[00:08] mnemo: I sketched out https://wiki.ubuntu.com/X/KernelModeSetting - please flesh it out further based on your testing [00:09] bryce: For me, UXA with KMS is completely stable, unlike without KMS. [00:10] wgrant: i have the opposite experience on G45 [00:11] mnemo: I'm an old i915. [00:11] The cursor flickers a lot more, but otherwise it's good. [00:13] wgrant, mnemo: thanks; I've added a section for reporting results to https://wiki.ubuntu.com/X/KernelModeSetting - mind adding your info? [00:13] one cool thing i915 is that windows vista never shipped WDDM 1.1 drivers for it so it cannot run Aero afaik... which means that with UXA ubuntu is basically delivering a better and more capable driver than windows.. that rocks... see for example: http://www.tinyscreenfuls.com/2007/04/video-why-intel-915-graphics-dont-have-a-wddm-driver-for-vista/ [00:13] bryce: its ok you can add it [00:14] bryce: exact card is: [00:14] 00:02.0 VGA compatible controller [0300]: Intel Corporation 4 Series Chipset Integrated Graphics Controller [8086:2e22] (rev 03) [00:14] mnemo: can you add to it please? [00:14] bryce: i've only tried KMS on jaunty with a kind of messy config actually [00:14] maybe I should upgrade my intel box to karmic as well [00:17] bryce: are you sure you can load the i915 module through /etc/modules like that? i mean doesnt modesetting have to load like really early for it to be useful? I thought thats why the debian kms wiki page said it needs to go into initramfs etc? [00:18] bryce: What do you want for 'version'? Driver version? [00:18] yeah the way he posted works fine for me but on 2.6.30-2 [00:18] mmkay [00:19] wgrant: yep [00:19] i think you have to manually load some stuff prior to i915 if you dont enable KMS by default in the kernel like in 2.6.30-1 like the debian wiki says [00:19] mnemo: no I'm not sure, so please correct it to what you find in practice. i'm just copying what others have told me [00:20] but all you need is the modprobe option on 2.6.30-2 [00:20] (maybe I should have just left the page blank) [00:21] maybe we can ask some intel guy to do some vetting on that page later on [00:22] anyway, don't take anything written there as gospel; rewrite the whole thing if it makes more sense [00:22] I just pass the option on the kernel line in GRUB. [00:22] only reason you need the modprobe.d option is because of this http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-karmic.git;a=commit;h=24ea0d5ee9c0bac3b9aab43e2761394dd3dcf413 [00:22] eventually I plan to work on kms stuff myself, but I need to focus on getting more UXA bugs upstreamed first [06:08] Hi. I'm trying to track down how wacom's don't report some information to Xinput. [06:08] heh [06:08] I'm wondering if part of the issue is using xf86Xinput.h rather than Xinput.h [06:12] Am I looking down the wrong path? The documentation on xinput seems light, but I(m confused about the relationship between libxi and xserver-xorg-input-foo. Is libxi the client library only, receiving events from the server? [06:22] At least I found out why there were four devices: pad, touch, tool nib, tool eraser. [06:34] Hrm. Seems it's running in some sort of compatibility mode, and was never ported to XInput. I'm still not sure I understand the architecure properly, but the complete lack of any XIfoo calls seems suspicious. [06:35] * persia defers to someone who actually has hardware, as at this level of surgery, random hacking has a low probability of success [10:51] persia: yes, libXi is the client side stuff. the driver shouldn't touch that. [10:52] persia: the driver calls server functions, which are declared in /usr/include/xorg/*.h [12:46] jcristau, OK, and things that manipulate properties should be including inputstr.h ? [12:48] persia: what "things"? [12:49] Well, I'm trying to reinstate the propagation of tool and pad serial numbers to X in the wacom driver (or at least interested in documenting what needs done, as I don't have a wacom to test). [12:50] The sane way to do it seems to be to set properties inside DeviceIntRec, from my limited understanding of how things work. [12:51] I'm suspecting these will be represented to clients and can be collected with XGetDeviceProperty() in a client. [12:53] But I'm not sufficiently confident of my understand of how things work to know if this is correct. [12:53] s/understand/understanding/ [13:02] right the driver keeps some stuff in its private data, and tells the server about it with XIChangeDeviceProperty [13:07] Ah, from /usr/include/xorg/exevents.h. I've been mistakenly looking in /usr/include/X11. Thanks! [13:08] and then clients use XListDeviceProperties/XGetDeviceProperty from libXi [16:36] any tips to get usable video on VGA compatible controller: Intel Corporation 82830 CGC [Chipset Graphics Controller] (rev 03) [16:37] characters get blurrier as time goes on [16:37] https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/213171 [16:37] Ubuntu bug 213171 