=== Amaranth_ is now known as Amaranth [01:33] heya [01:33] Good morning! [01:33] Or possibly evening? [01:34] You crazy past-livers :) [01:34] heh, yeah it's Sunday evening, just got back from trip to the park with my son [01:35] Sounds fun! [01:35] RAOF, btw I disabled the gestures patch (101) in -evdev. It was causing major X crashing fun [01:35] bryceh: Yeah, I noticed the bug. [01:35] seems to be an amd64 specific bug, which may explain why I didn't see it when I tested [01:35] Pity the backtrace is so anaemic :( [01:36] I tested on amd64 [01:36] hm [01:36] if you look at the bug report, one of the commenters suggested the issue was an abi incompatibility. you might take a peek at the comments [01:36] it was a bit speculative, I decided to just flip it off until someone could investigate more thoroughly [01:37] Yeah, I agree with that call. [01:37] probably best to leave it off until post-alpha-2 [01:37] It'd be nice if cnd would pop up, be a dear, and look at his gesture patches :) [01:38] fwiw, I didn't see the synaptics misbehavior LLStarks mentioned, but did mention the possibility for issues in the a2 release notes [01:38] RAOF, yah [01:38] duly noted [01:38] Their problem is that synaptics just isn't being loaded for their touchpad. I'm not sure why. [01:39] btw guys, is tty broken? i can't drop to tty1 or tty2. [01:39] blinking cursor only [01:39] Not broken on any of my systems. [01:40] was talking to the AMD guys friday, they're a bit concerned if xserver 1.10 goes in today, and there are any respins required for subsequent bug fixes, it could throw off their build schedule [01:40] can't kill x by killing gdm either [01:40] LLStarks, are you running xorg-edgers? [01:40] no [01:40] LLStarks, natty current has been working fine for me [01:41] i flip-flop as needed [01:41] bryceh: As in - if there are any further ABI respins required? [01:41] RAOF, as in, any bug fix uploads to xserver or mesa [01:42] How do non-ABI-changing bugfix uploads interfere with their build schedule? [01:43] they have to start it over [01:43] Because they need to certify it against a particular upload? [01:43] I guess so; they were complaining about it but was hard for me to get specifics [01:44] That's… awkward. How have they handled these things in past releases? We upload bugfixes all the time! [01:44] Perhaps they're wary about further ABI breaks? There's an ABI break between RC1 and master, and *possibly* another one between master and RC2. [01:44] nah, there's always been a window where we're frozen for a day or so leading up to the alpha release [01:45] apparently arm builds take longer than x86 ones [01:45] why can't ati devs be less anal and put out more releases like nvidia? [01:45] * RAOF will take specs and full-time open-source devs over a more responsive binary driver team anyday. [01:46] RAOF, anyway as a compromise I took out some of the drivers from the video-all arm script [01:46] waiting for their advance releases less than 2 weeks from an ubuntu release isn't fun [01:47] RAOF, you mean like the way intel handles things. that works really well. [01:47] RAOF, anyway my feeling is if we need a bug fix in, we need a bug fix in. But I suppose if we can batch up patches and minimize uploads it might make them a little happier :-) [01:47] Yeah. [01:47] I think the key thing is to get xserver in asap before the alpha-2 freeze is announced [01:49] RAOF, btw it looks like the .38 kernel upload Thurs/Fri nuked fglrx [01:50] I'm a bit concerned about further ABI breakage in xserver now. Keith's ‘I'm sick’ email promised to pull an ABI-breaking DGA branch sometime; I need to see how much that breaks ABI we care about. [01:50] hrm [01:51] If it *is* ABI breakage we need to care about, is it worth uploading Xserver pre A2, only to need another rebuild of the world post A2? [01:52] well, it simplifies a few things if we postpone xserver update 'til after alpha2 [01:52] but it somewhat reduces testing coverage we'll get [01:54] RAOF, how much is left to do to get xserver in? [01:54] Finish updating the gesture extension patch, or getting one from Chase. [01:56] I wonder if he's jetsetting, or just not