[Mesa-dev] [PATCH] dri_util: remove ALLOW_RGB10_CONFIGS option (v2)

[Tapani Pälli]

Hi;

On 11.01.2018 20:20, Mario Kleiner wrote:
> On 01/11/2018 09:14 AM, Tapani Pälli wrote:
>> Yes, but as it broke regular visuals (on some of our testing machines 
>> as well) we needed a fast fix for this.
>>
>> While this is an issue, I think the visual corruption has higher 
>> priority than this. This can be fixed meanwhile or afterwards when 
>> things work OK. Now it can be toggled true in intel_screen.c when 
>> debugging/investigating the issue.
>>
> 
> That regression in Bug 104536, could you only reproduce it on "Ubuntu 
> 16.04 or 17.04 respectively, with latest git versions of libdrm, mesa, 
> xserver, xlibs, and the drm-tip kernel." as David Weinehall reported in 
> private message?

So far this has been only seen in CI machines. (I've been busy with 
something else ATM but I will also try to get to this one soon.)

> Does it require X-Server from git master?

Yes, this is how those machines do, it's the latest components of the 
graphics stack.

> Because testing here on Intel Ivybridge and Skylake with KUbuntu 16.04.3 
> LTS and the distribution-provided X-Server 1.19.3 and the Unity-7 + 
> Compiz standard Ubuntu 16.04 GUI, i can't reproduce the visual 
> corruption from that screenshot in bug 104536?, neither with the 
> intel-ddx nor the modesetting-ddx used by default, neither depth 24 nor 
> depth 30.

What I've understood (from some bugs related to problems with sRGB 
visuals) is that with X server master there are a lot more visuals in 
general and as example can be multiple 32bit visuals (which did not 
happen before). This could possibly explain why these issues cannot be 
reproduced on LTS provided Xorg.

// Tapani