Skip to content

ci: report all workspace tests in hosted CI - #202

Merged
nishuzumi merged 2 commits into
nishuzumi:mainfrom
Anand-0037:anand/fix-ci-isolation
Sep 22, 2026
Merged

nishuzumi merged 2 commits into
nishuzumi:mainfrom
Anand-0037:anand/fix-ci-isolation

Conversation

@Anand-0037

@Anand-0037 Anand-0037 commented Sep 15, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Addresses #195.

Hosted CI stopped at the first failing workspace package, which could hide results from later packages when an unrelated live test failed.

This patch:

  • changes the Linux hosted test step to pnpm -r --no-bail test
  • keeps local pnpm test behavior unchanged
  • documents how to distinguish baseline failures from regressions

#200 already removed the expired Aave quarantine that originally exposed #195. This patch does not change Aave, Kuru, package code, or the live verification workflow.

Verification

Local:

  • pnpm build: passed
  • pnpm typecheck: passed
  • pnpm lint: passed
  • env MOSS_SKIP_E2E=1 pnpm -r --no-bail test: passed with all workspaces reporting
  • synthetic early failure: later workspaces still ran and the command exited non-zero
  • workflow YAML parsing: passed
  • git diff --check: passed

Hosted CI:

  • lint, build, and typecheck passed
  • Aave failed first, but later workspace tests continued to run
  • final recursive result: 3 failing workspaces, 18 passing
  • the test command still exited non-zero
  • windows-offline passed

The live failures are outside this patch. Aave is tracked in #201 and Kuru in #194. The no-bail run also surfaced a Merkl request failure that was previously hidden behind the earlier failure.

No package or lockfile changes.

@Anand-0037
Anand-0037 force-pushed the anand/fix-ci-isolation branch 2 times, most recently from ce73091 to 3fc1162 Compare September 18, 2026 07:54

@nishuzumi nishuzumi left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed 3fc1162 plus my one-line follow-up. pnpm -r --no-bail test in the Linux job is the right fix for the first-fail bail: the hosted run on this head reported all 21 workspaces (3 failing, 18 passing) instead of stopping at Aave, and the job still exits non-zero. Local pnpm test is unchanged; the CONTRIBUTING note tells contributors how to read a red job whose failing package is not in their diff.

The no-bail run immediately paid for itself: it surfaced that the Merkl live suite fails on the hosted runner with Merkl rewards request failed: fetch failed. That is the egress allowlist, not the adapter — harden-runner blocks api.merkl.xyz, which #187 needed and which the bail had hidden. I pushed api.merkl.xyz:443 into allowed-endpoints on this branch since it is the same file and the same "make hosted CI report the truth" concern.

This closes the bail half of #195. The tripwire half — expiry checks living in a scheduled live-verify run rather than in every PR's gate — is still open there.

@nishuzumi
nishuzumi merged commit 7943745 into nishuzumi:main Sep 22, 2026
1 of 2 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.

2 participants