Nicolai Hähnle wrote: > Recently, some optimisations have been introduced to texenvprogram.c so that > if we determine, based on vertex state, that certain fragment attributes must > be constant, they are read from the constant file instead of from the > fragment > input registers. > > However, the assumptions that lead to this optimisation fail when the > fragments are created by a call to DrawPixels. This can be observed, for > example, by the failure of the glean/pixelFormats test with the r300 driver, > in which output is produced alternately by rendering quads with constant > color > and with DrawPixels. > > The following patch fixes this problem by setting the vp_override flag around > the actual operation of _swrast_DrawPixels. It fixes the problem for me, but > it may not be the most elegant solution to the problem, so if you think there > is a better fix, please tell me. Otherwise I will commit this patch to master > in a few days.
We basically do the same thing in the Gallium state tracker's st_DrawPixels() code so this is probably fine. I've been doing a fair amount of glean/conform testing lately so if there's a regression I'll probably find it soon enough. -Brian ------------------------------------------------------------------------------ Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H _______________________________________________ Mesa3d-dev mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
