Open squell opened 3 hours ago
So it's not consistently the same test that fail, but I do see a pattern that they are time-stamp related (although sometimes a signal test fails as well). That might be due to poor performance of the CI.
Found the culprit: PR #887 completely broke the delicate /proc/stat
parser we had.
After the merger of #887 (which passed the CI while being 'up to date' with
main
, which is now the rule since the merge queue was disabled), CI onmain
seems to be consistently failing.Two possibilities:
We have made the CI more flaky by removing all the caching.