[xserver PULL master] randr fixes + primary GPU fallback

[Hans de Goede]

Hi,

On 06-12-16 19:06, Adam Jackson wrote:
> On Tue, 2016-12-06 at 15:22 +0100, Hans de Goede wrote:
>> Hi,
>>
>> On Mon Dec 5 16:54:51 UTC 2016, Adam Jackson wrote:
>>  > Not a fan of the autobind patch in terms of upstreaming, I feel like we
>>  > really should be able to do better.
>>
>> We've had years to do better
>
> And? This is seat configuration. It's something the system should tell
> us, not something we should guess at. That solves it both for all DEs
> and for all display servers.

No this is not seat configuration, when auto-configuring output devices
we already take the seat into account AFAIK, so we only bring up GPUs which
are part of the seat the xserver is running for, this is about actually
tying these GPUs together, so that they actually form a coherent seat,
rather then being claimed by the server (since they belong to the current
seat) but not being usable by the end-user unless the end-user first
types some black-magic xrandr commands.

I find the this is seat configuration pov quite interesting btw, that
does nicely solve the policy problem of which devices to bring up in
a DE agnostic manner, which means that the one thing which has stopped
this patch from going upstream actually has been solved long ago.

Regards,

Hans