update-webkit-chromium sometimes fails when run from the bots because gclient can be flaky, as demonstrated in https://bugs.webkit.org/show_bug.cgi?id=67857#c6. We should allow the script to retry the gclient call in case a given failure is transient.
Created attachment 107674 [details] Patch
Comment on attachment 107674 [details] Patch Do we want to back-off at all? There are also cases where gclient fails because you've screwed up editing the GYP file, and those shouldn't be retried... This is probably a good start though.
I'm not sure how to determine which failures should or shouldn't be retried. Maybe based on the specific errorcode? I'll think about ways to identify which ones would make sense to retry or not. I think the worst current case would be that if you break a gyp file, it will retry (with no chance of success) and cost a little extra time.
Comment on attachment 107674 [details] Patch Clearing flags on attachment: 107674 Committed r95363: <http://trac.webkit.org/changeset/95363>
All reviewed patches have been landed. Closing bug.