checking email/irc :/ [01:59] RAOF, aside from that, any other packaging work to do or is the xserver git tree good for upload? [01:59] It needs a changelog entry; that's it. [02:01] RAOF, I'm leaning towards just leaving the gesture stuff disabled if we don't hear from him by, say, morning my time, and uploading [02:01] then when 1.10 is officially released we can do another mega-driver rebuild [02:02] bryceh: Wilco. I'll have a package ready for you then, then :) [02:02] alrighty, I can upload if you want to do it now [02:03] Nah, I'll give cnd until your morning. [02:03] ok [02:03] wow take a gander - http://www.bryceharrington.org/X/Reports/ubuntu-x-swat/totals-natty-workqueue.svg [02:03] Incidentally, were you planning to push wayland and/or libxkbcommon to pkg-xorg git? [02:04] Spikey McSpike visited our graph [02:04] Is that Maverick trend the line at the same stage of release? [02:04] no, I might put it in bzr [02:04] yeah [02:04] Why do we have so many fewer open bugs?! Hurray! [02:05] I credit that to there being 2 X guys ;-) [02:05] Looks like Spikey McFglrx happened :) [02:05] And I see we've got some intel testers finally :) [02:06] I've been going through all the open natty bugs every morning, so that's kept things better under control than any release before [02:06] Aaah, yeah. That'll do it. [02:06] yeah the fglrx bugs are all dupes of "kernel broke it" [02:06] a bunch are probably from the evdev crash [02:07] we may also have a few of "xchat shows corruption with latest intel [02:07] KiBi's doing some wayland/xkbcommon packaging in pkg-xorg git; it might be nice to help him if you've got time. [02:07] Oh, yeah. That intel bug. [02:07] RAOF, I'm popping up :) [02:08] ah no, looks like mostly GPU lockup bugs [02:08] got back from bangalore yesterday [02:08] cnd: Time to make good on those “you'll not need to forward-port patches for us” promises :) [02:08] still beating back the remnants of a fever suffered while there [02:08] RAOF, is KiBi aware of my packaging? [02:08] Oh, sucks. [02:08] RAOF, yeah, so, things have changed recently in the utouch world [02:08] bryceh: Yeah, I think so. [02:09] rydberg is unfortunately no longer on contract with us [02:09] so we've had to scrap our plans to put the gesture recognition on the client side of X [02:10] Oh. So we'll still have a gesture extension for 1.10? [02:10] we will be forward porting the maverick gesture patches for xorg-server and xserver-xorg-input-evdev to 11.04 [02:10] I have begun this work [02:10] got started while I was in india [02:10] so here's the good news: everything forward ports pretty easily [02:11] here's the not as good news: it needs to be reworked a little if it's going to sit along side XI 2.1 [02:11] though mostly that rework sits in utouch-grail [02:11] not in any x components [02:11] (Although *my* attempt at forward-porting it in evdev results in Xserver crashes on hardware that isn't mine) [02:12] RAOF, my forward port was also based on evdev master [02:12] which has a *huge* masked valuators patch from me in it now :) [02:12] well, series of three patches [02:12] So we need a newer snapshot of evdev also? [02:13] RAOF, lets first go over what we need to accomplish, and by when [02:13] when are you wanting to push things to ubuntu? [02:13] We'd like to push xserver in before A2 freeze, so soon. [02:13] RAOF, can you be more specific? [02:13] just so I know [02:14] A2 freeze will be Tuesday. [02:14] *early* Tuesday [02:14] oh, so soon really means about 24 hrs :) [02:15] cnd, yep, preferrably more like 15 hrs [02:15] so here's two routes: [02:15] a. gestures in xorg-server and evdev [02:15] b. gestures and xi 2.1 in xorg-server and evdev [02:15] I feel about the same confidence in both at this point [02:16] mostly because I haven't tried a [02:16] b. also requires an xi 2.1 patch series, right? [02:16] however, b would require a new min version of utouch-grail, which would need to be pushed in as well [02:16] yes [02:17] RAOF, I would be happy to add the patches to xorg-server and evdev myself [02:17] to the git repo [02:17] so you would only need to verify and upload [02:17] That would be nice. I'm