/srv/irclogs.ubuntu.com/2010/06/09/#ubuntu-x.txt

RAOFAmong other things I think that steam does some GL vendor string parsing and freaks out about gallium drivers.00:00
RAOFThe nouveau drivers would be pretty featureful, but significantly less stable.00:01
RAOFAnd significantly less fast, too.00:01
lifelessbrb00:05
lifelessright, where was I00:08
lifelessoh yes00:08
RAOFIn Sacrifice?00:08
lifelessheh00:10
lifelessso in lucid00:11
RAOFCedega fullscreen being off by a couple of issues is likely to be either (a) cedega's fault or (b) a VGA connector issue. 00:11
lifelessand I just upgraded my games machine, so its new00:11
lifelesswhen it goes fullscreen, most of the time, its offset to the bottom-right00:12
lifeless*but the desktop shows in the gap*00:12
RAOFAh.  Silly cedega.00:13
lifelessworked in karmic00:13
Sarvattintels gpu's are such crap noone would buy a seperate card and they know it :)00:14
RAOFAnd it turned out doing a GPU in software on top of a bunch of tiny x86 cores wasn't as easy as they thought, too.00:16
RAOFWorked in Karmic, eh?00:16
lifelessRAOF: this occurs with both i915 or whatever my laptop has, and the nvidia in my games machine00:16
lifelessRAOF: and it was pixel perfect yesterday morning before the upgrade00:16
Sarvattwhat upgrade?00:17
lifelesskarmic-lucid00:17
lifelessmy games machine was shipped transtasmin so this was the first chance to do it00:17
lifelesshttp://imagebin.org/10051100:17
lifelessRAOF: ^00:17
RAOFI think this is unlikely to be an X problem.00:17
lifelessshows both the offset, and the other thing I wanted to query you on00:17
RAOFOh, a third thing? :)00:18
lifelessyes00:18
lifelesschoosepixelformat failed00:18
lifelesswhich google tells me was fixed in 2006 with glx1.300:19
lifelessand glxinfo claims the nvidia driver is glx1.400:19
lifelesswhats particularly special is alt-tab to and from the fullscreen app aligns it correctly on the screen00:19
RAOFI think google might be wrong, because the open drivers haven't been claiming glx1.3 support until… Lucid, until we disabled it.00:20
lifelesswell, I know that I know not, in this space.00:20
lifelessI'm going to get a vanilla wine install of steam up00:20
lifelessand see if its better00:21
Sarvattyeah that'd be the first thing i'd try too00:21
lifelessthough that itself appears to be black magic00:21
Sarvatt? it installs fine here00:22
Sarvatti'm in the process of installing counterstrike now in it00:22
Sarvattpretty sure you can just use the cedega one too since its just wine00:24
lifelessblows up when I run regular wine on it00:25
lifelesswith WINE_ENV or whatever set00:25
lifelesscleaner to start fresh, I think00:25
Sarvattlooks like wine with the blob starts things at 60hz when it changes resolutions00:28
Sarvattthe blob lies about what refresh rate you're at for its twinview crap00:28
Sarvatttry adding Option"DynamicTwinView""false" to your xorg.conf?00:29
Sarvattfound an old forums post from 07 with the same problem it looks like - http://ubuntuforums.org/showthread.php?t=62917300:30
lifelessI'm in awe of your googlejuice00:30
Sarvattwine fullscreen xserver 1.7 found it :)00:31
lifelessah 00:31
lifelessthats for the offset thing, not the pixel format00:31
lifelessright00:31
Sarvattlooks like disabling compiz will fix the choosepixelformat error00:33
lifelessI have visual effects set to None already00:34
lifelessI thought that disabled compiz ?00:35
Sarvattoh, darn :(00:35
Sarvatthttp://forums.steampowered.com/forums/showthread.php?t=128298100:36
Sarvattlooks like it picks directx mode by default and that switched it to opengl so it works00:37
lifelessthat screenshot was set to gl mode :)00:38
lifelessd3d mode goes boom00:38
lifelessI'm pulling down cs in a fresh wine setup00:39
lifelesswill report back in a couple hours I guess00:39
lifelessfor now, bzr stuff awaits00:40
Sarvattintel needs a rebuild :( looks like the server was stuck in NEW02:17
Sarvatt Provides: xserver-xorg-video-602:18
lifelessRAOF: so, lucid wine does a bit better02:43
lifelessSarvatt: the twinview thing helped02:43
rippsI'm trying to use ppa-purge to downgrade xorg-edgers to default maverick, but I'm having some kind of dependency problem with xserver-xorg-core and the virutal input packages.03:34
rippsSarvatt: ^03:36
RAOFripps: Yes, you will.  Not everything's built for the new server yet.03:36
rippsRAOF: ah, okay. Any idea when they'll roll out?03:37
RAOFToday.03:37
lifelessRAOF: cedega is still naffed :-) chalk one up for open sores03:42
RAOFHeh.03:43
lifelessRAOF: now, just have to remove pulseaudio and everything should be happy again04:08
RAOFOk.  Weirded out by that mesa backtrace.04:47
=== lifeless_ is now known as lifeless
ntr0pyI have a very strange bug indicating gdm starts too early in upstart on lucid resulting in low graphics mode with nvidia-current installed via jockey...09:12
ntr0pyThis is very irritating for people installing the suggested nvidia driver via jockey and on next reboot they get "ubuntu is running in low graphics mode" error message...09:15
brycehntr0py, http://ubuntu.com/support09:16
ntr0pyTo avoid the effect of this issue it is enough to delay the start of gdm for only 2 seconds on my system, but it leaves the causes (upstart race) untouched...09:17
ntr0pybryceh: What should i do with your url??09:23
brycehntr0py, if your interest is merely in getting your machine working with lucid, there are plenty of support channels to go through.  This is a development channel.09:25
ntr0pybryceh: I do have a working machine with a the dirty hack avoiding the effects of this issue (delaying gdm start), but i think this is a bug in upstart (race condition).09:28
brycehthen analyze it fully and submit a bug report (and perhaps a patch against the upstart rule if you're conversant in upstart)09:30
brycehhowever you might check if it's just that the nvidia kernel module didn't finish building before gdm tried to start X, which is already a known issue.09:32
brycehpatches definitely welcomed on that one, as the number of developers who support the proprietary driver is pretty limited.09:33
brycehanyway, nite.09:33
RAOFbryceh: Good night!09:33
ntr0pybryceh: I did analyse it finding out the hack and tried adding something in /etc/init/gdm.conf ("and graphics-device-added nvidiactl PRIMARY_DEVICE_FOR_DISPLAY=1"), but unfortunately im not familiar enough with upstart to do this...09:34
RAOFntr0py: Something like that would be the way to do it, yeah.09:34
* RAOF is also off.09:34
ntr0pyAny ideas how i could wait with gdm's start on nvidia module having build and beeing ready for use in upstart?09:38
seb128the current intel driver got built with the old xserver10:16
seb128should it be rebuilt to get the new abi?10:16
jcristauyes.10:17
seb128is there a list of things that need to be rebuilt still somewhere?10:17
seb128jcristau, do you know if the rebuild is hold off for a reason or if I should just do a no change upload for it?10:18
jcristaudon't know of a reason you should hold off10:19
seb128ok thanks10:20
seb128I will let a bit for somebody to comment if there is a reason or do an upload in an hour otherwise10:20
jcristauit's the middle of Sarvatt's night and RAOF said he was off an hour ago :/10:21
seb128seems quite some other drivers didn't get rebuild though10:22
seb128let's give some build retries10:25
seb128soyuz depwait handling got turned off yesterday10:25
seb128they had issues with it since the new update10:26
bigjoolsit's coming back RSN :)10:26
seb128how rsn is?10:26
bigjoolsit's getting cherry picked right now10:26
seb128which means it will land in production today?10:27
seb128ie it will be active once cherry picked?10:27
seb128or does it still need to wait a rollout?10:27
bigjoolsit'll be in production when it's picked10:27
seb128ok thanks10:27
seb128I will stop to manually retry all the drivers then10:28
asacwhats the story about gallium driver? what benefit/use case is that targetting?13:24
RAOFseb128: I've got a set of no-change rebuilds locally plus a new siliconmotion upstream to build against 1.8.  You're welcome to retry or upload no-change rebuilds for all the video drivers.13:30
seb128RAOF, seems they fixed depwait now13:32
