Skip to content

[notme-18450e] fix(ci): run the publish gate's tests from worker/, not the root#44

Merged
jamestexas merged 1 commit into
mainfrom
fix/publish-dpop-vitest-resolution
Jul 23, 2026
Merged

[notme-18450e] fix(ci): run the publish gate's tests from worker/, not the root#44
jamestexas merged 1 commit into
mainfrom
fix/publish-dpop-vitest-resolution

Conversation

@jamestexas

Copy link
Copy Markdown
Contributor

Second publish attempt failed: ERR_PNPM_RECURSIVE_EXEC_FIRST_FAIL Command "vitest" not found.

vitest is a devDependency of the worker package only — the repo root has none. pnpm vitest at the root resolves fine locally through hoisting, which is exactly why this passed every local check and failed in a clean CI install. Local success was evidence about my node_modules, not about the workflow.

worker's vitest config globs ../packages/dpop/__tests__, so running the suite from worker/ covers the SDK's own 67 tests and proves worker still resolves the package through its bare specifier — a stronger gate than testing the package in isolation.

Verified: cd worker && pnpm vitest run → 23 files / 339 pass; pnpm --filter @agentic-research/dpop build → exit 0.

Both failures so far were caught before the publish step, which is the design working — an npm version is immutable once it lands, so failing early costs a re-tag and nothing else.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UNahXs8dda3vttpWrVwVaA

…t the root

Second publish attempt failed: ERR_PNPM_RECURSIVE_EXEC_FIRST_FAIL,
Command "vitest" not found.

vitest is a devDependency of the worker package only — the repo root has
none. `pnpm vitest` at the root resolves fine LOCALLY through hoisting,
which is exactly why this passed every local check and failed in a clean
CI install. Local success was evidence about my node_modules, not about
the workflow.

worker's vitest config globs ../packages/dpop/__tests__, so running the
suite from worker/ covers the SDK's own 67 tests AND proves worker still
resolves the package through its bare specifier — a stronger gate than
testing the package in isolation.

Both failures so far have been caught by the gate BEFORE the publish step,
which is the design working: an npm version is immutable once it lands, so
failing early costs a re-tag and nothing else.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@jamestexas
jamestexas merged commit a8520c1 into main Jul 23, 2026
2 of 3 checks passed
@jamestexas
jamestexas deleted the fix/publish-dpop-vitest-resolution branch July 23, 2026 03:30
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