[00:00] tjaalton: Certainly. Could you add the reproduction steps from the patch to the bug? [02:25] RAOF: hm, am I mistaken or is 1aud$ still .50$ in terms of what you can buy with it, except the echange rate says otherwise? [02:26] ml|vacation: Depends on what you're buying. [02:26] food [02:26] ml|vacation: *Domestic* prices haven't changed much with the higher AUD, obviously. [02:26] It's now substantially cheaper to buy things from the US in AUD, though. [02:27] yeah :s [02:27] but as an outsider paying 4$ for some honey... [02:29] used to pay half of that or less :) [02:31] Yeah. Sucks to be travelling *to* Australia. [02:34] unfortunately for you if the exchange rate n [02:34] ormalizes again, they use that as excuse to bump the prices* [02:41] RAOF: alrighty, I'll add that [02:42] actually those steps need a newer xinput than what precise has, but you can simulate it with suspend/resume [02:43] Whatever works :) [02:44] ml|vacation: hey, your valgrinding helped fixing the synaptics bug :) [02:44] RAOF: ha, dronus added it already [02:44] but I'll edit the header [02:50] tjaalton: great :) [02:50] will it be sru'd to precise too? :) [02:51] uploaded already :) [02:51] and accepted, it seems [02:51] and works, I verified it [02:52] so all the scary x bbugs are gone now? [02:52] nice that the new version of xinput lets you disable/enable devices [02:52] not exactly [02:52] there's still this weird intel gpu hang causing compiz to spin (black screen, cursor changes). happens on quantal too :/ [02:53] oh, but the xi bugs are gone? [02:53] and thumper said he could reproduce it immediately after login with the proposed unity stack from a ppa [02:53] ah, hope so [02:53] (i tried to reproduce the hang but couldn't) [02:53] new driver? [02:53] oh hey :) [02:54] ok well great, back to vacationing, ttyl :) [02:54] ml|vacation: yeah, cya [02:54] thumper: so, I can reproduce it but only on resume, and not that frequently [02:55] thumper: was wondering if your compiz/unity setup was something else than the ~default I have [02:56] yes, yes it is [02:56] I'm running pre-quantal branches [02:56] as in, stuff that should land in quantal soon [02:56] me too, but the configuration of compiz? [02:56] ppa:unity-team/staging has new compiz and unity [02:56] nothing special [02:56] k [02:57] sorry, have to go [02:57] time to get kids [02:57] and take the rest of the weekend off [02:57] kiwi pycon \o/ [02:57] mine will wake up in 1-2h [02:57] ew [02:57] :P [02:57] early [02:57] that's rough [02:57] tjaalton: Is that compiz spinning, or compiz blocked waiting for a vblank? [02:57] hi RAOF [02:58] thumper: Yo! [02:58] thumper: got the strace somewhere? [02:58] RAOF: it is inside the driver [02:58] * thumper digs [02:58] could get it from irclog.. [02:58] actually, my other laptop is hung like that now [02:58] Is compiz burning CPU, or is it just not doing anything? [02:58] I could dig one from that too [02:58] not burning cpu here [02:58] unless I try to restart it iirc [02:59] http://pastebin.ubuntu.com/1173379/ [02:59] RAOF: I do have to run [02:59] RAOF: no, compiz is idle [02:59] Ooooh. [02:59] 100% cpu on something [02:59] but none of the htop line items [02:59] so kernel maybe? [02:59] * thumper really runs [03:00] tjaalton: You've got the same thing? [03:00] RAOF: dunno of the trace, but hung yes [03:00] Can you attach gdb to compiz? [03:00] probably, should get the symbols first [03:01] Yeah. [03:02] From thumper's trace, I guess one of two things: (a) compiz has taken out a ServerGrab and either it's on one of its other connection or mesa has its own X connection, or (b) compiz is currently GLX throttled, which would happen if it's got a SwapBuffers pending. [03:02] this is bug 966744 btw [03:02] Launchpad bug 966744 in compiz (Ubuntu Quantal) "[i965] Resume from suspend leaves me with black screen or a screen of the desktop before it suspended. Compiz hung in intel_update_renderbuffers() from intel_prepare_render() from brw_draw_prims()" [Critical,Triaged] https://launchpad.net/bugs/966744 [03:07] Right. So, Chris' UXA fix would be a fix for a pending swap that'll never complete, suggesting (b) [03:08] those are already in quantal [03:08] and on the package for precise I put there, but someone could reproduce it still [03:08] Which means we've got another bug; and the same reasoning applies. [03:09] right, entirely possible :) [03:09] If you can attach gdb to compiz, I'd guess you're blocking in _XReply, like thumper. [03:10] The next thing to do is to attach to the server, and see if you can match up compiz' Client. [03:10] Well, actually, the *next* thing to do is to attach to the server and see if someone's got a servergrab, because that's easy. [03:11] http://pastebin.com/raw.php?i=qBtsGKcE [03:11] it's the same [03:11] (got some more symbols) [03:11] Right. So, it's most likely either a server grab or the client being GLX throttled. [03:12] We can't really tell from the client side. [03:12] ok how do I check the server= [03:12] ? [03:12] Attach to the X server; there's a global serverGrab symbol you can print. [03:15] Sorry, grabState [03:16] $1 = 0 [03:17] Ok, so the quick and easy test is failed. [03:17] bummer [03:18] Oh. Is the server