in xorg "[i830] Unable to install with GUI on Fujitsu Lifebook C7651" [Unknown,In progress] [18:47] bryce: you there? [18:48] yes [18:49] any help on all the characters on my screen turning in ugly blurred blocks over time [18:49] say in 3-5 minutes after login [18:49] 00:02.0 0300: 8086:3577 (rev 03) [18:50] I can't even really see what I'm typing, really nasty stuff [18:51] alex_mayorga: sorry no idea, sounds weird [18:52] yeap, really weird, never seen this before [18:53] actually I had to paste your reply on a terminal to see what it said [18:53] oddly the characters on the terminal look fine [18:53] what might it be? [18:55] is there a way to query my LCD for its native resolution? [18:56] xrandr [18:57] alex_mayorga: install orca (screenreader) :-P [18:58] JanC: definitely an option hehe [18:58] xrandr marks the current res with * and the preferred one with + [18:59] so 1400x1050 60.0*+ it is then [18:59] alex_mayorga: yup [18:59] alex_mayorga: which driver version do you have? [18:59] mnemo I had to resort to 2.4 I believe https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/213171 [18:59] Ubuntu bug 213171 in xorg "[i830] Unable to install with GUI on Fujitsu Lifebook C7651" [Unknown,In progress] [19:00] that bug has my details [19:00] well the comments with my name on it that's it [19:00] mnemo: anything I can do to get this one resolved I will [19:01] alex_mayorga: step up as the maintainer for those older chipsets :) [19:01] If I use 2.7 is like the four quarters of the screen were overlaid in the top left quarter [19:01] alex_mayorga: but no corruption with 2.7 ? [19:01] mnemo: that's the plan, but I'm pretty much a n00b here [19:02] mnemo: yes it's way more unusable [19:02] bryce mentioned something on the modaliases [19:03] alex_mayorga: how old is this machine btw? [19:04] mnemo: you don't want to know :) [19:04] but is all that I've got, probably 7-9 [19:04] :( [19:10] blockiness is coming back :( [19:12] seems like I can get back to usable if I change the resolution [19:18] but from there it starts to degrade to blocky stuff [19:19] I guess downgrading to 8.10 is my only way to ubuntu on this machine [19:19] mnemo: any other leads? [19:27] what's in theory the best driver for i830M cards? === virtuald_ is now known as virtuald [19:40] bryce: do you know what was the last driver to support i830M? [19:41] alex_mayorga: dunno, maybe i810 [19:42] current code supports i830 [19:42] thought not all configurations [19:42] some dvo chips aren't handled [19:46] might this be it? https://bugs.freedesktop.org/show_bug.cgi?id=16928 [19:46] Freedesktop bug 16928 in Driver/intel "[830M] Garbage and incorrect redraws" [Normal,New] [19:49] "disappears with "Option" "AccelMethod" "XAA"." where do I put this? [19:52] should I give up and just start saving for a new laptop :( [19:54] alex_mayorga: well you could jump onto that bug and say you're willing to test fixes [19:54] you could also try disabling render accel [19:55] jbarnes, the freedesktop one? [19:55] yes [19:55] if a bug's not filed at fd.o it doesn't exist ;) [19:55] yeah, that's my saying, bug or it didn't happen [19:56] how do I go about "disabling render accel"? [19:56] with exa it's Option "exanocomposite" [19:57] you can also completely disable acceleration with: [19:57] Option "NoAccel" "true" [19:58] bear with me here, I guess all that you mention goes under xorg.conf somewhere right? [19:58] yeah [19:58] yes, inside the "Device" section [19:58] OK, now that's a start :) [19:59] please forgive me, this are my first dabblings at xorg bugs [19:59] what should I try first? [20:00] pick one [20:00] exnocomposite is probably a bit faster I think [20:00] should I first go to 2.7 driver [20:01] I'm on 2.4 and it kind of works, except for the blocky/blurry characters over time [20:02] jbarnes: are you on #intel-gfx as well? [20:03] alex_mayorga: yes [20:05] mnemo is exanocomposite or exnocomposite [20:05] exa [20:06] OK, I'll try these 3 options [20:06] there's no X restart in jaunty anymore, right [20:06] 'sudo killall X' [20:06] is there a way other than reboot? [20:06] logout should do [20:07] OK, let see how it goes [20:07] or maybe it's 'sudo killall Xorg' instead [20:07] since it restarts the server [20:08] killing xorg shouldn't kill my pidgin and such, right? [20:09] umm yes it does [20:09] kills your session... [20:09] I'll wait a bit then, rather not close my VPN at this time [20:19] one question remains, should I stay on 2.4 or try the options suggested on 2.7? [20:20] try both [20:20] you need to systematically document your issue to zero in on the bug [20:24] mnemo [20:24] thanks I'll try that out [22:42] http://people.ubuntu.com/~bryce/totals.svg [22:47] bryce: would it be easy to do other similar graphs for other set of components? [22:48] possible, but not easy; I have a lot of hacks that are kind of xorg package specific there [22:50] seb128: as a starting point, I use packages xorg-swat is subbed