Bug 213382 – [ews] commit-queue should comment on bug if it fails to get contributors information

• Status: RESOLVED
• Resolution: WORKSFORME
• Component: Tools / Tests

[Aakash Jain]

commit-queue should comment on bug and set cq- flag if it fails to get contributors information (and fails to commit because of that). This is so that people can easily notice that commit-queue failed, and take necessary action. 

However, the chances of commit-queue hitting this issue (of failing to get contributors information) is almost zero after the improvement in Bug 213377.

[Aakash Jain]

Created attachment 402291 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 402291 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=402291&action=review

> Tools/BuildSlaveSupport/ews-build/steps.py:758
> +            self.setProperty('bugzilla_comment_text', 'commit-queue failed to get contributors information, please cq+ patch again.')

It is very useful to have specific guidance about what the person needs to do. This one doesn't sound right to me - if the action were to simply mark as cq+ again, we wouldn't need to mark as cq-. Also, this information is noise to everyone except for the person who has set cq+ originally.

So I think that two things need to change:

1. Make the guidance accurate (the whole reason to do this is to have people reach out to admins, so we should advise on how to do this).

2. Make it private (e-mail, and don't change cq+ to cq-).

Looking at these together, I'm beginning to think that the right action is to notify admins directly, not to ask contributors to do it.

[Jonathan Bedard]

(In reply to Aakash Jain from comment #0)
> commit-queue should comment on bug and set cq- flag if it fails to get
> contributors information (and fails to commit because of that). This is so
> that people can easily notice that commit-queue failed, and take necessary
> action. 
> 
> However, the chances of commit-queue hitting this issue (of failing to get
> contributors information) is almost zero after the improvement in Bug 213377.

I feel like if we manage to hit this issue with the fix from bug 213377, infrastructure is so far broken that there is nothing anyone except for an admin could do to help.

[Aakash Jain]

As noted above, the chances of hitting is extremely low (since local file reading should pretty much work). And if we hit this scenario, infrastructure is too much broken already. It's probably not worth to spend more time in trying to improve automatic handling of this rare scenario.

Developing mechanism for notifying admins would be done in Bug 207475.