On Wed, 3 Oct 2018 at 09:19, Mathias Fröhlich <mathias.froehl...@gmx.net> wrote: > > Hi Emil, > > On Tuesday, 2 October 2018 11:59:23 CEST Emil Velikov wrote: > > > > + return dev->device->nodes[DRM_NODE_PRIMARY]; > > > ... we probably want > > > return _eglGetDRMDeviceRenderNode(dev); > > > > > That isn't quite possible, as discussed in 2016's thread > > "EGL_EXT_*_drm - primary vs render node". > > > > The extensions is (was?) not too clear that a card node must be > > returned, yet there are applications depend on it. > > Hmm, I feared that I get back something like that when I wrote the comment. > Then let it be it. > > What happens here is that your piglit test that takes the file returned here, > opens it and wants to create a new display from it fails on my test system. > ... obviously... > > > As mentioned in said thread we could add another extension which adds > > support for EGL_DRM_RENDER_DEVICE_FILE_EXT. > > But I'd suggest keeping that as a follow-up - hence the comment above > > _eglGetDRMDeviceRenderNode() > ... but yes. Probably the better approach then. > It's more like making all tests pass. May be we should not fail the test of > opening the file from there is failing because of permissions. > I agreed having the test fail is bad, yet the permissions handling of /dev/dri/card* is fundamental issue. One not specific to the extension or piglit tests.
The only thing that comes to mind is to SKIP the (sub)test if open() fails? I'll send a patch for that in a bit. -Emil _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
