration) exceeds its explicit 30s timeout under concurrent full-suite load. Commit identity now goes through GIT_AUTHOR_*/GIT_COMMITTER_* env vars on the commit call instead of two separate `git config` subprocess spawns, cutting setup from 5 sequential git invocations to 3. The remaining spawns still wait on real OS process scheduling under load, so the timeout is also widened to 60s based on repeated measurement under simulated CPU contention (16 processes oversubscribing a 12-core machine): clean single-attempt runs ranged ~0.3-20s, with one outlier at ~54s and none exceeding 60s across 10 runs.
Contributor
|
Superagent didn't find any vulnerabilities or security issues in this PR. |
loopover-orb
Bot
added
the
gittensor:bug
24 tasks
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #6871 +/- ##
=======================================
Coverage 93.64% 93.64%
=======================================
Files 684 684
Lines 68193 68193
Branches 18707 18707
=======================================
Hits 63857 63857
Misses 3350 3350
Partials 986 986
Flags with carried forward coverage won't be shown. Click here to find out more. |
24 tasks
JSONbored
added a commit
that referenced
this pull request
Jul 25, 2026
* fix(deps): resolve postcss + tar audit findings via overrides npm audit flagged postcss <=8.5.17 (path traversal in previous-sourcemap auto-loading, GHSA-r28c-9q8g-f849, high) and tar <=7.5.20 (uncontrolled recursion DoS via crafted long-path tar, GHSA-r292-9mhp-454m, moderate), both transitive. Pin both via overrides, matching this file's existing pattern for other transitive-dep security fixes: bump the existing tar override past its first-patched version (7.5.19 -> 7.5.21) and add a new postcss override (8.5.18, first patched version). Closes task 1 of #8588's checklist. The remaining findings in that issue (eslint v10 bump for the brace-expansion/minimatch chain, the @esbuild-kit->tsx trace, and the recharts v3 migration) are unrelated, each need their own scoping, and are now tracked as sub-issues #8608, #8609, #8610. * build: add the missing ui-kit build step to the local test:ci gate Discovered incidentally while running the full local gate for #8588: test/unit/check-ui-kit-package.test.ts's own regression guard needs packages/loopover-ui-kit/dist/ to exist, but the root test:ci script never built that workspace before test:coverage -- unlike the real CI validate-tests job, which already builds it via `npx turbo run build --filter=@loopover/ui-kit` right before its own coverage step (added in #8592 to close this exact gap, but only in the CI workflow, not the local script this repo's own contributing guide points everyone at). Anyone running `npm run test:ci` from a clean checkout hit a spurious failure that real CI never showed. Adds `npm run build --workspace @loopover/ui-kit` alongside the other per-package builds already in the chain (engine, discovery-index, mcp, miner), so the local script matches CI again. * fix(mcp): stop CLI stdout truncation on >64KB piped output Discovered incidentally while running the full local gate for #8588: test/unit/mcp-cli-profiles.test.ts's changelog test started failing with a JSON parse error once packages/loopover-mcp/CHANGELOG.md grew past the OS pipe buffer size (verified: `loopover-mcp changelog --json | wc -c` returned exactly 65536 bytes instead of the real 65603+ -- the classic 64KB pipe-buffer boundary). Root cause: process.stdout/stderr writes to a POSIX pipe are asynchronous, but the CLI entrypoint called process.exit() immediately after the command's async work resolved, cutting off any pending write larger than one pipe buffer's worth before it finished flushing. This silently truncates any command's output over ~64KB for any real consumer piping the CLI (not just this test) -- e.g. `loopover-mcp changelog --json | jq` would get invalid JSON. Fix: drain both streams (wait for their internal buffer to empty) right before calling process.exit(), instead of changing when/whether process.exit() itself is called -- keeps the existing fast-exit behavior for everything else (no risk of hanging on a lingering fetch keep-alive socket) while fixing the truncation at its actual source. * test(miner): raise timeout for build:verify syntax-check regression guard Discovered incidentally while running the full local gate for #8588: this test spawns a real subprocess (node --check over every dist/bin+dist/lib file in packages/loopover-miner) -- 5.5s in isolation, but it exceeded the default 15s testTimeout once under the full suite's parallel load. Genuinely necessary real subprocess work with no redundancy to cut, matching this repo's established real-subprocess timeout-flake pattern (agent-sdk-driver.test.ts, miner-attempt-worktree .test.ts, miner-repo-clone.test.ts, #6869/#6871): raise to the same evidence-based 60000ms ceiling already used for that class of test, rather than reflexively widening without a reason.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
test/unit/agent-sdk-driver.test.ts's real-git-subprocess test (enumerates tracked and untracked worktree changes with git) spawnsgit init, twogit configcalls,git add,git commit, and the driver's own diff enumeration (two more, parallelized) against a real temp repo.534dad77alongside the sibling fix now being applied totest/unit/ai-summaries.test.tsin a separate PR).GIT_AUTHOR_*/GIT_COMMITTER_*env vars on the commit call instead of two separategit configsubprocess spawns, cutting setup from 5 sequential git invocations to 3.test/unit/ai-summaries.test.ts, separate PR) -- both tests were widened in the same historical commit for the same reason, and a full-suite verification run for that fix surfaced this one still flaking.Scope
type(scope): short summaryConventional Commit format.CONTRIBUTING.mdand does not reintroduce GitHub Pages, VitePress,site/, orCNAME.Validation
git diff --checknpm run actionlintnpm run typechecknpm run test:coveragelocally -- test-only change totest/**, which Codecov does not measure, so there is no patch-coverage obligation.npm run test:workersnpm run build:mcpnpm run test:mcp-packnpm run ui:openapi:checknpm run ui:lintnpm run ui:typechecknpm run ui:buildnpm audit --audit-level=moderateAdditional validation beyond the standard checklist: the fixed test was run 10x under simulated heavy CPU contention (16 CPU-bound processes oversubscribing a 12-core machine) with zero outright failures (see timing data above). A full local
npm run test:cirun also passed this file cleanly (agent-sdk-driver.test.ts, 9.3s, no retry needed), though that run also showedtest/unit/ai-summaries.test.tsfailing (expected -- its fix lives in a separate, not-yet-merged PR; this worktree branched before that fix existed) and surfaced two additional, unrelated flakes (test/unit/github-labels.test.ts,test/unit/github-pr-actions.test.ts) that only appeared while two full 18k-test suites were running concurrently on the same machine -- self-inflicted, unrealistic contention well beyond normal CI conditions. Flagging separately rather than expanding this PR's scope further.Safety
UI Evidencesection below. (Not applicable -- no visible/UI changes.)UI Evidence
Not applicable -- test-only change, no UI/frontend/docs surface touched.
Notes
test/unit/ai-summaries.test.tsflake, and a same-pattern timeout fix for the#5132miner clone/worktree test suite.