seb128RAOF, do you plan to upload your no change rebuilds?13:32
RAOFI don't have upload rights for them.  I was going to beg for for sponsorship from Luke tomorrow.13:33
seb128do you have them ready somewhere?13:34
RAOFYeah.  I'll put them on my webserver13:34
seb128so it's just debsign and upload?13:34
RAOFRight.13:34
seb128can you give the url on #ubuntu-desktop?13:35
rippsgrr... still can't downgrade from xorg-edgers yet... when are they gonna finish building those packages.13:37
RAOFasac: The gallium architecture is meant to make it easier to write graphics drivers for modern cards.  The idea is basically to have a low-level architecture which maps easily to modern graphics chips (which are all about programmable stuff) that you can build APIs on top of (OpenGL, OpenVG, DirectX, etc)13:40
asacRAOF: so is that something that will become interesting for ARM (or already is?)13:45
jcristauas long as arm means sgx...13:46
asacok so its vendor responsibility to make that happen i guess ...13:47
asacbut is that of benefit for us?13:47
asace.g. should i go to vendors and tell them to do gallium?13:47
RAOFOnly if it makes sense for their hardware.13:48
asacplease elaborate a bit ;)13:49
RAOFThe gallium architecture only makes sense for GPUs which look sufficiently like current desktop GPUs - where there's not much fixed-function hardware, and pretty much everything is programmable and shader-y and such.13:55
RAOFIf your GPU hardware looks sufficiently like that, then building your driver using the gallium code will make it easier to support lots of API frontends - OpenGL, GL|ES, OpenVG, etc.  Video decode acceleration is also likely to be implemented there (at some point).13:59
RAOFIf your GPU hardware doesn't look like that - it doesn't have flexible shaders, it's got a lot of fixed-function stuff - then using the gallium code won't be helpful.14:02
RAOFThe classic mesa infrastructure is more suitable for fixed-function hardware, and is not going to be going away.  We still need it to support the fixed-function desktop hardware, like pre r300 ATI, and older intel chips.14:04
asacRAOF: whats your personal opinion? do you think the win through reduced efforts in writing driver frontends is worth this extra level of abstraction?14:04
asacok ... so all modern stuff is using gallium?14:05
RAOFA lot of modern development is using gallium.14:06
RAOFAlmost all of the recent work on r300-r500 cards has been in the gallium driver, and gallium offers a much easier path to get to OpenGL 3 and 4 than the classic mesa drivers.  We'll likely be shipping the r300 gallium driver instead of the r300 classic driver for Maverick.14:07
asacok so there is consent that gallium is the way forward for modern hardware. thanks for the background14:08
RAOFNouveau only has a gallium driver for nv30+.  The older chips (nv04-nv2x, corresponding to TNT → GeForce 3) have a classic mesa driver.14:08
RAOFasac: Right.  Anything which looks like a modern desktop GPU should be using gallium.  ARM GPUs have significantly different constraints, so I'm not sure how much like a modern desktop GPU they behave.14:09
RAOFHm.  Anything which is supposed to support GL|ES v2 is probably enough like a desktop GPU to benefit from gallium.  From what I read of GLES v2 it's very much about programmable pipelines and shaders, which is what gallium requires.14:10
Dr_JakobRAOF: Anything that has shaders and want to implement either GL or DX is suitable for gallium14:23
Dr_JakobWhich is all modern hardware14:24
Dr_Jakobincluding low powered ARM gpu's..14:24
RAOFDr_Jakob: Didn't know you idled here!14:25
Dr_JakobI do now :)14:25
RAOFI think I'd talked myself around to the position that ARM gpus probably wanted gallium, since they GLES v2 appears to be all about the shaders, too. :)14:26
Dr_JakobI dunno about the mali GPU but for SGX most people pay PowerVR for their closed source driver.14:28
RAOFI think we're aiming to be in a position to show demand for opensource drivers, rather than actually having any drivers currently.14:31
hyperairaaargh17:46