blocked? [03:18] ie: where was the server when you connected gdb to it? [03:19] WaitForSomething [03:19] Yeah, that'd be right. [03:26] Hm. [03:26] You know, you could iterate through the Client list, and see if any are throttled by GLX. [03:28] ok, how do I do that? [03:31] my gdb-foo is limited :/ [03:34] There's a global clients[] array. [03:36] should I know it's length to print it? [03:36] It's of length MAX_CLIENTS [03:37] Each client will have an ->ignoreCount; if that's positive, the X server is ignoring it. [03:38] sure it's 'client[]'? [03:38] uh [03:38] too early :) [03:39] ok, got pointers [03:40] I mean, it returns ClientPtr [03:42] 30 clients [03:43] Right. [03:43] Are any of them ignored? [03:44] You can use such constructs as for() and such :) [03:45] still figuring out how to get that info from the pointer :) [03:48] It should be the ignoreClient member of the ClientPtr; ?print clients[0]->ignoreClient? should print an integer? [03:48] * RAOF spots a probable bug in IgnoreClient/AttendClient [03:49] There is no member named ignoreClient. [03:49] Sorry, ignoreCount [03:49] Sorry. [03:49] yeah [03:49] ok let's see.. [03:51] ok got one with = 1 [03:51] ignoreCount = 1 [03:52] That's *likely* to be the compiz one. [03:52] Particularly if it's the only one with ignoreCount = 1 [03:52] yeah just one [03:54] To be sure, we'd want to work out whether any of Compiz's drawables have DRI2DrawablePtr->swapsPending > 1. [03:54] I can't, off the top of my head, think of a way to verify that, though. [03:55] I think its' reasonable at this point to blame this on the intel driver scheduling a swap that never completes. [03:56] yeah, I could show the trace to #intel-gfx once someone is awake [05:04] tjaalton: \o [05:04] so here's some fun news, I got some time and messed about with grub a bit [05:04] I can now disable the ATi discrete graphics [05:05] still have to boot with nomodeset since the lvds dual channel patches are not there in the ubuntu kernel [05:05] in dmesg : [ 20.258413] i915: `2' invalid for parameter `lvds_channels' [05:06] unfourtunately, I don't how to compile my own kernel with lvds patches [05:20] shadeslayer: huh, someone clearly failed to send them upstream then [05:21] dunno, I think I saw them on lkml at one point [05:21] though I don't follow lkml so I don't know [05:21] atleast now the thing doesn't heat up like a frying pan [05:22] right === chrisccoulson_ is now known as chrisccoulson [09:18] tjaalton: yay just applied the fixed version to this laptop, think it was having the same issue on suspend ;) [09:19] yep [09:19] wait, another bug?= [09:19] -= [09:19] no, just didn't have proposed enabled [09:19] ah [10:08] mesa 9.0 branched [10:08] great :) [13:03] * apw is seeing some background corruption on quantal, trying to work out if this is X or compiz [13:04] apw, might be our new cubistic theme for ubiquity .... [13:04] https://launchpadlibrarian.net/113965811/IMG_0170.JPG [13:04] :P [13:05] ogra_, luckily not that :) [13:21] apw: there's a known regression on intel < 965 [13:21] compared to mesa 8.0.x [13:22] tjaalton, we got any idea of a fix or is beta shipping wit this? [13:23] tjaalton, this is an atom netbook, would i expect that to have the right version [13:27] if it's i945, that's your bug [13:27] tjaalton, how do i tell [13:31] i just learned about the bug today [13:32] lspci [13:32] also, we have a snapshot that's over a week old [13:35] bug 1042211 [13:35] Launchpad bug 1042211 in mesa (Ubuntu) "[quantal] [regression] [i915] Corrupted display, desktop and menus don't repaint correctly using Mesa 9.0 (8.0.4 works)" [Critical,Triaged] https://launchpad.net/bugs/1042211 [13:36] tjaalton, sysfs says this is an is_g33, and the pci_id points to intel_pineview_info ... so i assume its not 945 [13:37] the damage i am seeing could easily be compiz as some of it is aligned well with the places it puts things [13:37] tjaalton, that said the description is definatly the same i recon [13:38] apw: yeah sounds like it's no the same bug [13:38] oh :) [13:38] well, try the previous packages [13:38] bacground a solid colour, menus appear and return to menu background [13:40] so the background being mashed is being triggered by nautilus running there [13:45] tjaalton, yeah seem to be this issue as far as i can tell, all the mitigations work as per daniel, though performance is mindbogglingly slow with them [13:47] apw: but does downgrading fix it? [13:49] which packages do nead, there seem to be 40 binary packages here for mesa [13:52] libgl1-mesa-glx, -dri should be enough [13:53] the bug has links [14:06] tjaalton, thanks, here goes nothing [14:07] needs -glapi as well for reference [14:08] tjaalton, ok confirmed, downgrading those mesa packages fixes it here [14:09] ah [14:10] tjaalton, added the information to the bug, looks like basic netbooks are affected [14:11] i believe it has i945 gfx, it should use i915_dri [14:15] but anyway, too late to update the snapshot today, but monday :/ [14:15] hehe :p [14:16] argh forgot to copy the whole git tree over, linux kernel has no git history now :s [14:16] started looking at the update but didn't get far before eod [14:18] gone again -> [14:25] ml|vacation, sounds like a mistake :)