[Intel-gfx] Policy for dealing with CI failures?

[Intel-gfx] Policy for dealing with CI failures?

[Hans de Goede]

Hi All,

I'm wondering if there are any (unwritten) guidelines for how to deal with CI
failures. Although it is great to have such an extensive CI systems, it seems
to result in quite many false-positives.

After a couple of resubmits I've gotten 2 patch-sets to pass the BAT tests,
but both have failed the IGT tests on what I'm reasonably sure are unrelated
failures.

Some examples:

This single patch series has a Reviewed-by and I would like to push it to
dinq, but it keeps failing CI:

https://patchwork.freedesktop.org/series/69902

This is a new series, which still needs reviews, etc. but it too is
currently failing CI for what are likely reasons which are unrelated to
the series:

https://patchwork.freedesktop.org/series/70945/

So my main, generic question is how am I supposed to deal with these
cases. Resubmit with a [CI] prefix in the subject until things pass?
Take a good look at the CI results and then use common sense?

Have it at least pass BAT and use common sense for IGT?

Regards,

Hans


_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [Intel-gfx] Policy for dealing with CI failures?

[Hans de Goede]

p.s.

Reading one of the IGT failure summary emails I see:

"If you think the reported changes have nothing to do with the changes
  introduced in Patchwork_15773_full, please notify your bug team to allow them
  to document this new failure mode, which will reduce false positives in CI."

Is there say an email address I can forward CI failure reports which
I believe to be false-positives to for this? What is the best (non IRC /
non real-time) method to contact "your bug team" ?




On 15-12-2019 22:09, Hans de Goede wrote:
> Hi All,
> 
> I'm wondering if there are any (unwritten) guidelines for how to deal with CI
> failures. Although it is great to have such an extensive CI systems, it seems
> to result in quite many false-positives.
> 
> After a couple of resubmits I've gotten 2 patch-sets to pass the BAT tests,
> but both have failed the IGT tests on what I'm reasonably sure are unrelated
> failures.
> 
> Some examples:
> 
> This single patch series has a Reviewed-by and I would like to push it to
> dinq, but it keeps failing CI:
> 
> https://patchwork.freedesktop.org/series/69902
> 
> This is a new series, which still needs reviews, etc. but it too is
> currently failing CI for what are likely reasons which are unrelated to
> the series:
> 
> https://patchwork.freedesktop.org/series/70945/
> 
> So my main, generic question is how am I supposed to deal with these
> cases. Resubmit with a [CI] prefix in the subject until things pass?
> Take a good look at the CI results and then use common sense?
> 
> Have it at least pass BAT and use common sense for IGT?
> 
> Regards,
> 
> Hans
> 
> 

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx