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
We are currently using CIFuzz in our Github CI. It turns out that at the end of each CIFuzz job that is kicked off by the Github Action, a followup job called ClusterFuzzLite/CIFuzz is also run. It was interfering with one of our one our CI workflows, which uses an action that is designed to intelligently enforce that CI checks are passing. See unicode-org/icu#3160 for our experience with the interference.
The reason for the interference is that, empirically, the ClusterFuzzLite/CIFuzz job starts off initially in a "neutral" state, but then appears to run to completion. The neutral state is interpreted as unsuccessful by the wait-for-status-checks job (example).
So the fact that the job ClusterFuzzLite/CIFuzz starts off reporting its status in a "neutral" state seems like a bug. Is there a way that it can start off with some type of "in progress" status of some kind instead? And only change the status to something else once the job conclusively ends successfully one way or another?
The text was updated successfully, but these errors were encountered:
I don't have any docs about how Github Actions interprets things. All I have is empirical conclusions based on observations. One of the links that I included above is PR unicode-org/icu#3160 above. In that PR, you can also look at the discussion and inline "code" comments to get a better sense of what happened and how the workaround needed to be designed to avoid those specific issues.
We are currently using CIFuzz in our Github CI. It turns out that at the end of each CIFuzz job that is kicked off by the Github Action, a followup job called
ClusterFuzzLite/CIFuzz
is also run. It was interfering with one of our one our CI workflows, which uses an action that is designed to intelligently enforce that CI checks are passing. See unicode-org/icu#3160 for our experience with the interference.The reason for the interference is that, empirically, the
ClusterFuzzLite/CIFuzz
job starts off initially in a "neutral" state, but then appears to run to completion. The neutral state is interpreted as unsuccessful by thewait-for-status-checks
job (example).So the fact that the job
ClusterFuzzLite/CIFuzz
starts off reporting its status in a "neutral" state seems like a bug. Is there a way that it can start off with some type of "in progress" status of some kind instead? And only change the status to something else once the job conclusively ends successfully one way or another?The text was updated successfully, but these errors were encountered: