[04:01] <BenC> linux-restricted-modules-2.6.19-4-generic_2.6.19.1-1_i386.deb
[04:01] <BenC> sweetness
[04:01] <zul> nice
[04:02] <infinity> Does it work? :)
[04:14] <BenC> infinity: I upgraded to latest unified ati shar file
[04:14] <BenC> I don't have any ATI to test with though
[04:14] <BenC> rest of the stuff just got the normal request_irq() fixups, and removal of linux/config.h
[04:15] <BenC> getting ready to test nvidia
[04:19] <BenC> infinity: I plan on having lrm/linux-meta/linux-source-2.6.19 uploaded tomorrow
[04:19] <BenC> infinity: want to test lrm+ati?
[04:22] <BenC> infinity: if you feel like testing it, I'm uploading the x86 packages including fglrx stuff to rookery:~bcollins/ now
[04:24] <infinity> lrm+ati won't work, guaranteed.
[04:24] <infinity> Unless you did the necessary (and hush-hush, we don't talk about it) hex-editing.
[04:32] <BenC> nope, give me some hints :)
[04:33] <infinity> Probably easier shown than spoken.  I can mangle the orig for you for this upload, and show you at UDS-MV.
[04:33] <BenC> can we do it in debian/rules?
[04:33] <infinity> Note thta we have an unwritten (but oft confirmed) pseudo-license with ATI to blatantly violate their license (which disallows binary patching), hence why this is a bit hush-hush.
[04:34] <BenC> I'd rather we keep the shar file intact without modification
[04:34] <infinity> I've tried a few ways to automate it, and it never quite turns out right.
[04:34] <infinity> We always just rolled our own tarballs in previous releases.  *shrug*
[04:34] <BenC> is this on the kernel module or the libaries?
[04:34] <infinity> (We did even before I started patching binaries)
[04:34] <infinity> This is libGL.
[04:34] <BenC> is it rpath related?
[04:35] <infinity> Vaguely, yeah.  They hardcode the (incorrect) path to the DRI modules.
[04:36] <infinity> It could be fixed with a symlink on i386 (though ugly), but it can't really be fixed on amd64 at all without binary patching, since they build for SuSE which uses lib/lib64 while we use lib32/lib
[04:37] <infinity> And, of course, as any kid who used to do binary patching in the bad old days knows, you must make sure to pad out the replaces string with nulls, and make sure the binary is the same length at the end. :)
[04:37] <infinity> Trivial to do with a hex editor by hand, I had a hard time coaxing sed or perl into wanting to give me love.
[04:39] <BenC> infinity: That may no longer be the case
[04:39] <infinity> I'd be ecstatic if that were so.
[04:39] <BenC> infinity: I can't find any hint of that path in any of the binaries
[04:39] <infinity> Let me grab the latest.
[04:39] <BenC> 8.29.6 is what I have
[04:40] <BenC> single shar for x86 and x86_64
[04:40] <infinity> We want 8.30.3
[04:40] <infinity> 8.29.6 is known-broken.
[04:40] <infinity> Like, moreso than usual. :)
[04:40] <BenC> 8.30.3 is beta isn't it?
[04:41] <infinity> Not according to their press release that says it was released yesterday. :/
[04:41] <BenC> wasn't on the download when I went to "current", just a few hours ago
[04:41] <infinity> Gah, ati.com goes to ati.amd.com now.
[04:41] <BenC> yep :)
[04:41] <infinity> Man, I need to keep up with tech news.
[04:43] <infinity> Curernt is 8.30.3 when I go there. :)
[04:44] <infinity> Thouhg, where are you finding this unified shar?
[04:44] <infinity> Also, I'm womdering if we should do an fglrx-legacy, or if we should just tell poeple to use the radeon driver if they've got old cards.
[04:45] <BenC> Linux Drivers -> FireGL Series -> Linux x86 Display Driver... -> any "Current Driver"
[04:45] <infinity> Hrm, I went a different route.
[04:46] <BenC> I went to "Drivers & Software"
[04:46] <infinity> But, kay.  While they server the 32/64-bit "driver installers" out of different directories, they do appear to be the same file. :)
[04:46] <infinity> serve, even.
[04:47] <BenC> yeah, md5sum proved that to me :)
[04:47] <BenC> got a url got 8.30.3?
[04:47] <infinity> I'm low on bandwidth, I just trusted file lenght. :)
[04:47] <infinity> https://support.ati.com/ics/support/KBAnswer.asp?questionID=1176
[04:48] <BenC> I'll drop that in and rebuild
[04:49] <infinity> Want to unshar it, "find . -name libGL\* -exec strings \{\} \; | grep usr" for me?
[04:49] <infinity> If they've really removed those hardcoded paths, I'm a happy man.
[04:49] <BenC> if not, I'll see about hacking them in-build
[04:49] <infinity> I'm sure it's not too much effort, I just gave up cause I had better things to do.
[04:51] <infinity> If I did the regex matching (and replacing) in pure escape codes, and found the magical invocation for "stop trying t null-terminate the string, you dimwit", I'm sure it'd go smoothly. :)
[04:52] <infinity> Of course, the other advantage to distributing it pre-patched was that it wasn't obvious from the build system that we were violating the license.
[04:52] <infinity> And ATI doesn't really want us to mention that we're doing so (which is why it's not in debian/copyright), cause then everyone will want to.
[04:53] <infinity> "Well, if Ubuntu can add a binary patch for their pet issue, why can't we add binary patches for PCI IDs?" etc.
[04:54] <infinity> I realise the irony of me discussing this in a public channel, but it's not like it's a SECERET that we provide a patched libGL (any idiot can verify that with md5sum), it's just something we prefer to pretend we're not doing.
[05:02] <BenC> why do we move the dri modules if they are hard coded?
[05:04] <infinity> In the i386 cause, it's because we don't ship /usr/X11R6 anymore, in the amd64 case, it's because of that, and also because of the lib/lib32/lib64 confusion.
[05:04] <infinity> Our base system (lib) is 64-bit, SuSE's base system is 32-bit, so the biarch logic is ass-backwards there.
[05:05] <infinity> While we can symlink lib64->lib, we clearly can't symlink lib->lib32. :)
[05:14] <BenC> right
[05:20] <infinity> Oh, this could prove.. Interesting...
[05:20] <infinity> http://projects.info-pull.com/mokb/
[07:59] <fabbione> BenC: are you still around?
[06:07] <HeMing> anybody in here?
[09:40] <tonfa> gug
[09:41] <tonfa> just curious, has anybody managed to boot -rc4-mm2 (with or without SYSFS_DEPRECATED)
[09:41] <tonfa> ?
[09:41] <tonfa> and thanks for the UUID stuff, switching to libata works fine!