Philipp Klaus Krause <p...@spth.de> writes: > Am 28.10.2012 20:51, schrieb Kenneth Graunke: > >> This makes sense to me. The original reason for splitting out the ANGLE >> vs. EXT bits was that the ANGLE extension only requires decompression, >> while the other two require both compression and decompression. So that >> one could be advertised even without libtxc_dxtn. >> >> Of course, now that we've decided to simply lie and fall back to >> GL_COMPRESSED_RGB[A] when the user asks for online compression and we >> don't have libtxc_dxtn, this isn't a problem. So we can simplify. > > I can see how one wants to lie about EXT_texture_compression_s3tc and > give users a broken GL by default, since that's what they want. AFAIK > previously it was a driconf option, but enabling it was too much of a > burden on the user. > Still there might be some users that want standard-compliant behaviour > without having libtxc_dxtn installed. Why not give them a driconf option > they can use to disable EXT_texture_compression_s3tc when libtxc_dxtn is > not there?
We used to have a more-strict-correctness flag like that in the intel driver for other cases of strict correctness. It turned out that users actually didn't want it, and it was a trap for QA which would say "ooh, more correctness, we should be testing that code instead, right?" and thus not test the code that shipped. So I deleted it.
pgp32vgu065hK.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
