We only need the config attribute list from the caller,
we can chose the configs and free them ourselves
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Now that we're first trying no-config contexts,
there is no need to inject an `EGL_NO_CONFIG_KHR`
as the first config in the returned list
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Only the render acceleation and xv work for some reason.
glx only wants to create indirect contexts, which don't work.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
It will likely be slower that unaccelerated rendering.
Command-line flags were added for overriding this.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
This makes it so that glamor glx can properly initialize and
choose hw accelerated `FBConfig`s, instead of just using softpipe/llvmpipe.
This commit fixes the issue described in https://github.com/X11Libre/xserver/pull/1832#issue-3827524680
Now, `FBConfig`s corresponding to the choosen accelerated `EGLDevice`
are choosen by `glXChooseFBConfig`.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Now gbm required for the modesetting driver.
`GLAMOR_HAS_GBM` is defined even if glamor isn't built.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Since libgbm was split from mesa, and a backend that works
with only dumb buffers was written, there is no reason why we
should roll out own dumb buffer implementation.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
If we do this in the generic present `check_flip`, we would have to
do it before we call the driver's `check_flip`.
This means that we wouldn't get a reason for the flip failure,
which we still need to report if we are in the middle of a flip.
Thanks to @nkalkhof For helping bisect this in
https://github.com/X11Libre/xserver/issues/1812
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
If glamor isn't used, gl isn't initialized.
This can lead to the wrong cursor mask interleave value to be used.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Adding a #define in xlibre-server.h that drivers can #ifdef on,
in order to check for the new EDID parsing API.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
The old ones didn't know the block size, so couldn't deduce the block
type version. With upcoming new features, eg. HDR, we need to know the
block type version in order to know what we can extract from it.
This new function should now be used by all drivers, the old ones shall
be phased out.
That commit should be backported to 25.0 and 25.1 releases, so drivers
can remain compatible with all existing release lines.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
This is (rarely used) special mode where, at server regeneration
(when last client exited), Xnest is additionally terminating and
restarting it's upstream connection, thus also recreating it's
main window, etc.
That mode is only meant for special debug scenarios, not suited for
practical use. And the same can be achieved by just terminating the
whole Xnest process (-terminate cmdline arg) and restrting it.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Not used by any external module/driver, so no need to keep it
public. Also move it to the new dix settings code and rename it.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
