[PATCH 11/12] dri2: Don't block GLX clients on VT switch

[Adam Jackson]

On 10/17/12 12:11 AM, Dave Airlie wrote:
> On Wed, Oct 17, 2012 at 11:18 AM, Eric Anholt <eric at anholt.net> wrote:
>> Adam Jackson <ajax at redhat.com> writes:
>>
>>> This was necessary in DRI1 for a variety of reasons (SAREA ownership,
>>> VRAM layout, etc).  None of that is relevant to DRI2.  Which is good,
>>> because VT switch is an xf86-ism and I'd like to make DRI2
>>> ddx-independent.
>>>
>>> Signed-off-by: Adam Jackson <ajax at redhat.com>
>>
>> We still do this?  Not cool.
>>
>> The only risk I see is if it means that new clients race ahead to trying
>> to do dri2 auth and fail because we've dropped master.
>
> yeah don't VT switch during a piglit run, it won't end well. The thing
> is I suspect some app have grown used to this behaviour and will act
> all crappy on VT switch without it.

We already lose window images on VT switch, which (still) routinely 
surprises people.  Which is sort of vestigial from UMS and XFree86 3 
days where pulling the whole framebuffer down to host memory was both 
too slow and too much host memory to be practical.

It's not especially hard to fix that these days.  It only requires ze 
vill to do so.  I'll take a look.

- ajax