hyperairthird time crashing due to intel gpu hang, third time re-running mp3gain, and third time fixing corrupted banshee sqlite db.17:47
* hyperair groans17:47
ilmarihyperair: gpu hang or driver deadlock?17:48
hyperairer17:48
hyperairhow do i tell which?17:48
ilmariwhich gpu?17:48
hyperairintel gpu17:48
hyperairer17:48
hyperair2a0217:48
hyperair96517:48
hyperairX310017:48
hyperairthe first number is my pci-id17:48
hyperairthe last two are the two different model numbers my gpu appears to go by17:49
hyperairwell i'm using xorg-edgers, so i can't really blame anyone but myself for getting these wonderful hangs >_.17:49
ilmariah17:49
ilmariI used to get gpu lockups on lucid, but that stopped. now I get x/kernel deadlocks instead :(17:51
ilmarion arrandale (core i7 integrated graphics)17:51
hyperairilmari: how do you tell the difference between a gpu lockup and an x/kernel deadlock?17:53
ilmarihyperair: by SSHing in when it happens and looking at dmesg, Xorg.0.log, intel_gpu_dump18:04
hyperairilmari: ah i see. well soudn hangs for me, so i'm sure it's a deadlock then =p18:05
hyperairtime to start mp3gain again..18:05
hyperairand hope that it doesn't fail miserably this time >_>18:05
ilmarialso SysRq-d (show held locks, requires a kernel with lock debugging), SysRq-w (show blocked tasksk)18:07
hyperaireh?18:07
ilmarito show  what's blocking18:08
hyperairno sysrq keys all crash for me18:08
hyperairi mean18:08
hyperairthe sysrq keys no longer do anything.18:08
ilmarinot even sysrq-b? wow18:08
ilmari(preferrably preceded by s and u)18:08
hyperairnot even that.18:09
ilmarithat sounds like a GPU or other hardware hang18:09
hyperair=\18:10
ilmarimy deadlock is https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/56813818:11
ubot4Launchpad bug 568138 in xserver-xorg-video-intel (Ubuntu) "[arrandale] deadlock in i915_gem_madvise_ioctl (affects: 6) (dups: 1) (heat: 36)" [Undecided,Confirmed]18:11
hyperairAAAAAAAAAAAAAAAAAAAAAAAARGH!18:29
hyperairAGAIN!18:29
Sarvatthyperair: 945 and g33 are fine here, no idea whats up with yours18:39
* hyperair switches from forbid-version to hold18:39
Sarvattwhat kernel?18:39
hyperair.35-rc218:39
Sarvattmaybe try rc1 again and see if its ok there still?18:41
hyperairnah, things only went sour after last night's upgrade18:43
hyperairafter mp3gain finishes, i'll turn on X to see what's going on.18:44
Sarvattmachine dead when it hangs or can you get any more info out of it?18:45
Sarvattthere were huge changes in 965 mesa yesterday18:45
hyperairhaaaaaaaangs.18:45
Sarvattmaybe try disabling compiz?18:46
Sarvattso ya can blame mesa18:46
hyperairand go non-composited for hours on end until it hangs?18:47
hyperairnothanks.18:47
hyperairi'd end up ignoring the machine, which would in turn ignore me and refuse to hang.18:47
Sarvattmetacity compositing? :D18:49
ThisOtherGuyhi all - can anyone help me with this: http://pastebin.com/7wdqrVdH18:54
hyperairSarvatt: ._. that sounds terrible.19:11
hyperairSarvatt: i really meant compizless though. gnome-shell is still tolerable, but i can't really go without compiz's extra features, particularly scale for long before cracking and switching to my desktop19:12
Sarvatthyperair: build compiz from git where opengl is optional? :D19:13
hyperairSarvatt: don't have the plugins require opengl?19:14
hyperairSarvatt: actually i've been meaning to, but at the rate i'm crashing, i'll never get to clone compiz from git.19:14
Sarvatti dunno, lets see. i dont think they do though19:15
Sarvatt    <requirement>19:15
Sarvatt<plugin>opengl</plugin>19:15
Sarvattso scale does :(19:15
hyperairheheh19:16
Sarvattcould always just purge xorg-edgers and use 2.11 from x-updates :D19:20
Sarvattof course mesa is going to suck then19:20
=== asac_ is now known as asac
Sarvattwhy did we make nvidia-graphics-drivers provide a video abi again? it supports multiple abi's with the same package..19:31
Sarvatt(needs to be no change rebuilt too)19:31
Sarvattnvidia-173 and lower dont work with 1.8 either yet, i think they'll work with the IgnoreABI option though19:33

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