[fixtures] Skip dev-registry tests on Windows - #15018
Conversation
These tests are consistently failing on Windows in CI and are blocking other work. Skip them on Windows as a temporary measure until the underlying issue is fixed. There is still value in running them on macOS and Linux.
|
|
All five The change is correct:
The author's own PR description already thoroughly documents the tradeoffs (blanket skip masks future Windows regressions, TODO not enforced). These are acknowledged design decisions for a temporary unblock, not bugs. The implementation is correct and minimal. No logic bugs, security issues, backward-compat violations, or incorrect API behavior found. LGTM |
dario-piotrowicz
left a comment
There was a problem hiding this comment.
Thanks @petebacondarwin 🙏
workers-devprod
left a comment
There was a problem hiding this comment.
Codeowners reviews satisfied
@cloudflare/autoconfig
@cloudflare/build-output-utils
@cloudflare/config
create-cloudflare
@cloudflare/deploy-helpers
@cloudflare/kv-asset-handler
miniflare
@cloudflare/pages-functions
@cloudflare/pages-shared
@cloudflare/unenv-preset
@cloudflare/vite-plugin
@cloudflare/vitest-pool-workers
@cloudflare/workers-auth
@cloudflare/workers-editor-shared
@cloudflare/workers-utils
wrangler
commit: |
|
Codeowners approval required for this PR:
Show detailed file reviewers |
|
Let's merge this one for today, so that we can get the release out and unblock other PRs. |
Reverts the temporary skip from #15018, which disabled the whole suite on Windows to unblock other work "until the underlying issue is fixed". The preceding commit fixes it: the timeouts were a workerd abort during teardown (cloudflare/workerd#6913), triggered by killing a dev session that a still running session was forwarding tail events to, and the fixture no longer sets that up. Windows is the only platform the dev registry flake ever appeared on, so leaving these skipped there would mean the suite never guards the thing it exists to guard.
Measures the baseline Windows flake rate under the same 6x harness. Reverts both the teardown fix and #15018's Windows skip, so the original cyclic tail_consumers are back and the tests actually execute on Windows.
Reverts the temporary skip from #15018, which disabled the whole suite on Windows to unblock other work "until the underlying issue is fixed". The preceding commit fixes it: the timeouts were a workerd abort during teardown (cloudflare/workerd#6913), triggered by killing a dev session that a still running session was forwarding tail events to, and the fixture no longer sets that up. Windows is the only platform the dev registry flake ever appeared on, so leaving these skipped there would mean the suite never guards the thing it exists to guard.
The
fixtures/dev-registrysuite is consistently failing on Windows in CI and is blocking unrelated work. This skips it on Windows as a temporary measure to unblock the pipeline, while keeping full coverage on macOS and Linux.The fixture has five top-level
describeblocks, so rather than wrapping the whole file in a platform conditional (asfixtures/interactive-dev-tests/tests/index.test.tsdoes), this aliases the vitest import and derives a locally-skippeddescribe:That covers all five blocks in a five-line diff with no re-indentation, which keeps the eventual revert trivial.
Verification:
describeblocks — no top-levelit/testescapes the skip.truelocally reports27 skippedand exits 0, so Windows goes green rather than erroring on an empty test file.vitest liststill enumerates all 27 tests on macOS.oxfmt --checkandcheck:typeboth pass. Lint is not applicable —.oxlintrc.jsoncignoresfixtures/**.Note that this is a blanket file-level skip, so any new Windows regression in the dev registry will also go unnoticed on that platform, not just the current failure. The
TODOmarks it as temporary but nothing enforces cleanup, so this should be reverted once the root cause is fixed.A picture of a cute animal (not mandatory, but encouraged)