[08:52] Hi. Again i am living this unendable ATI nightmare. Ubuntu switched X-Sserver three times in for distributions and all the time it was like starting from scratch with an ATI RS690 (x1250). Now testing 10.10 here we go again. With xserver 1.9 I have again a flickering screen with HDMI, a dissorted Image in Googleearth, a glxgear framerate limited to 300frames in 5 seconds, and not even gallium to test. I can test that with the x-edgers dr [08:52] Can we hope for gallium in 10.10? [08:54] minimec: In 10.10? No. [08:54] It's not a panacea, either. At this point it's more differently broken :) [08:55] RAOF: definitly not? So I don't see a reason to continue with 10.10... Let's hope for a good 2.6.35 Kernel in lucid to get the new ATI features... [08:56] Incidentally, glxgears going at 300frames / 5 seconds is totally intentional. [08:56] This is why we say glxgears Is Not A Benchmark™. [08:57] You'll notice that 300 frames / 5 seconds corresponds to… 60 frames/sec, which is the 60Hz of your screen's refresh rate. [08:57] Behold! For vsync now works! [08:58] RAOF: I know. In fact... I have a regression in 'Look+Feel' of the whole system. That GPU runs definitly slower on 10.10... [08:59] Also, the X server itself doesn't have a particularly large influence on the open drivers - you say that HDMI worked for you at some point? [09:00] It's probably kernel changes which have broken that. [09:00] RAOF: HDMI worked first time in Karmic and worked in Lucid.. Now I get the sreen, but the image is adjusting all 10 sec or so. [09:01] have you tried with the lucid kernel? [09:01] RAOF: At leat I have to say, that video playback has evolved on the ATi opensource driver and there is (almost) nothing to complain anymore. [09:02] tjaalton: Talking to me? I see no reason to use an 'obsolete' kernel (missing some ATI features) on a testing system... [09:02] tjaalton: ... using a ATI device ;) [09:03] minimec: Using a lucid kernel (and getting HDMI working) would mean that your HDMI woes are kernel related. [09:03] It's a… test. For your testing system :) [09:04] RAOF: THat is in fact some an argument. I installed the x-edgers ppa to have my HDMI working. [09:04] For bonus points you could test to see _when_ HDMI broke; you could go through all the released kernels from Lucid to current Maverick and find the last one which worked. [09:04] It works in x-edgers? Do you have kernel modesetting disabled? [09:04] RAOF: So HDMI is working with the current 10.10 kernel with the x-edgers ppa [09:05] RAOF: No. KMS is default and activated. [09:05] Colour me surprised; userspace doesn't do very much on the modesetting front now. [09:06] yeah, I would've thought it was in the kernel.. [09:07] RAOF: the RS690 is some special GPU based basically on the r300. That's why I guess we have 'good' x-edgers support for this card (kms included) [09:08] xorg-edgers has the same x server as Maverick, so we can pretty much discount that... [09:14] Hm. And xserver-xorg-video-ati doesn't seem to have any interesting commits re: HDMI for your card. Unless it's really an r600 :) [09:17] RAOF: http://www.x.org/wiki/RadeonFeature It is definitly not. [09:30] RAOF: You are right... I also have the HDMI problems with the x-edgers ppa. I switched too much from my 'productive'- to my 'test'- system. Ok this will end in a bug report. [09:31] try with the lucid kernel to see if that helps.. [09:32] tjaalton: Yep. THis will be part of bug tracking (or an 34 kernel from the kernel mainline) [10:55] using xorg-edgers ppa(linux-image-2.6.35-19-generic) on Lucid. if I install libgl1-mesa-dri-gallium using nvidia 9800gtx+ keyboard and mouse does not work at gdm. though I can still kill X by alt+sysrq+k any help? [14:42] bryceh, jcristau: I've pushed up lucid and maverick branches for evdev to git://git.debian.org/users/corp186-guest/xserver-xorg-input-evdev.git [14:42] http://git.debian.org/?p=users/corp186-guest/xserver-xorg-input-evdev.git;a=summary [14:42] in case you want a gitweb version [15:08] Sarvatt, RAOF: is it possible to get fglrx on maverick yet, or are we still waiting for updated drivers? [15:10] cnd: there's no driver yet [15:10] I'm the maintainer BTW [15:12] tseliot, ok, thanks! [15:13] tseliot, do you know when it will be available? [15:14] cnd: maybe... but I can't discuss it here [15:14] tseliot, ok [15:14] it's not information I *need* [15:14] just want it for my laptop :) [15:14] so don't worry about it [15:14] cnd: ok, good to know ;) [15:24] cnd: oh I totally missed your messages here, sorry! [15:25] Sarvatt, np :) [15:26] Sarvatt: BTW my late congrats on joining Canonical :-) [15:26] cnd: looks good (evdev git branches) [15:27] cnd: from what I have read that is public information, fglrx 10.10 will be the first one with new features and 10.8-10.9 will just be bug fix releases, ubuntu may get a beta of 10.10 if it supports xserver 1.9 I imagine somehow :) [15:27] cnd: btw mesa in xorg-edgers supports 3D on hd 5xxx chips now [15:27] thanks tseliot! [15:29] cnd: but not 2D acceleration, they are doing that on a seperate branch of the driver because its got freezing problems at the moment [15:29] -radeon without acceleration is probably faster than fglrx at 2D though :) [15:30] heh [15:31] Sarvatt, can you briefly walk me through the steps to tell if I have an hd 5xxx chip? [15:31] just lspci? [15:31] do you know if its a HD 5000 or higher series? [15:31] I have an RV710/730 [15:32] I don't know anything other than what I can find in ubuntu [15:32] ah that uses r600 [15:32] Sarvatt, so is there any hardware 3d acceleration I can get in maverick? [15:32] yes [15:32] you have it already [15:32] oh, then it's just plain borked :) [15:32] it's significantly better in mesa 7.9 though [15:32] unity is all messed up [15:33] Sarvatt, do I just need to use the xorg-edgers ppa to get the new mesa? [15:33] r600 works great here but you need the latest mesa to use unity [15:33] oh yeah unity is messed up on all ati's it seems like at the moment :( [15:33] i thought it was just r500 and older ones though === JanC_ is now known as JanC [15:33] unity is broken on intel 945 and 965 and nouveau here too [15:34] ah [15:34] cnd: just curious - glxinfo | grep GL_ARB_texture_non_power_of_two [15:35] I think that should be in the new mesa [15:36] Sarvatt, I do have that [15:36] i thought it was in mesa 7.8 for r600+ [15:36] without any new mesa [15:36] ah ok not the same problem [15:37] that wasn't the only problem [15:38] cnd: try CLUTTER_VBLANK=none unity [15:38] that actually fixes it for me on intel [15:38] Sarvatt, ok [15:39] it was all white before [15:39] Sarvatt, it's all screwed up now [15:40] Sarvatt, should I just try out the latest mesa? [15:40] thought it was before? :) [15:40] I just don't know where it is [15:40] Sarvatt, more screwed up :) [15:43] cnd: can you push the maverick branch as ubuntu instead for consistency? [15:44] Sarvatt, should I keep lucid as is and just call rename maverick to ubuntu? [15:45] sure if ya want, i doubt we'll use the lucid one since SRU's are usually by people not on pkg-xorg :) xserver uses ubuntu-lucid though, might make sense to use that too [15:47] ok [15:53] is there no way to configure unity? its taking up a bit too much space on this netbook [15:54] it's odd, looks like it could have just been a docky theme [16:02] what's the command to check for nvidia alternatives? [16:13] update-alternatives --display gl_conf [16:18] this guy is having trouble playing interlaced european video using vdpau on his ion netbook [16:18] but the hardware should be good enough [17:11] ugh, a mesa 7.9 merge in debian git is rough, trying to work out how to do this without manually fixing hundreds of conflicts :) [17:18] it's always like that, you can't avoid it ;) [17:24] Sarvatt: remove/modify conflicts, hopefully? [17:26] http://paste.ubuntu.com/483534/ [17:26] almost done fixing it up, its not as bad as I thought :D [17:29] never done mesa before and thought i might have been missing a merge strategy that makes it easier or something [17:29] merge -s ours $old_upstream_branch into the new one helps as a first step [17:41] there should be something on the wiki about that [17:41] iirc [17:41] it's somewhere hidden in http://wiki.debian.org/XStrikeForce/git-usage [17:43] and also on https://wiki.ubuntu.com/X/GitUsage [17:50] afternoon Sarvatt [17:54] Sarvatt: I'm struggling trying to build a RC2 kernel to try a suggestion of jbarnes. Where would I find directions? [17:59] yeah XStrikeForce/git-usage is insanely helpful and on my bookmark bar :) jg: hey there! I haven't built a kernel myself in about 6 months and i'm really not sure with all of the changes in the kernel-package package, maybe this wiki will help? https://help.ubuntu.com/community/Kernel/Compile [18:01] jg: yeah the section starting with "The following instructions are based on this link:" near the end is what you should use [18:04] the way i do it doesn't build an initrd or anything and I do it all manually, be sure to disable comedi and ti-st from staging when you do a make menuconfig because they don't build currently [18:08] I noticed :-(. I want to be fully upstream, so I can easily test with jbarnes. [18:18] sudo apt-get install fakeroot kernel-wedge build-essential makedumpfile libncurses5-dev git-core kernel-package; sudo apt-get build-dep linux; cp -r /usr/share/kernel-package $HOME; git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git; cd linux-2.6; cp /boot/config-$(uname -r) .config; make oldconfig; make menuconfig (disable comedi and ti-st from staging drivers) then make-kpkg clean; CONCURRENCY_LEVEL=`getconf _NPROC [18:18] ESSORS_ONLN` fakeroot make-kpkg --initrd --append-to-version=-jg --overlay-dir=~/kernel-package kernel_image kernel_headers [18:18] hopefully that should work :) [18:20] * Sarvatt writes that down for next time he's asked [18:21] probably could just use /usr/share/kernel-package as the --overlay-dir=, not sure why they did it that way in those steps [18:22] i would add a git checkout v2.6.36-rc2 in there after cloning, i think make-kpkg had a problem with non tagged versions recently [18:25] also be prepared to delete /etc/kernel/postinst.d/nvidia-common or remove nvidia-common, it doesn't like kernels made with make-kpkg :) [18:31] dang, not having any luck working out this merge without making a mess of the history. [18:35] creating a new branch off of upstream/master, merge -s ours upstream-experimental into it, then merging ubuntu on top of that for debian/ and merging that into ubuntu sound right? that gets me the least amount of conflicts but 2 merges in the history [18:35] yeah that's basically it [18:38] seems like this will make a mess when 7.9 is in experimental, and oops make that 3 merges in the log since i haven't commited this one yet :) [18:38] wait, why are you merging ubuntu twice? [18:39] it deletes debian/ when i make a branch off upstream/master, merge -s ours upstream-experimental then merge that new branch into ubuntu [18:41] uh. [18:43] i probably did one of the steps wrong with a pull instead of a merge or reverse, will try again [18:46] bryceh, jcristau, Sarvatt: sorry for the delay, but I finally did get ubuntu evdev changes pushed to git.debian.org [18:46] ubuntu became ubuntu-lucid [18:46] and the maverick branch name became ubuntu [18:46] cnd, excellent [18:56] Sarvatt, the catalyst 10.8 driver is out. [19:27] dandel: thanks for the heads up, will package it up when i get some free time [19:31] oh yeah, lunch break. I guess that time is now :) [19:32] the release notes are for 10.7... [19:33] hah [19:34] 6% performance drop is no longer observed while running fgl_glxgears and glxgears on specific cards === yofel_ is now known as yofel [19:47] Sarvatt, the #1 feature change is: OpenGL ES 2.0 support [19:47] I dunno, 6% performance increase in glxgears has my vote [19:48] OpenGL ES 2.0 support is good for firefox v4.x [19:48] so that's what has my vote. [19:48] GL ES support will be interesting to the ubuntu mobile guys I'd bet [19:49] yes, and it's also interesting for console devs. [19:51] increasingly the Consoles and mobile devices are gaining OpenGL ES Support [19:52] yeah I was just being sarcastic :) chromium has had welGL/gles support for almost a year now too [19:53] WebGL rather [19:53] its just too bad this doesn't work on maverick [19:57] well they added support for xserver 1.8 silently in 10.6, maybe it will [20:04] ok its uploaded to x-updates now [20:27] any idea mesa has or will implement framebuffer objects in swrast? [20:33] howdy [20:34] I got a SSD for my laptop yesterday, and did a fresh install of maverick with BTRFS [20:34] the 1st few boots I had 13 secs with nouveau. once I installed nvidia blob, it went to 23 sec. [20:34] so I'm willing to give the newer nouveau with 3D support a try. [20:34] what should I know? how stable and fast is it? what ppa has it? (x-edgers?) [20:43] asac: it has for a long time afaik? I see GL_ARB_framebuffer_object/GL_EXT_framebuffer_object under swrast [20:43] BUGabundo: you can just install libgl1-mesa-dri-experimental in maverick [20:44] Sarvatt: thanks. no ppa?? how do I do it with jockey ? [20:44] dont need a PPA, just deactivate nvidia-current in jockey and enable effects [20:45] hit cancel when it pops up saying the only way to have desktop effects is by installing the blob :) [20:45] I'm lost [20:45] (after installing libgl1-mesa-dri-experimental) [20:45] so I isntall libgl1-mesa-dri-experimental [20:45] then deactivate nvidia-current, then reboot and you have 3D [20:45] disblabe blob in jockey [20:45] okay [20:45] eheh [20:46] 10 sec boots here I came [20:46] when you go to enable desktop effects it'll say you need to install the blob but you can just hit cancel [20:46] LOL [20:46] hope that gonna be fixed :D [20:46] how fast and well support is it? [20:46] dri-experimental isn't supported :) [20:46] multi monitor, and stuff [20:46] ahhhh [20:47] it runs fine though [20:47] depending what GPU you have [20:47] geforce 7xxx and older is probably bad :) [20:48] t8300 (2.4gh) with nvidia 8400m G [20:48] its fine on that, i've had my wifes laptop on the same thing pretty much with nouveau for about 8 months now [20:49] turion x2 tl-60 with a 8400m gs [20:49] as always there's always xorg-edgers if you have problems that might be fixed :) [20:50] remove dri-experimental first though i've got nouveau in the main dri package there [20:50] ill try to remember [20:50] since its not supported, are you interested in any kind of reports [20:51] ? [20:52] yeah if its broken it'll be nice to know when not to update her laptop :D [20:52] AHAHAHAH [20:52] nouveau is in ia32-libs in edgers too for wine [20:53] ? [20:53] wine needs 32 bit DRI libs, its not in ia32-libs in maverick [20:53] ahhh [20:53] not installiung wine for now [20:54] pretty sure google earth needs 32 bit too if you use that [20:54] yeah, that's just wine [20:59] $ sudo aptitude install libgl1-mesa-dri-experimental [20:59] let the mayhem begin [21:03] oop [21:03] forgot to re-enable 3d [21:04] oi, its already enabled :S [21:04] rebooting [21:08] hoy this nouveau 3D is FAST [21:13] ok [21:13] that didn't make my boot any faster :((( [21:14] you could reduce the grub timeout from 10 seconds down to 1 or 2 [21:14] and its X that is taking 5 precious seconfs [21:14] bjsnider: there's no timeout at all [21:14] it boots directly [21:16] well, i don't approve of that [21:16] BUGabundo: try after doing a echo FRAMEBUFFER=y | sudo tee /etc/initramfs-tools/conf.d/splash && sudo update-initramfs -u [21:16] bjsnider: stock maverick install [21:17] sudo delete /etc/initramfs-tools/conf,d/splash and sudo update-initramfs -u after to revert that chane [21:17] sudo delete? [21:17] sorry doing 10 things at once :) [21:18] let me upload bootcharts [21:18] so you guys can look at them [21:18] did you reboot twice? maybe ureadahead was working that time [21:18] only once [21:18] only hve 6 boots since install [21:26] oh wow, git rerere is my new best friend [21:27] Sarvatt: http://bootcharts.f.bugabundo.net/ [21:28] BUGabundo: thats unreadably small [21:28] click to expand [21:28] right side, full size png [21:29] what kind of SSD did you get? I'm