On 09/11/15 05:19, Dave Airlie wrote:
On 21 August 2015 at 22:08, Marek Olšák <mar...@gmail.com> wrote:
Hi Alex,
The tests which don't specifically test the behavior of the default
framebuffer should use -fbo. I think this is done by marking a test as
"concurrent" in all.py. A lot of tests don't do that, because nobody
had the time to update all.py and check if it doesn't break them and
if it's okay to do that with regard to the purpose of the tests. (i.e.
do they specifically test the default framebuffer or something else?)
Just FYI classic swrast suffers from this issue, and since I added front buffer
support to softpipe/llvmpipe they now also.
We try to call XGetImage when no window has been mapped, and the X server
correctly throws BadMatch. I'm not sure how to detect this is going to
fail in this
case, as I don't think inside libGL I can detect the unmapped front buffer.
Dave.
Dave, since you see a problem on swrast too, are you willing to give
your R-b on this change?
I guess this will force us to spend time on moving tests to using fbo,
to speed things up and be less annoying, but we should not rely on
unspecified behaviour anyway...
Martin
_______________________________________________
Piglit mailing list
Piglit@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/piglit
