[igt-dev] ✗ Fi.CI.BAT: failure for tests/kms_chamelium: add hdmi-audio test

[Arkadiusz Hiler]

On Fri, Apr 26, 2019 at 05:13:21PM +0000, Patchwork wrote:
> == Series Details ==
> 
> Series: tests/kms_chamelium: add hdmi-audio test
> URL   : https://patchwork.freedesktop.org/series/60003/
> State : failure
> 
> == Summary ==
> 
> CI Bug Log - changes from CI_DRM_6006 -> IGTPW_2929
> ====================================================
> 
> Summary
> -------
> 
>   **FAILURE**
> 
>   Serious unknown changes coming with IGTPW_2929 absolutely need to be
>   verified manually.
>   
>   If you think the reported changes have nothing to do with the changes
>   introduced in IGTPW_2929, please notify your bug team to allow them
>   to document this new failure mode, which will reduce false positives in CI.
> 
>   External URL: https://patchwork.freedesktop.org/api/1.0/series/60003/revisions/1/mbox/
> 
> Possible new issues
> -------------------
> 
>   Here are the unknown changes that may have been introduced in IGTPW_2929:
> 
> ### IGT changes ###
> 
> #### Possible regressions ####
> 
>   * igt at kms_chamelium@hdmi-edid-read:
>     - fi-kbl-7567u:       [PASS][1] -> [FAIL][2]
>    [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6006/fi-kbl-7567u/igt@kms_chamelium@hdmi-edid-read.html
>    [2]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_2929/fi-kbl-7567u/igt@kms_chamelium@hdmi-edid-read.html

This is a real regression:

tests/kms_chamelium.c:1763
----------------------------------------------
connector_subtest("hdmi-edid-read", HDMIA) {
	test_edid_read(&data, port, data.edid_id,
		       igt_kms_get_base_edid());
	test_edid_read(&data, port, data.alt_edid_id,
		       igt_kms_get_alt_edid());
}
----------------------------------------------

data.edid_id is this new edid for HDMI, yet the blob we provide for
comparison is the "base igt edid".

Having to pass both representations is rather error-prone.

I think we should just use IDs and when we are registering new EDID,
we should store it in a map id -> EDID. Then such tests would take
edid_id only, and retrieve the blob from the map.

-- 
Cheers,
Arek