Currently, the testing infrastructure uses a commit hook trigger to queue up a branch for testing on qa.d.o.

However, as I understand it, when a maintainer creates a series of commits, and pushes them all at once, the commit hook fires once per commit (instead of once for the push event) ... thus queuing mutiple redundant instances of the branch test those commits belong to.

For the testing infrastructure, we really only need to test each 'push' as opposed to each commit, as the test pulls the current state of the branch for each test (not, for example, a specific commit ID).

Does this capability currently exist? And if not, can we make it a feature request?

Comments

marvil07’s picture

marvil07 commented

Actually, that was one of the main refactor tasks when event object was introduced.

The used hook, hook_versioncontrol_code_arrival() acts after the repository have made the synchronization, which in the case of git, is when the the queued sync is run. That naturally assumes the git post-receive hook was added/configured correctly to add queue items per push.

So, yep, it is once per push, can you point me to the place where you saw that behavior? It sounds really strange.

jthorson’s picture

jthorson commented
Status: Active » Fixed

It was an anecdotal report that the bot was queueing up multiples on every push, from a core committer ... and with collaborating reports. I simply made an assumption on what could be causing this.

Managed to clear things up (clarified that this was on push, not commit) with Sam, but hadn't got a chance to get back online and close the issue, due to travel.

Thanks for the response!

Status: Fixed » Closed (fixed)

Automatically closed -- issue fixed for 2 weeks with no activity.