/srv/irclogs.ubuntu.com/2010/02/23/#ubuntu-x.txt

rickspencer3bryceh, I'm installing Sauerbrten right now00:06
brycehhmm, with current lucid bits, I'm seeing a [drm:drm_mode_getfb] *ERROR* invalid framebuffer id and some garbage. 00:06
brycehrickspencer3, ok00:06
rickspencer3*the* definitive graphics stress test00:06
brycehhah00:06
brycehit's a good one00:06
BUGabundoeheh00:08
RAOFHow are we with the work-items for nouveau?  In particular, what can we do about nouveau-firmware?00:08
BUGabundoit really pushes my laptop to its limits00:09
* bryceh uninstalls plymouth and eliminates the fb id issue00:09
BUGabundobryceh: :D00:09
BUGabundohalf of lucid test users, will end up without it :D00:09
brycehRAOF, I think the kernel team intends to obsolete nouveau-firmware with the generator thingee upstream is doing00:09
brycehI don't know if it's at the top of their todo list at the moment though, but it might be good to remind them if they don't get to it in the next few weeks00:10
RAOFOk, but that's not going to land before A3 :).  I've been tracking it in #nouveau, though, and it looks like mwk's making progress.00:11
brycehRAOF, in the meantime I agree we should add it as a dependency, although I'm not sure exactly where00:11
brycehRAOF, I need to read your email more closely00:12
brycehRAOF, does it need to be a kernel-level dependency?  If so then I suggest we bring it to apw and sconklin and let them decide00:12
RAOFnouveau-firmware?  It pretty much is a kernel-level dependency, in that the kernel module shipped in lbm-nouveau wants the firmware.00:13
superm1why not just put it in lbm-nouveau for now then and pull it out when it's time to go?00:14
superm1would keep things far simpler without needing to obsolete the package00:14
RAOFOne reason might be it's uncertain licensing, but it *would* be simpler it it were just shipped in lbm-nouveau.00:15
superm1what's this generator thing that upstream is working on?  it generates it on the fly for the HW?00:17
RAOFYes.00:18
RAOFThey're presume that's how the blob does it, too.00:18
superm1what's it generating it from then?  the VBIOS?00:18
RAOFFrom primitive operations & a description of the actual card hardware, I think.00:19
RAOFThere's already a voodoo generator for nv4x, which is why my laptop doesn't need the nouveau-firmware package.00:19
superm1so then where did the firmware that exists today actually come from?00:19
RAOFMMIO traces of the binary blob.  It's a dump of how the binary driver initialises the card.00:20
superm1yucky00:20
RAOFHence “uncertain licensing”00:20
superm1so more importantly, what's the HW do w/o that firmware in place then?00:20
RAOFIt's not quite firmware.  It is really a small GPU program that's run on context switches.00:21
RAOFSo, without the voodoo you don't get context switches, which means you don't get any acceleration.00:21
superm1so not the end of the world, just fairly inconvenient00:23
RAOFYeah.00:23
RAOFIt interacts badly with the fact that nv5x+ chips don't expose a linear framebuffer, though, which means that CPU fallback is quite a lot slower.00:23
RAOFSo nouveau needs to read VRAM, unswizzle the bits, do the CPU fallback, reswizzle, then push back to VRAM.00:24
RAOFCurrent status of the voodoo generator for nvAx cards appears to be “[In quake 3] depending on place in the level where you're standing, and the direction you're looking, you can get anything from perfect display, through textures replaced with pink, up to totally bogus geometry”00:28
rickspencer3bryceh, so, this is totally subjective, but saurbraten seems smoother and more stable in Lucid00:57
rickspencer3I suppose that's mostly -intel00:57
brycehnice00:59
brycehthere has been some performance work done with -intel00:59
rickspencer3bryceh, yeah, I'll really test it out this weekend when I have more than 10 mins ;)01:00
RAOFtseliot's done a bang-up job on making the binary nvidia driver work nicely - Enabling & Disabling nvidia in Hardware Drivers just works.01:30
