On 1/22/2020 18:24, Patchwork wrote:
> == Series Details ==
>
> Series: drm/i915/guc: Update to GuC FW v40 (rev3)
> URL   : https://patchwork.freedesktop.org/series/72032/
> State : failure
>
> == Summary ==
>
> CI Bug Log - changes from CI_DRM_7786_full -> Patchwork_16198_full
> ====================================================
>
> Summary
> -------
>
>    **FAILURE**
>
>    Serious unknown changes coming with Patchwork_16198_full absolutely need to be
>    verified manually.
>    
>    If you think the reported changes have nothing to do with the changes
>    introduced in Patchwork_16198_full, please notify your bug team to allow them
>    to document this new failure mode, which will reduce false positives in CI.
>
>    
>
> Possible new issues
> -------------------
>
>    Here are the unknown changes that may have been introduced in Patchwork_16198_full:
>
> ### IGT changes ###
>
> #### Possible regressions ####
>
>    * igt@kms_atomic_transition@5x-modeset-transitions-fencing:
>      - shard-tglb:         NOTRUN -> [SKIP][1]
>     [1]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_16198/shard-tglb8/igt@kms_atomic_transition@5x-modeset-transitions-fencing.html
According to the log, this test failed because it ran on a device that 
only had one display device attached and not the five required by the test:

IGT-Version: 1.24-g5cf58d947 (x86_64) (Linux: 5.5.0-rc7-CI-Patchwork_16198+ x86_64)
Starting subtest: 5x-modeset-transitions-fencing
Test requirement not met in function run_modeset_transition, file ../tests/kms_atomic_transition.c:887:
Test requirement: num_outputs >= requested_outputs
Should have at least 5 outputs, found 1
Subtest 5x-modeset-transitions-fencing: SKIP (0.000s)


I'm not sure how that could be called a regression in the GuC FW patch. 
I also don't see any reason why the test would previously have been a 
'NOTRUN' and now is being attempted. Changing the GuC FW should not 
affect which KMS tests do or do not get run!

I don't have a system with five display devices so I can't actually run 
the test myself either. However, I do not see how this could be affected 
by changes to the GuC. Especially when the GuC is only being used for 
HuC authentication.

So I think this definitely counts as an issue with CI not this patch.

John.