The purpose is to include the work of Roland and enable Texture Compression for XOrg X11 users of ATI Radeon (r250) and Intel Graphics (i830, i915) hardware. There are some concerns about intellectual property, which are handled by sticking the icky parts in an external library. See http://homepage.hispeed.ch/rscheidegger/dri_experimental/s3tc_index.html for exact details.
Created attachment 40575 [details, diff] v1, 6.8.0-dri-s3tc-040908.patch files/6.8.0-dri-s3tc-040908.patch hand patched from the tarball available on Roland's site.
Created attachment 40576 [details, diff] v1, patch against x11-base/xorg-x11/xorg-x11-6.8.0-r1.ebuild patch against x11-base/xorg-x11/xorg-x11-6.8.0-r1.ebuild
Icky parts go in bug #65607
finished compiling /installing xorg-x11 build with these patches. doesn't hurt anything, xorg-x11 functions just like it did before. i have no idea how to test if this texture compression thing works or not, but i ran glxgears and got debug output as follows. erisha@necro ~ $ glxgears Mesa: software DXTn compression/decompression available 8409 frames in 5.0 seconds = 1681.800 FPS 10189 frames in 5.0 seconds = 2037.800 FPS 10187 frames in 5.0 seconds = 2037.400 FPS 10192 frames in 5.0 seconds = 2038.400 FPS
i just tried out a copy of nwn i haven't played in a long time, because of graphic glitches and lack of s3tc. all i can say is wow, it works wonderfully now. tested working on x86 (athlon-xp k7), everything on this box was 'emerge -e world'ed with gcc-3.4.2-r2
As mentioned, there are concerns about IP. IMHO, they aren't sufficiently addressed by the separate library solution, but I will defer to upstream if they say the problems are nonexistent. I'm not willing to add this patch until it's added upstream. Is it?
the patch + library solution has been proposed - and rejected - upstream long ago. S3TC appears to be a live and defended patent, and no one has reached an agreement with Via about implementing it in Mesa. Brian Paul is not willing to add patent-encumbered code to Mesa. the situation is similar to the Apple freetype patent, but they take the more liberal stance of ifdefing the encumbered code out. until an IP lawyer negotiates an agreement between Via and Mesa, upstream will never include S3TC support. patent-encumbered code violates clause 7 of the DFSG: "The rights attached to the program must apply to all to whom the program is redistributed without the need for execution of an additional license by those parties." but i don't recall how closely Gentoo hews to the DFSG, so that may not be relevant.
http://www.gentoo.org/main/en/contract.xml "Any external contributions to Gentoo Linux (in the form of freely-distributable sources or binaries) may be incorporated into Gentoo Linux provided that we are legally entitled to do so. However, Gentoo Linux will never depend upon a piece of software unless it conforms to the GNU General Public License, GNU Lesser General Public License or some other license approved by the Open Source Initiative (OSI.)" Willfully and knowingly contributing to patent infringement is not legal. The "depend" clause applies to "system" software only, so that's not particularly relevant to this bug, only to ajax's implied question. I'm closing this as WONTFIX pending an upstream solution.