brycehsweet01:32
RAOFSarvatt: Your xorg-edgers mesa packages don't actually build nouveau's gallium driver because you've misspelt “--enable-gallium-nouveau” - you've got “--enable-nouveau-gallium”.01:47
* Sarvatt slaps forehead01:48
Sarvattuploaded a fixed one just now01:49
Sarvattsorry about that01:49
Sarvattwas just that last upload with the problem, ran into some problems with the rules and did it by hand01:51
RAOFYou can make it up to me by telling me about how you build those packages, so I can fix it myself next time :).  I understand that there's a script that you run to automate the package uploads.01:51
Sarvattdo you have auto-xorg-git?01:52
Sarvatt./auto-xorg-git -H hooks -g -d origin/ubuntu -p mesa -a 0ubuntu0sarvatt is what I use for mesa, I need to update the hooks though in bzr01:53
RAOFWhere's auto-xorg-git?01:53
Sarvatti'll update xorg-pkg-tools in bzr now01:54
Sarvatthttps://code.edge.launchpad.net/~xorg-edgers/xorg-server/xorg-pkg-tools01:54
snap-lHaving some problems with an Intel card and a new load of Karmic01:59
snap-lThe graphics for programs like Nexuiz and Stellarium are corrupted01:59
snap-lNexuiz won't display other players/ floors, etc. intermittently01:59
snap-lStellarium displays the toolbars with some fuzzed out look (like some memory corruption).02:00
snap-lAny thoughts on how I can get 3D working properly on this machine?02:01
snap-lIt's an Asus K60I with an Intel® Graphics Media Accelerator 4500M02:03
snap-l32 bit load of Ubuntu 9.10. Seems to be using the i915 driver02:04
RAOFsnap-l: Hm.  I've got the same chipset, but on Lucid, and I haven't noticed any corruption.02:13
RAOFThat said, I'm not the most demanding of 3D users.02:13
snap-lYeah, Compiz works just fine, but this is a bit strange02:13
snap-lit seems like anything that requires several layers of textures gets messed up02:13
snap-lNot sure if it's memory allocation or something else.02:14
SarvattRAOF: for the nouveau ddx I use ./auto-xorg-git -H hooks -t + -g -d origin/ubuntu -a 0ubuntu0sarvatt nouveau    mesa: ./auto-xorg-git -H hooks -g -d origin/ubuntu -p mesa -a 0ubuntu0sarvatt02:15
RAOFThere have been any number of fixes to the Intel drivers in Lucid; you might want to try an alpha 3 livecd in a couple of days to see if it's fixed there.02:15
Sarvattanything with a + in the changelog is something i do by hand02:15
RAOFSarvatt: Ta.  I should probably make it so that the xorg-edgers nouveau DDX contains the driver date; it's a useful piece of debugging information.02:16
RAOFAlso, nouveau gallium works again on nv4x, modulo crazy update slowness.02:16
Sarvatti have to disable that snapshot date patch because you try to patch configure that doesnt exist after a git checkout, it clones then patches then runs autoreconf after the patching02:18
RAOFOooh, and wobbly works, too!02:19
RAOFI think that you should actually be able to apply that patch now.  Why don't I check? :)02:22
Sarvattcomment out the line in hooks/xserver-xorg-video-nouveau.prepatch02:23
Sarvattactually just drop the -H hooks line from it completely02:24
Sarvattthat patch removal is the only hook02:24
Sarvatt ./auto-xorg-git -t + -g -d origin/ubuntu -a 0ubuntu0raof nouveau would work :)02:25
RAOFBah.  No, of course it doesn't just work.  In fact, it'd be wrong, wouldn't it?02:27
RAOFWhat actually needs to happen is to run get-orig-source and import that...02:27
Sarvattcould just run the script with the hook, let it build then reenable the patch in the series and debuild -S -sa again02:34
Sarvatti'll just do that next time i update it02:36
RAOFI was thinking of running get-orig-source in the prepatch hook, git-import-orig'ing it, and then letting it do it's funky thing.02:37
RAOFAlso, nouveau gallium is awesome until compiz stops updating the screen.02:37
bjsnideri wonder if the nvidia drivers are really working as well on lucid as you think RAOF 02:40
Sarvatthmm maybe try toggling the unredirect fullscreen windows option RAOF?02:41
