Emil Velikov <emil.l.veli...@gmail.com> writes: > On 28/09/13 01:41, Eric Anholt wrote: >> Emil Velikov <emil.l.veli...@gmail.com> writes: >> >>> * clone the drienv to driswenv and adjust approapriately >>> * export driswenv and use it in dri-swrast >>> * ensure __NOT_HAVE_DRM_H is defined for drisw, similar >>> to all other common_drisw users >> >> I'm confused where __NOT_HAVE_DRM_H comes from. I don't see any >> references to it in the tree until your patch. >> > Yes that one did my head in a bit, here is what I've gathered > > If you decide to omit it, build will fail due to missing drm.h in the > include path, coming from dri_interface.h. > > Take a look at dri_interface.h, it has a very interesting heuristics - > #if def __APPLE__ || __CYGWIN__ || __GNU__ > #ifndef __NOT_HAVE_DRM_H > #define __NOT_HAVE_DRM_H > #endif > > Thus the obvious question, why did it work before and not after - I'm > assuming that scons plays "nicely" with the __GNU__ define. > > The last one is only speculation as I've ran out of patience at that > moment :\ Although the following info wrt __NOT_HAVE_DRM_H is quite > interesting. > > Three out of four automake swrast providers define it > * src/mesa/drivers/dri/swrast/Makefile.am > * src/gallium/targets/dri-swrast/Makefile.am > * src/gallium/state_trackers/dri/sw/Makefile.am (swrast libGL.so) > > and only one Scons target provides is > * src/gallium/state_trackers/dri/sw/SConscript > env.Append(CPPDEFINES = [('__NOT_HAVE_DRM_H', '1')]) > > Note #ifdef __NOT_HAVE_DRM_H vs #if __NOT_HAVE_DRM_H
Not sure what I did before to fail at grep, but yeah, it's obviously in the tree. I've squashed your patch into mine -- sound good?
pgpBUK7YCrfmQ.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev