ci: Reduce test runner startup overhead - #334833
Conversation
Run Agent Host E2E beside the remaining Electron integration tests, split Electron smoke coverage into core and chat shards, and isolate the macOS proxy scenarios from the main smoke critical path. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.
Note
This error may be related to your runner configuration. You can now configure runners for Copilot code review separately from Copilot cloud agent by creating a copilot-code-review.yml file with your setup steps. Read the docs for details.
Skip redundant prelaunch work on prepared CI runners, avoid Browser-only Copilot compilation, disable login-shell probing on test servers, and reduce Browser unit console output while preserving JUnit results. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.
Note
This error may be related to your runner configuration. You can now configure runners for Copilot code review separately from Copilot cloud agent by creating a copilot-code-review.yml file with your setup steps. Read the docs for details.
roblourens
left a comment
There was a problem hiding this comment.
[Experimental performance review bot]
Automated experimental performance review.
(Written by Copilot)
|
Going to do some more work on this as feedback is valid, the current refactoring may be putting more pressure on the pool while the wall time win is small. |
Run Agent Host E2E concurrently inside the existing Electron integration jobs and recombine core and chat smoke coverage, eliminating repeated checkout, dependency restore, transpilation, and Electron setup. Keep only the macOS network proxy smoke job isolated because it changes global network state. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This reverts commit 5612608.
Summary
Expected impact
A focused Browser smoke run took 15.4 seconds with prelaunch synchronization and 12.2 seconds without it. Browser jobs start 17-18 test servers, projecting approximately 50-60 seconds saved per Browser job.
Skipping Browser-only Copilot compilation removes another 5-8 seconds of unnecessary work. The Browser unit reporter benchmark reduced console output from 2,761 to 263 lines and from 177KB to 80KB for the same 2,700-test subset.
The workflow job matrix and test coverage are unchanged.