On 07/09/2012 10:59 AM, Eric Anholt wrote: > Chad Versace <[email protected]> writes: > >> In GLES, glReadPixels does not accept format=GL_FLOAT, but it does accept >> format=GL_BYTE. >> >> This shouldn't make any tests begin failing. If anything, this commit may >> make pass some almost-passing tests because there is less precision in >> probing with glReadPixels(GL_BYTE) than glReadPixels(GL_FLOAT). > > Choosing to compare with bytes should be based on the type/size of the > renderbuffer's channels, imo.
But in GLES, GL_BYTE is the only valid format (other than luminance formats) that can be given to glReadPixels. Do you suggest that we have two implementations for this function, one for GLES and one for GL, that only differ by GL_BYTE vs GL_FLOAT? That seems to be the only sensible alternative to this patch, and I don't think it's very sensible. _______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
