[python-committers] AppVeyor off?

[Brett Cannon]

On Mon, 17 Jul 2017 at 05:52 Nick Coghlan <ncoghlan at gmail.com> wrote:

> On 17 July 2017 at 21:29, Antoine Pitrou <antoine at python.org> wrote:
> >
> > Nevermind, it seems to have come back online.
>
> I've occasionally seen that not just with Appveyor, but also with
> other pre-merge CI systems: it seems like their event monitoring queue
> and/or task scheduler can get backed up (sometimes significantly), so
> it takes them a while to report back that they've seen the PR and have
> a test run scheduled.
>
> Travis is the most reliable I've seen when it comes to getting their
> "Result pending" status back in promptly, so my assumption is that
> they actually send that back before (or in parallel with) scheduling
> the CI run itself.
>

There's also race condition issues between GitHub's webhook events and when
the API actually exposes the data that triggered the webhook (this is why
both the Knights Who Say Ni and Bedevere bots have a forced delay before
updating a PR). I don't know if AppVeyor heavily delays trying again if the
initial status setting fails, but it's possible.

I should also mention I try to announce anything that is a visible change
to our workflow here, so if I haven't said anything was turned off, assume
it broke. ;)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-committers/attachments/20170717/1eaffd83/attachment.html>