You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Investigate workflow job failing on main: e2ePerformanceTests / Find the baseline and delta refs, and check for an existing build artifact for that commit #54478
hm, for some reason when that PR got merged the job tried to look for a PR with your user instead of using the PR number. Maybe there is something broken about that logic (I think the PR number should've been used)
While the above error might have been a GitHub hiccup, it seems that the e2e tests are still broken though.
All runs are currently failing after ~ 1h45m on AWS.
Checking whats wrong with that …
@perunt it seems that the tests are failing due to this error in the linking test:
@kirillzyusko I know that we have a mechanism for marking a test as failed in a way that it won't fail the whole pipeline, but just the faulty test. Can you please let us know what the API for that is? Then @perunt can open a PR to fix that not the whole pipeline fails.
After that we should probably investigate why that error is happening though.
Yes, we have a fault tolerance mechanism, but it's used to safely handle flaky tests I believe.
But this test explicitly fails with error:
E2EClient.submitTestResults({branch: Config.E2E_BRANCH,error: 'Linked message not found',name: `${name} test can't find linked message`,}).then(()=>E2EClient.submitTestDone());
And for such cases we don't have safe failure handling, because they are clearly indicate that it's something wrong with the test itself (i. e. it's not flakiness) 🤷♂️
And it forces us to investigate a reason instead of silently ignoring them 👀
This issue has not been updated in over 15 days. @hannojg eroding to Monthly issue.
P.S. Is everyone reading this sure this is really a near-term priority? Be brave: if you disagree, go ahead and close it out. If someone disagrees, they'll reopen it, and if they don't: one less thing to do!
Activity
marcaaron commentedon Dec 23, 2024
Looks like some kind of Http error on GitHubs end possibly
hannojg commentedon Dec 30, 2024
hm, for some reason when that PR got merged the job tried to look for a PR with your user instead of using the PR number. Maybe there is something broken about that logic (I think the PR number should've been used)
hannojg commentedon Dec 30, 2024
While the above error might have been a GitHub hiccup, it seems that the e2e tests are still broken though.
All runs are currently failing after ~ 1h45m on AWS.
Checking whats wrong with that …
hannojg commentedon Dec 30, 2024
@perunt it seems that the tests are failing due to this error in the linking test:
@kirillzyusko I know that we have a mechanism for marking a test as failed in a way that it won't fail the whole pipeline, but just the faulty test. Can you please let us know what the API for that is? Then @perunt can open a PR to fix that not the whole pipeline fails.
After that we should probably investigate why that error is happening though.
hannojg commentedon Dec 30, 2024
Issue is being investigated over here:
kirillzyusko commentedon Dec 30, 2024
Yes, we have a fault tolerance mechanism, but it's used to safely handle flaky tests I believe.
But this test explicitly fails with error:
And for such cases we don't have safe failure handling, because they are clearly indicate that it's something wrong with the test itself (i. e. it's not flakiness) 🤷♂️
And it forces us to investigate a reason instead of silently ignoring them 👀
melvin-bot commentedon Jan 23, 2025
This issue has not been updated in over 15 days. @hannojg eroding to Monthly issue.
P.S. Is everyone reading this sure this is really a near-term priority? Be brave: if you disagree, go ahead and close it out. If someone disagrees, they'll reopen it, and if they don't: one less thing to do!
hannojg commentedon Feb 17, 2025
This is fixed by now, can we please close this @mountiny (I don't have the permission for some reason)