bjsniderthere are loads of complaints all day in +102:41
RAOFbjsnider: And when I followed up with one of them, they'd manually installed the drivers from the nvidia website and broken everything.02:41
bjsniderhahahaa02:41
bjsnideryou're kidding02:41
bjsniderare you kidding?02:42
RAOFWell, no.  They had, at one point in the recent past manually installed the drivers from nvidia.com.02:42
bjsniderbut you'd have to remove jockey for that02:42
RAOFDo you?02:42
RAOFWhy?02:42
bjsnideryou'd have to remove ubuntu-desktop02:42
RAOFDo you?  Why?02:42
bjsniderbecause jockey blocks the nvidia installer02:43
Sarvattjockey is optional :)02:43
bjsniderjockey isn't part of ubuntu-desktop?02:43
Sarvattnope, i dont have it installed and ubuntu-desktop is installed02:43
bjsnidersomehow i automatically associate everything on a default system with a metapackage02:44
bjsniderit's on a default system anyway02:44
RAOFJockey blocks the nvidia installer?  AWESOME.02:45
* Sarvatt would love a ncurses jockey he could actually use02:45
bjsniderwhat about the guy who said the 195 driver is crashing his rig every few minutes? what about all of the people who say they've lost opengl?02:46
Sarvattmost of the problems are from nvidia.com or installing PPA versions and screwing things up from what i see02:46
RAOFI guess that upgrades might be broken?  But my best guess would be Sarvatt's, yeah.02:46
bjsnidercan someone go over to that channel and explain that? these complaints are getting annoying02:47
bjsniderall issues with jockey not being able to activate the appropriate alternatives are solved right?02:48
Sarvattyep, i did see a bug earlier where someone was upgrading from hardy-lucid and didn't have a new enough dpkg to set up the alternatives though02:50
Sarvattbryceh: we're going to get those record extension patches in the next xserver update already02:58
Sarvattjust a heads up they'll be obsoleted with the next merge from debian, they'll be on 1.7 branch next point release (i think friday)03:00
Sarvatti've been watching that bug for a few weeks getting ready to pounce on it when they hit, just went on server-1.7-nominations 2 hours ago03:02
bjsniderSarvatt, you're going to pounce on a defenseless little bug?03:05
Sarvattbjsnider: are you using the blob right now? can you tell me the number returned from xlsclients | wc -l if so?03:12
bjsnidernot on lucid, no03:14
bjsnideron karmic it's 2203:14
Sarvattwas just wondering how the heck BUGabundo had >256 the other day03:17
brycehSarvatt, ok no prob, just wanted to ensure the patch didn't get dropped03:34
brycehoh there's another 1.7 point release in the works?  nice03:35
Sarvattwonder how much space we could save on the cd dropping some dri drivers that arent even usable like sis and mach64, they're about 2.3MB each03:45
bjsniderSarvatt, a guy in +1 just left a list of things the nvidia-installer left behind: http://paste.ubuntu.com/382004/03:46
bjsniderfascinating03:46
Sarvattso he didnt even run the installer with --uninstall then...?03:46
bjsnideri'm not sure03:46
bjsniderbut it leaves stuff behind even if you do03:47
Sarvatti think its --uninstall anyway, --help tells you the uninstall command03:47
bjsniderand now mysteriously his lucid system has no opengl03:47
Sarvattit deletes that stuff when i use the uninstaller03:47
bjsnidergee, i wonder why...03:47
Sarvattheck, i think its actually called nvidia-uninstall03:47
Sarvattinstalls a progam called that i mean03:47
bjsniderdo you trust it?03:48
Sarvattanother case of pebkac if he didnt run that :)03:48
Sarvattyeah i use it all the time03:48
bjsniderof course these people could say they uninstalled it whent hey really didn't03:48
Sarvatti've always used them straight from nvidia.com instead of using ppa ones, one of the reasons i dont like having an alternative installed in mesa :)03:49
bjsniderif you don't know enough to use the uninstaller between drivers, should you be using the nvidia-installer?03:49