perfectly happy to apply patches myself if that's easier. [02:18] it's probably easier if I do it anyways [02:18] and I don't mind [02:18] Is the utouch-grail min version the one in bug #702637 ? [02:18] Launchpad bug 702637 in utouch-grail (Ubuntu) "Upload utouch-grail 1.0.18 to Ubuntu (affects: 1) (heat: 186)" [Undecided,Confirmed] https://launchpad.net/bugs/702637 [02:18] no, it would need to be 1.0.19, which doesn't even exist yet [02:19] without changes I made, touchscreen devices would no longer work [02:19] they wouldn't move the pointer [02:19] but they aren't huge changes [02:19] I can try to make the packages tonight [02:20] what do you think? a or b? [02:20] it's really ok if you want a for alpha 2 [02:20] the xi 2.1 work isn't final yet either [02:21] still some bits needing implementation [02:21] Are we more likely to get more useful testing out of (b) than (a)? I think we will, but I think that's the main criterion at this point. [02:21] bryceh, any news about the 16-bit color appearance on intel with .38? [02:21] LLStarks, news? [02:21] RAOF, there's already someone on our multitouch mailing list asking about xi 2.1 for games [02:21] he's just waiting for it to hit [02:21] x session-wide color banding [02:22] That sounds fun. Would you *like* his testing at this point? [02:22] the sooner the better :) [02:22] cnd, are a and b about the same level of risk of breakage? [02:22] LLStarks: Sounds like a failure to set an appropriate dither mode. [02:22] bryceh, b would be slightly more risky for MT touchscreens [02:23] I don't think it should be any more for MT touchpads [02:23] LLStarks, since it's the weekend I've only been looking at critical bugs, but don't recall seeing that one. bug #? [02:23] and should not be a factor for non-mt devices [02:23] **** [02:23] i never filed one [02:23] LLStarks, ah, then that would be your next step [02:23] i'll do it after the ppv ends [02:23] need to boot into .38 anyway [02:23] ok [02:24] LLStarks: FWIW I don't get it on my 6-bit panel. [02:24] . [02:25] i'd post camera pics, but it's kinda hard even with a 14mp one [02:25] bryceh, if we go with b and find it breaks things, we should be able to revert to a before alpha 2 actually ships [02:25] cnd: And the risk on those MT touchscreens is that the pointer doesn't work on the touchscreen, yes? Rather than it causing puppies to combust? [02:25] RAOF, yes [02:25] so you could just use a mouse or trackpad if something is awry [02:26] As someone without access to a MT touchscreen, that sounds an acceptable risk :) [02:26] it's just cause of the interaction between utouch-grail and xi 2.1 [02:26] yeah at this stage the risks we're mostly averse to is risks of widespread breakages (ala the -evdev breakage) [02:27] bryceh, which evdev breakage? [02:27] but also we want to minimize the number of times we upload xserver between now and alpha-2, so the fewer bug fixes needed the better [02:27] I forward-ported the gesture patch to 2.6 in such a way that caused flaming destruction on !my hardware. [02:27] ahh, that breakage :) [02:27] cnd, oh yeah you need to take a look - the gestures patch hosed the evdev upload [02:27] (bug #709977 for your delectation) [02:27] Launchpad bug 709977 in xserver-xorg-input-evdev (Ubuntu Natty) (and 1 other project) "Xorg crashed with SIGSEGV in RemoveDevice() - segfault at 1010 error 4 in evdev_drv.so (affects: 37) (dups: 15) (heat: 210)" [Critical,Fix released] https://launchpad.net/bugs/709977 [02:28] RAOF, ahh yes, I know exactly where the forward port went wrong :) [02:28] unfortunately I didn't catch it before uploading (but I *did* test this time!) [02:29] I tested on multiple machines, none of which exploded. [02:29] On the other hand, none of which had anything multitouch, either. [02:29] anyways, if you guys give me the thumbs up I'll go make gesture + xi 2.1 packages for you [02:29] You have a go from me. [02:30] yeah I'm ok if we can get it in quickly [02:30] RAOF, would it be reasonable to collapse all the xi 2.1 git commits into one big uber-patch in our package? [02:30] that uber-patch would then be updated as we update xi 2.1 work [02:31] I don't have