guessing an intel x25-v [21:29] oh nevermind thats a nice max speed :) [21:30] 22:27 < asac> LIBGL_ALWAYS_SOFTWARE=1 glxgears [21:30] 22:27 < asac> 1180 frames in 5.0 seconds [21:30] 22:27 < asac> that cant be right ;) [21:30] is that not supported anymore? [21:30] BUGabundo: btrfs? RAOF was mentioning how btrfs had a *huge* performance regression in maverick right now [21:30] sure it can asac. its magic [21:31] Sarvatt: I read about it [21:31] but seems to be workign fine for me [21:31] asac: whats wrong about it? :) [21:31] I told you that, when I entered the room [21:31] Sarvatt: i dont see a fbs difference for glxgears with and without it [21:31] on full screen [21:31] is that ok? [21:32] yeah glxgears isn't a benchmark of anything useful, those numbers dont look wrong to me [21:33] good. i confirmed in glxinfo that its really software rast [21:33] thanks [21:33] * asac goes test unity with that ;) [21:34] 922 frames in 5.0 seconds = 184.349 FPS on a turion x2 [21:36] WOW [21:36] jumping to TTYs and back is WAY faster [21:38] 208 frames in 5.0 seconds [21:38] asac left ? :( [21:39] Sarvatt: what did asac do to get that!? [21:39] probably has a fast cpu :) [21:39] :(( [21:41] and no compiz running [21:41] ahh [21:41] FAIL [21:42] well going for a new reboot [21:42] to see if that command you gave did anything [21:48] Sarvatt: I'm still losing: near the end of building the packages, it says: dpkg-gencontrol: error: package linux-image-2.6.36-rc2-jg+ not in control info [21:49] jg: yeah thats the error you get if you dont build from a tag and use master, I'm not sure how to fix it outside of just building from the tagged release [21:49] git checkout v2.6.36-rc2 [21:50] I did.... [21:50] oh? hmm [21:50] mew [21:50] 20 sec [21:50] a bit lower [21:50] still not the 13 sec of the 1st boots [21:50] uploading now [21:52] maybe it's because you had local changes, ugh [21:52] local changes? [21:53] I have a new kernel ... LOL [21:53] and yeah, I instaled a bunch of packages [21:53] that was directed at jg [21:53] let me remove some stuff from the start [21:53] ahhhhhh [21:53] not in my log [21:53] LOL [21:53] jg: I'm really sorry but I don't know how to fix it and am in the middle of some work, maybe someone in #ubuntu-kernel would know? [21:54] Sarvatt: thanks for your help. [21:54] will look in a bit otherwise [22:07] jg: http://ftp.us.debian.org/debian/pool/main/k/kernel-package/kernel-package_12.036_all.deb [22:07] that should fix it [22:10] need to rm -rf ~/kernel-package and cp -r /usr/share/kernel-package $HOME again after updating it, looks like the version in maverick is screwed up with .35-rc4 and newer kernels. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=588126 [22:10] Debian bug 588126 in kernel-package "kernel-package: Fails to build linux-image on 2.6.35-rc4 if CONFIG_LOCALVERSION is set" [Important,Fixed] [22:23] jg: if you're on i386 and can send me the patch I can try building you one on my vps in about 45 minutes [22:25] RAOF, ping [22:31] Sarvatt: that seemed to finally fix it.... I have a couple packages to try to install.... [22:41] Sarvatt: ok, if I want to rebuild a single driver, now what do I do? [22:42] hope you used ccache the first time? :D [22:43] really though you might be able to just run the make-kpkg line again without cleaning it first and have it only build that again, i'm not sure [22:59] jg: thanks for getting me to look into it because I've been wanting to figure out how to do it these days :) [22:59] hyperair: have you seen that linux-phc has dkms packages now?? [23:15] jg: btw there should be no need to use a tagged release in that new kernel-package afaik [23:35] did the latest kernel fix wacom bamboo? [23:35] input: mt: Add support for the Bamboo Touch trackpa [23:58] Daekdroom: Contextless pong. [23:59] RAOF, bug 617201 isn't fixed yet [23:59] Launchpad bug 617201 in xserver-xorg-video-ati (Ubuntu) (and 1 other project) "Xserver crash in radeon_frame_event_handler (affects: 4) (dups: 2) (heat: 30)" [High,Fix released] https://launchpad.net/bugs/617201