Closed dragos closed 11 years ago
I think https://github.com/typesafehub/ghpullrequest-validator/blob/master/src/main/scala/backend/PullRequestCommenter.scala#L49 would be a better place, so we can still update status, but generate less noise. The reason we weren't doing this kind of detection any more is that the actor could in principle crash, and when we restart the bot more than 10 minutes after the job, it won't see the results.
I see... Unfortunately I'm not sure how to fix this anymore. PLS SYNCH
would probably be broken, if I understand the code correctly. I think I explained how this affects us in the call today, but here's a summary for the record:
This leads to a 'failed' comment being added immediately. As Adriaan rightly pointed out, "we're using it the wrong way". Unfortunately we'll have to live with this.
I never meant to imply you need to live with this. I've been thinking about how to improve this, and I'm sure it can be done.
... hours) before the kitty kicked-off another build.
We suppress the comment for a build that occurred in the past. We detect that by comparing the build timestamp with the timestamp when the
JenkinstJobStartWatcher
actor was started, and add a healthy 10 minutes epsilon.