On 04/07/2016 12:40 PM, Stéphane Marchesin wrote:
On Thu, Apr 7, 2016 at 3:37 AM, Miklós Máté <mtm...@gmail.com> wrote:
On 04/07/2016 12:25 PM, Stéphane Marchesin wrote:
On Apr 7, 2016 02:27, "Michel Dänzer" <mic...@daenzer.net> wrote:
On 07.04.2016 18:01, Marek Olšák wrote:
On Thu, Apr 7, 2016 at 7:32 AM, Ian Romanick <i...@freedesktop.org>
wrote:
Why would you do that? I've NAKed this patch several times.
You NAKed the previous version of the patch, not this one. I guess
that means NAK for this one too.
I'd be interested in a specific justification for NAKing this version.
Seems to me like this might be the best that can be done.
We asked ourselves the same question.
It is the best that can be done to fix glx 1.2 for sure. But it will break
glx 1.3, so for example it will break chrome (I have a vested interest here
:).
So what we figured out is, if we have to choose between the old and new api,
why break the newer (and arguably better) one? Instead we should consider
the old one deprecated/legacy and move on.
Stéphane
How does this break glX 1.3, and how does this break Chrome?
I don't have the source here, so it's off the top of my head. But
isn't this going to start leaking dri drawables if you constantly
makecurrent?
No, it doesn't leak anything in makecurrent.
By itself this is ok (because they all point to the same
buffer) but there are pieces of glx 1.2 (like swapbuffers) that
iterate over all the dri drawables and then they become really slow.
Stéphane
The problem with this patch, as I was informed, is that people don't
call glXDestroyPixmap (or glXDestroyGLXPixmap) before calling
XFreePixmap, and thus the resources allocated by glx are leaked (until
the process exits). In current Mesa, the dri drawable associated with
the glx object only lives while the glx object is current to a context,
which prevents the above problem in a crude way.
MM
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
