Closed kbrenneman closed 3 years ago
Seems OK to me. @oddhack - how does this look?
Seems OK to me. @oddhack - how does this look?
I think it would be good to get signoff from both Mesa and NVIDIA on this change - I am not sure which, if either, @kbrenneman represents here. The actual change seems fine to me but I don't use all those combinations of platforms.
@kbrenneman , can we get an review/ compile check/ reassurance from x11 community on this change?
@kbrenneman is with NVIDIA. @nwnk for feedback from X.org side and @ianromanick because he objected when we tried to redefine EGLint one time, and this is vaguely reminiscent of that but much less invasive IMHO.
Thanks James. I was just hoping for someone from x.org to poke for input. Just to be careful.
I'm very pleased to see this, with my Mesa and Gentoo/Linux hats on.
This changes eglplatform.h so that it'll use the
EGL_NO_X11
typedefs by default.This has come up a couple times in libglvnd bug reports, where applications that use EGL but not X11 don't build on systems where the Xlib headers aren't installed. The
EGL_NO_X11
macro can fix that, but adding that to every broken app out there is a tall order, and adding it to libglvnd's pkg-config files seems like a crude workaround at best.This change should be pretty safe, I think. Applications that currently pass a
Display *
value to eglGetDisplay will still build, because you can always pass aDisplay *
to avoid *
.EGLNativePixmapType
andEGLNativeWindowType
won't change at all, because in both the Xlib and non-Xlib cases, they're typedefs forunsigned long
.Existing calls to eglQueryNativeDisplayNV could break, but I'd expect those to be relatively uncommon, and would just need a trivial typecast to fix. Alternately, you can still define
USE_X11
to get the current behavior.