Sarvatti think the nvidia.com installers should work again once libgl is moved back to /usr/lib though03:50
* bryceh flushes xserver-xorg-video-nv bugs04:55
* bryceh ker-puuushes04:55
RAOFWhoop!  Whoop!04:58
brycehallgone05:00
RAOFWhat is occasionally killing X when I press enter?05:05
RAOFIs that plymouth attacking me *again*?05:05
brycehyep05:12
RAOFArgh.05:13
RAOFNext question: what's causing evolution-data-server to consume a core!  It's a non-stop cavalcade of fun105:13
brycehnfi05:14
brycehmaybe ask on #ubuntu-desktop once seb128 is online05:14
RAOFI'll see if there's anything obvious in gdb first.05:15
brycehinteresting, when I try forcing fbdev to load via xorg.conf I get lots of lines like this:06:35
bryceh(EE) FBDEV(0): FBIOPUTCMAP: Invalid argument06:35
alkisgI installed Lucid a while ago, and the -pae kernel was automatically selected for me. On some updates the non-pae kernel was pulled in, so I now have both of them installed. The problem is that I can't uninstall the non-pae kernel, without removing xorg:07:29
alkisgsudo LANG=en_US.UTF-8 apt-get purge linux-image-2.6.32-14-generic07:29
alkisgThe following packages will be REMOVED:07:29
alkisg  linux-backports-modules-nouveau-2.6.32-14-generic* linux-backports-modules-nouveau-lucid-generic* linux-image-2.6.32-14-generic* xserver-xorg-video-all*  xserver-xorg-video-nouveau*07:29
alkisgIs that problem caused by the nouveau driver?07:29
RAOFbryceh (should have) fixed that with the most recent xserver-xorg-video-nouveau updload.07:30
alkisgAh thanks, let me try to update, I haven't updated in the last 2 days...07:31
brycehyeah alkisg let me know if it does not fix it07:50
alkisgbryceh: thanks, I think it's not published yet, so I'll wait until it does and give some feedback then.07:50
kklimondahmm, my laptop doesn't wake up from sleep anymore.. damn07:58
kklimondahow can I get some informations?07:58
alkisgkklimonda: I've read this a while ago, maybe it could help you: https://wiki.ubuntu.com/DebuggingKernelSuspendHibernateResume08:05
kklimondaalkisg: thanks - I'll probably wait till a3 is released and try it on the fresh system08:14
alkisgBTW, would anyone care to help me troubleshoot nouveau? It doesn't work for me, I have a laptop with 8600 GT M.08:32
brycehhmm, to -intel 2.10 or not -intel 2.1008:37
brycehtjaalton, any strong opinions?08:37
tjaaltonbryceh: no, haven't tried it myself08:38
tjaaltonbut maybe it's more tested against the .33 drm08:38
tjaaltondunno08:38
brycehok.  I don't know of pressing issues impelling us to 2.10, but since it drops UMS I'm hesistant to go with it.  Hard to make a firm decision though.08:39
tseliotbryceh: a colleague reported that -sharevts + rootless X causes X to crash with Ctrl-C (it works fine without that paramter)15:10
jcristaumeans the vt is not put in raw mode15:11
jcristau(also i assume by rootless you don't mean rootless, but euid != 0)15:11
tseliotrootless as in run without root privileges15:13
brycehtseliot, filed a bug?16:35
tseliotbryceh: nope16:35
jcristautseliot: see hw/xfree8/os-support/linux/lnx_init.c.  we only set RAW mode if (!ShareVTs).  my guess is, whoever owns the vt should do that.16:37
tseliotjcristau: ok, thanks16:48
tseliotbryceh: do you mind if I fix bug #467490?16:49
ubottuLaunchpad bug 467490 in nvidia-graphics-drivers-180 "nvidia drivers don't work due to -Q in obsolete /etc/modprobe.d/lrm-video" [High,Triaged] https://launchpad.net/bugs/46749016:49
brycehtseliot, sure feel free to reassign16:50
tseliotok16:50
SarvattRAOF: you're getting the sigquit when pressing enter after first boot problem?17:14
jcristauSarvatt: i've seen a sigquit report on bugzilla a few days ago... #2668917:15
tjaaltonfdo bug 2668917:18
ubottuLaunchpad bug 26689 in devmapper "Device mapper devices have incorrect ownership and permissions." [Medium,Fix released] https://launchpad.net/bugs/2668917:18
tjaaltoni'm lazy :)17:18
tjaaltonbah17:19
Sarvatthttp://bugs.freedesktop.org/show_bug.cgi?id=2668917:19
