Hello, I just ran into this issue. I'm running piglits on a NV17 card, which uses the nouveau_vieux (dri) driver. It does not support the DRI_IMAGE stuff.
Taking a sample piglit test -- tests/spec/arb_clear_texture/cube.c -- it requests either GL 2.0 or GLES 2.0. The latter would attempt to use EGL. However since EGL is completely non-functional, it hits: _eglLog(_EGL_FATAL, "DRI2: did not find extension %s version %d", matches[j].name, matches[j].version); in egl_dri2.c, which causes the test to exit and fail instead of skip due to insufficient version. There are only a handful of uses of eglLog with EGL_FATAL: src/egl/drivers/dri2/egl_dri2.c: _eglLog(_EGL_FATAL, "DRI2: did not find extension %s version %d", src/egl/main/eglcurrent.c: _eglLog(_EGL_FATAL, "failed to initialize \"current\" system"); src/gallium/state_trackers/egl/common/egl_g3d_api.c: _eglLog(_EGL_FATAL, "destroy a context with an unitialized display"); src/gallium/state_trackers/egl/common/egl_g3d_api.c: _eglLog(_EGL_FATAL, "destroy a surface with an unitialized display"); Does it really make sense to terminate the process in these cases? Should I downgrade the egl_dri2.c one to an error (since it seems to handle the failure just fine), or should I just make fatal not exit (which seems a little nasty to do from a library)? -ilia _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