any real feelings either way. I expect that you'll be the one dealing with any fallout from that patch, so I think do whatever is most convenient for you. [02:31] k [02:31] that would be it [02:31] I will go cook some packages [02:32] Thanks. [02:32] btw, don't announce xi 2.1 support if it does go in [02:32] alright, gonna go play with some blocks and crayons. I'll check back in later. [02:32] We should stop getting you to jetset around. [02:32] heh, talk to marketing :) [02:33] and I found out after the fact that qatar airlines isn't actually in star alliance, so I don't get my 16,000 status miles either... [02:33] Or maybe hire a clone of you. Or, maybe, even Rydberg. [02:34] cnd: I was discouraged to find that I'm the person with the most miles logged in the Canonical tripit group :) [02:34] heh, no comment on that, other than that I wish we still had him around [02:34] RAOF, you're in australia! [02:34] every time you travel for canonical it's gotta be 20,000+ miles [02:35] Not quite, but close :) [02:35] RAOF, how long are your longest flights in hrs? [02:35] 16, I think. [02:36] from washington, dc to doha qatar was 13:30 for me, the longest I've ever been on [02:36] eek [02:36] That's longest *continual* flight. Sydney→Heathrow is like 22 [02:36] But with a stop in Singapore. [02:37] yeah [02:38] I suspect I'll beat that if we go to Dallas again. Qantas now flies Sydney→Dallas direct. [02:40] cool [02:41] Yeah. That'll be much more fun than SYD→LAX→DFW [02:55] RAOF, bryceh: I realized that the new grail depends on a new package called utouch-frame [02:55] it would require an upload, a review, a MIR request and review, etc [02:55] hmmm [02:55] …which isn't all going to get done before A2 freeze. [02:56] Is that a new source package, or new binary package? [02:56] I might need to branch off utouch-grail 1.0.16 [02:56] which is in maverick [02:56] and I guess currently in natty [02:57] Right. [02:57] Is utouch-frame a new source package, or a new binary package of the existing utouch sources? [02:58] If the latter, that'd only require an upload + archive admin prod. [03:00] new source package [03:00] I'm not sure I like the proliferation of utouch source packages [03:00] Urgh. Ok, that's not going to fly. [03:00] but I was apparently the only one on the team who felt so [03:02] as I look more closely at all this, I don't think the xi 2.1 + gesture work will be well tested enough [03:03] due to requiring some last minute changes [03:03] if we go with just gesture support [03:03] then I just need to add the gesture patches into xorg-server and evdev [03:03] I would like to update evdev to master though [03:04] How risky is that? [03:04] which part? [03:04] evdev → master. [03:04] I'll look at the git log. [03:04] let me check what's all in there [03:05] Hm. Looks to me like master is 2.6.0. [03:06] Lucky you! That's what we've got in Natty. [03:06] hmm... I guess peter hasn't pushed my patches yet [03:06] then I would like to add my patches in [03:06] they're the foundation of xi 2.1, and have been working well for quite some time now [03:07] On lots of MT and non-MT hardware? [03:07] RAOF, take a look at whot's evdev repo [03:07] RAOF, yes, both [03:07] at least, on trackpads and touchscreens [03:07] and if it breaks others I'd rather know sooner than later [03:08] Wow. That's a lot of deletions. [03:08] heh [03:08] actually, what is the server at for input abi [03:09] Yeah, Mr Remove Support for ABI < 12.2 [03:09] it's at 12.2 in 1.10 right now I think [03:09] I just want to be sure [03:09] anyways, I'll double check it [03:09] the patches add in masked valuator support [03:10] in xi 2.1 you can give incremental updates on valuators [03:10] so if the pressure hasn't changed, then that valuator axis isn't sent in the device event [03:10] That makes sense. [03:10] it's mostly a bandwidth save in the x protocol [03:10] not huge for ST devices [03:11] but pretty good for MT devices that support tens or more pointer [03:11] points [03:11] Woah! I just managed to crash zsh. [03:11] heh [03:11] did it give you a quip out of a dungeon crawler game? [03:12] I can't remember what utility does