ubottuFreedesktop bug 26689 in DDX/xorg "Crash in WaitForSomething() (maybe?) on first execution on radeon-git/KMS" [Normal,New]17:19
tjaaltonfreedesktop bug 2668917:19
ubottuFreedesktop bug 26689 in DDX/xorg "Crash in WaitForSomething() (maybe?) on first execution on radeon-git/KMS" [Normal,New] http://bugzilla.freedesktop.org/show_bug.cgi?id=2668917:19
tjaaltonyeah17:19
SarvattRAOF: can you compile this and load it before hitting enter and reproducing it? http://sarvatt.com/downloads/trace-signal.c17:21
Sarvatt touch Makefile ; make obj-m=trace-signal.o -C /lib/modules/`uname -r`/build M=`pwd` modules    will compile it17:25
Sarvattwonder if gentoo is using plymouth too, that'd rule that out if not.. it didn't happen without plymouth here before and I cant reproduce it at all anymore even with plymouth :( the majority of people hitting the bug are using the blob though17:27
Sarvatthttps://bugs.freedesktop.org/show_bug.cgi?id=2267917:32
ubottuFreedesktop bug 22679 in Server/general "Xorg sporatically crashes during first session after boot running compiz" [Normal,New]17:32
SarvattOsSigHandler (signo=3 there too, thread on it here http://www.mail-archive.com/xorg@lists.freedesktop.org/msg07782.html17:35
jcristauoh, i'd forgotten about that threa17:37
jcristaud17:37
_stink_hey folks.  i'm trying to debug an X driver using the gdb info here: http://www.x.org/wiki/Development/Documentation/ServerDebugging18:59
_stink_the problem is that X never really starts using this driver.18:59
_stink_how can i still hook gdb in if X doesn't stay running, but crashes during the startup procedure?18:59
_stink_fwiw, it's kind of an obscure driver, and not one that's widely used or anything.19:00
_stink_so i'm not reporting a huge bug or anything. :D19:00
jcristauif you're lucky you can start X from gdb19:01
jcristauotherwise, get X to dump core, and run gdb on that19:01
_stink_jcristau: thanks!19:02
_stink_i should be using something other than startx to start it if i'm invoking from gdb, right?19:03
_stink_looks like xinit is what i want19:11
baptistemmhi there, do you need other information for https://bugs.edge.launchpad.net/ubuntu/+source/xorg/+bug/52666220:55
ubottuLaunchpad bug 526662 in xorg "[lucid] Screen randomly goes off (aka IDLETIMER issue)" [Undecided,New]20:55
jcristauyes.  need a reliable way to reproduce.20:58
baptistemmjcristau, that's the problem ...20:59
jcristaubecause afaik 1.7.5 has all the existing fixes in this area, so if you still have issues there then you need to tell us how to reproduce, or give us a patch20:59
baptistemmchrisccoulson, ^^ you said this afternoon some bits could be missing, and you had a patch21:00
chrisccoulsonbaptistemm, jcristau, yeah, i think i might know what happens there. baptistemm, do you see the screen dim first (if you're on battery)?21:05
chrisccoulsoni'll discuss it in a bit though21:05
chrisccoulsoni'm busy with other things atm21:05
baptistemmyeah, not a problem21:05
bjsniderRAOF, is docky using xrender or opengl?21:07
chrisccoulsonbaptistemm, jcristau, see gnome bug 609720 for the remaining race i found in gpm21:14
ubottuGnome bug 609720 in gnome-power-manager "Can sometimes miss idle reset alarm, causing display to blank when it shouldn't do" [Normal,Resolved: fixed] http://bugzilla.gnome.org/show_bug.cgi?id=60972021:14
chrisccoulsonwe committed a patch to fix it, but it actually caused more issues21:14
chrisccoulsonso it's been reverted again for now21:14
baptistemmchrisccoulson, thanks a lot21:14
chrisccoulsonbaptistemm, i'm not saying for certain that's your issue, but i can occasionally hit that issue21:14
baptistemmchrisccoulson, should I reopen the bug as the patch was reverted21:15
chrisccoulsonbaptistemm, please do. that's on my list of things to look at soon21:15
jcristauhrm, there's http://bugs.freedesktop.org/show_bug.cgi?id=26469 which might be related, i guess?21:29
ubottuFreedesktop bug 26469 in Lib/Xext "Multiple xsync timers fail to fire" [Normal,New]21:29
