https://bugs.kde.org/show_bug.cgi?id=358750
--- Comment #24 from Niels Ole Salscheider <niels_...@salscheider-online.de> --- (In reply to Thomas Lübking from comment #23) > The relevant code (initEGL()) is in platform agnostic in libkwineffects - > we'd either have to transfer it into the platform implementations to make us > of eglGetPlatformDisplayEXT (or pass it down alongside the EGL_EXT_platform* > tests incl. an explicit query for client extensions on EGL_NO_DISPLAY) Yes, that's why I haven't proposed a patch yet... > => Simply setting EGL_PLATFORM avoids the crash? The crash is already avoided by the patch I sent to mesa-dev. But simply setting EGL_PLATFORM makes mesa choose another platform that uses render nodes instead of /dev/dri/card0 so that I do not see "radeon: Failed to get PCI ID, error number -13" anymore. Also, it would avoid the crash even when my patch was not applied to mesa. EGL_PLATFORM was a workaround by mesa for this problem when eglGetPlatformDisplayEXT was not yet around. It still works but I think that the latter is the cleaner solution. > I'm however not sure whether that's a general solution to the problem (see > eglinfo situation) I'm afraid that there is not really a general solution, because unfortunately eglGetDisplay made it into the standard in its current form without considering that there might be drivers with multiple platforms. Applications should probably use glGetPlatformDisplayEXT when possible. And for all that don't, distributions should set a sane default backend that will hopefully not fail (surfaceless?). -- You are receiving this mail because: You are watching all bug changes.
