Repository navigation
Description
Is your feature request related to a problem? Please describe.
Since #875, build-and-test no longer waits for clippy, and since #718 the E2E gate opens once build-and-test has finished with success or failure. A run where clippy fails therefore still spends a full E2E job (about 5 to 7.5 minutes of runner time on the runs I looked at) on a change that the required clippy check will block regardless. In the sample #875 used, clippy failed in 10 of 192 completed pull request runs, so the saving per run is small and this is only worth doing if the post-merge measurement shows the same.
Describe the solution you'd like
Add clippy to the E2E job's needs and open the gate only when clippy is success or skipped, the pattern pr-linux-binaries already uses. clippy finished in 1.5 to 4 minutes and build-and-test in about 8.5 minutes on the green runs of #875, so waiting for clippy should add no wall-clock time when lint is green. The workflow contract test that pins the E2E gate (xtask/src/workflow_contract.rs) would need a matching pin for the new condition, and the comments on the gate and on build-and-test would need updating.
Describe alternatives you've considered
- Do nothing. A clippy-red run keeps spending the E2E runner time, and the required
clippycheck still blocks the merge. This is a fine outcome if the measurement shows few lint-red runs reach E2E. - Skip E2E for the lint-red case another way, for example by re-running
clippyinside the E2E job. That costs more than reading its result.
Additional context
The E2E check is required, so on a clippy-red run it would report green without having run any scenarios. That is harmless while clippy is also required, but it is the same green-but-empty shape that the gate's own comment (from #718) calls out for a failed matrix leg, so it should be a deliberate choice rather than a side effect.
Suggested sequencing: do this after the post-merge measurement promised in #875, which counts how many lint-red runs now continue into build-and-test and E2E. It edits the same gate that #871 may touch, so the two should be ordered rather than done in parallel.
Raised as a non-blocking suggestion in the automated review of #875.
Activity