chrisccoulsonjcristau, that's possibly related21:41
chrisccoulsonalthough less exposed now, as gnome-screensaver doesn't directly rely on the idletime counter now21:41
chrisccoulsonalthough, thinking about it, gpm only registers a single alarm anyway21:42
RAOFbjsnider: docky is using cairo, which is using xrender.21:45
bjsniderRAOF, understood, sir21:45
Sarvattbryceh: looks like 2.9.1 with some way to selectively disable KMS by default on 8xx is the way to go. can you think of an easier way to implement that logic outside of patching the kernel? a kernel patch to do it wouldn't be that hard if not I'm pretty sure21:46
brycehSarvatt, "that logic" == blacklisting kms on 8xx?21:48
brycehyeah apw has talked about that but I don't know the implementation details of how that's done21:48
Sarvattyeah using modeset=0 just on specific pci id ranges21:48
brycehright, I think we need to mindmeld with apw and sconklin about that21:48
brycehbut yeah I'm also getting more and more certain about 2.9.1 rather than 2.1021:49
RAOFSarvatt: X died just as I was building trace-signal on my first boot; I've upgraded, rebooted, insmoded, and will drop you dmesg when X does it's swandive.22:18
* BUGabundo comments nouveua PPA22:19
BUGabundoare you serious?22:19
BUGabundoI was about to upgrade22:19
Sarvattif it didnt happen the first time you pressed enter it probably wont happen this boot RAOF, no need to leave it modprobed22:19
Sarvatteverything in xorg-edgers/nouveau is already in lucid BUGabundo 22:19
Sarvatti'm curious how fbdev over inteldrmfb works for 8xx people22:19
BUGabundoSarvatt: so where's the 3D?22:20
Sarvattxorg-edgers/ppa22:20
Sarvattif we did ship 2.10 and 8xx isnt fixed, can we just provide say a xserver-xorg-video-intel-legacy with 2.9.1 that has a modprobe.d conf with options i915 modeset=0?22:20
BUGabundoSarvatt: did that nvidia settings got updated?22:22
BUGabundoso I can test that comand you told me to improve speed of nvidia blob22:22
Sarvattno i have no idea if thats why, that option doesnt work if your card isnt new enough anyway so that might be why22:22
BUGabundo8400mG22:23
BUGabundo2 yo22:23
RAOFSarvatt: Aaaand there it goes!  Crash when I pressed enter to say “It doesn't always crash on the first press of enter” :)22:23
Sarvatti think only G96+ supports the stuff, GT200+ and some older things like 8200 integrated and ion22:23
BUGabundoSarvatt: I think mine is g85 or older22:24
SarvattRAOF: got the dmesg output when it sigquit?22:26
RAOFSarvatt: Have an (unfortunately trunkated) dmesg: http://paste.ubuntu.com/382574/22:26
Sarvattok its in there at line 1599 at least, looks the same as that fdo bug :(22:28
RAOFThe rest made it into syslog, so if you'd like (much, much!) more context, I can fish it out of there.22:28
brycehSarvatt, re: splitting out a -legacy intel driver yeah that's an option.  honestly though, looking through the changelog for 2.10 most of it is code deletions.  There's the Xv stuff which is new, and a few random things that might be bug fixes22:30
Sarvattbryceh: http://sarvatt.com/downloads/patches/103_increase_idgng_stride.patch http://sarvatt.com/downloads/patches/102_disable_bo_reuse_on_shadowfb.patch  2 backports from 2.10+ that I saw that would  be relevant on 2.9.122:31
brycehok22:32
Sarvatthmm src/libplybootsplash/ply-terminal.c is screwing with the tty22:47
apacheloggerhullos23:24
apacheloggercould someone help me debug a problem with xsetroot?23:24
Sarvatti'm guessing theres something racy, and ply_terminal_set_buffered_input (around line 196 of src/libplybootsplash/ply-terminal.c) is sometimes not getting the original term attributes and resetting the ISIG flag on X's VT so the enter key (0x1c) becomes the quit character (0x1c)23:24
apacheloggeror at least I think it is because of xsetroot :D23:24
Sarvatthmm http://suse-moblin.com/obs-status/Moblin:Factory/aaa_base/aaa_base-bootsplash.patch23:41

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!