to - https://bugs.edge.launchpad.net/~ubuntu-x-swat/+packagebugs [22:50] we have a similar list of desktop-bugs [22:50] if the code you use for the graphs somewhere online? [22:50] yes it is in my arsenal tree [22:52] seb128: I'll put on my todo list to clean it up and post it for you when I get some time [22:52] thanks [22:52] no hurry, I was rather curious about it but we don't have a real need for stats [22:52] oh ok [22:53] if that was just a "it works for any team just change the url or packages list" I would have given it a try [22:54] but it seems that's not the case which is ok too [22:55] yeah I know not everyone is as obsessed with stats as me, so haven't really bothered to generalize the tools that much ;-) [22:55] but maybe someday [23:33] bryce: more evidence of vbetool fail [23:33] bryce: running it will cause interrupts to break on some machines, leading to rendering hangs [23:33] jbarnes: ok [23:34] jbarnes: details? [23:34] so on intel it should never be run... sounds like some of the thinkpad tools do it in X startup scripts [23:34] bryce: fdo bug 20896 [23:34] Launchpad bug 20896 in firefox "Launching Firefox twice produces Choose Profile dialog" [Low,Fix released] https://launchpad.net/bugs/20896 [23:34] I think we've mostly excised use of it in ubuntu setup scripts [23:34] ubottu: freedesktop bug 20896 [23:34] Freedesktop bug 20896 in DRM/Intel "[GM965 KMS] X does not draw untill mouse is moved. Probably IRQ problems" [Major,New] http://bugzilla.freedesktop.org/show_bug.cgi?id=20896 [23:34] Error: I am only a bot, please don't think I'm intelligent :) [23:35] bryce: great [23:35] bryce: just a heads up [23:35] cool, thanks I'll keep an eye out [23:35] do you know of any plans to modify vbetool to not trigger the error? [23:36] I don't think vbetool is the problem per-se [23:37] all it does is call into the VBIOS afaik [23:37] and on some machines apparently the VBIOS goes off and clears regs it shouldn't [23:37] bryce: apparently it's the acpi-support package from debian that caused this particular problem (or maybe vbetool itself) [23:37] not sure which X session scripts either of those patch at install time [23:38] oh interesting, ok I'll check that out [23:38] we used to have a package xresprobe which wrappered vbetool [23:38] but we knew about the problems it could cause on -intel, I recall gutting the tool to prevent it calling it [23:39] for some reason I recall it affecting laptops more than desktops, but I could be misremembering [23:39] anyway, we've moved away from using xresprobe for anything now [23:39] cool [23:40] ah yes, acpi-support calls vbetool [23:40] ./resume.d/55-screen.sh: vbetool dpms on [23:40] ./resume.d/15-video-post.sh: vbetool post [23:40] ./suspend.d/90-framebuffer-stop.sh: vbetool dpms off [23:42] theres no /etc/init.d/vbesave in ubuntu's acpi-support like that guy in #intel-gfx has on debian though [23:42] https://bugs.edge.launchpad.net/ubuntu/+source/acpi-support/+bug/31425 [23:42] Ubuntu bug 31425 in acpi-support "garbled or no video output after resume from suspend (Inspiron 4150)" [Medium,Confirmed] [23:43] jbarnes: ^that sound like a match? [23:43] yeah could be [23:44] hmm, I would expect to see a larger number of bug reports [23:48] I don't think all vbioses do it [23:48] also it mainly affects kms apparently [23:48] due to the ordering of irq enable [23:49] also from the fdo bug report the symptoms are kind of generic so could be matches without any indication vbetool was involved [23:50] jbarnes: do you know which specific symptoms were the result of the vbetool call? Or are all the symptoms caused by it? [23:50] symptoms would have been "all rendering stops" [23:50] mouse movement or some other interrupt source may have allowed it to continue [23:50] but that's config dependent I think [23:51] ok, it doesn't sound super familiar but I'll troll through our bugs and see [23:51] actually [23:51] I just finished going through ALL of the UXA bugs and didn't see that issue [23:51] ok so far I've only seen it with KMS [23:51] so maybe you don't have a report [23:51] if it mainly affects just kms... yeah [23:52] oh btw [23:52] https://wiki.ubuntu.com/X/KernelModeSetting [23:52] jbarnes: that's a start of a howto page for KMS, with a place for gathering test results === maco_ is now known as maco [23:53] we've not started pimping it yet since there's still some infrastructural bits to square away but I'm planning on promoting that for more extensive KMS testing [23:53] bryce: cool [23:53] hopefully we can get enough of the UXA bugs resolved to switch over to UXA/DRI2/KMS by default [23:54] I just forwarded you a list of upstreamed uxa bugs I sent to yingying btw [23:54] ok [23:55] btw, are you guys no longer accepting EXA-specific bugs in the bug tracker now? [23:57] bryce: if they're high prio and affect 2.7 we could look at them [23:58] since we're doing 2.7.x releases [23:58] I hope yingying's team could help with those too