that, maybe screen? [03:12] No, it just SEGVd [03:12] Screen, in nethack mode? [03:12] I'm unaware of a nethack mode [03:12] but if screen crashes it gives a nethack-like message [03:12] Hm. 1.10RC1 has input ABI 12.0, master as of not a long time ago has 12.1, and I'm fetching a fresh master... [03:13] RAOF, well, if we're using 1.10RC1, then I'll just leave the patches alone for now [03:13] We're not using 1.10 [03:13] it would be nice to have, but I don't want to deal with breakage [03:13] RC1 [03:13] what are we using? [03:13] some snapshot of master? [03:13] Yeah. [03:13] Master as current as possible. [03:14] ok [03:14] Because there are ABI breaks between 1.10 and master, and I don't particularly want to rebuild the world more times than necessary. [03:14] yeah [03:14] in that case, it may be easier for me to give you an updated gesture patch [03:14] unless you've already forward ported it [03:14] I have not. [03:14] that one forward ported with ease [03:15] just conflicts in configure.ac that wiggle handled [03:15] hmm... master is still at 12.1 [03:15] Ah, yes. [03:15] I think peter has been meaning to bump it to 12.2, and the stuff in evdev is in xserver master [03:16] but it's just not worth fiddling with that much [03:16] if I can't git format-patch and shove it in debian/patches, it's too much work :) [03:16] Looking at the patch, it didn't seem to actually check for 12.2, just use things unconditionally. [03:17] Which means that if you're right, and master has what evdev needs, it'll just build? [03:17] hmmm, let me double check [03:18] RAOF, no, check evdev.h [03:18] Yeah, just saw it. [03:18] Although if you don't feel like fixing *that* up, I'd suggest you're feeling quite lazy :) [03:18] heh [03:18] well, I don't want to be patching things in and out as numbers change upstream [03:19] and it's 10:18 pm here [03:19] RAOF, so here's what I suggest: [03:19] I can forward port the gesture patch for xorg-server and xserver-xorg-input-evdev tonight [03:20] are the debian git repos up to date? [03:20] if so, I can push the updated patches in there once I've got them working [03:20] Evdev is up to date, xserver is a couple of commits behind. If you push to xserver I can happily just merge those changes in, though. [03:21] ok [03:21] Oh, no. It's not based on RC1+git in git. I'll push that for you first. [03:27] gah, apt-get is useless if you have broken deps [03:27] anyone know how to tell it to stfu and do what I say? [03:29] No, sadly. [03:29] I either apt-get -f install to resolve things, or just apt-get download + dpkg --install [03:29] RAOF, so what's going on with vmmouse and wacom? [03:30] they still seem to have broken depends in xorg-edgers [03:30] Oh, really? [03:30] at least on my computer [03:30] which just as xorg-edgers and I just ran apt-get update [03:30] * RAOF checks. [03:31] vmmouse depends on xorg-input-abi-12.1 [03:31] wacom depends on xorg-input-abi-11.0 [03:33] Aaah, yeah. I see what's happened there. [03:33] wacom in edgers has been superceded by the archive version, which unsurprisingly isn't build against ABI 12.1 [03:33] yeah, I just noticed that too :) [03:34] ok, I can get around this [03:34] Remove wacom? [03:34] And, looking at that list, I suspect synaptics, too? [03:34] no, not good enough [03:35] because xserver-xorg-input-all [03:35] yada yada [03:35] Just remove -input-all [03:35] but that's a dependency of something else [03:36] hmm... maybe not [03:36] maybe things just got in such an inconsistent state that I had to install it for some reason [03:37] anyways, my machine is back to what it should be again [04:44] RAOF, looks like there's another abi breakage [04:45] I can't build the xorg-server package that exists in git.debian.org [04:45] cnd: In master right now, or the upcoming DGA thingy? [04:45] in the ubuntu branch [04:45] I'm hitting a build error in hw/vfb/InitOutput.c [04:46] That's odd. It's *just* finished building right now locally. [04:46] struct _Screen has no member named index [04:46] RAOF, from the ubuntu branch? [04:46] Yeah. [04:46] hmm... [04:46] I, um, think so? [04:46] RAOF, what's your top commit hash? [04:46] fbfe7a1ec506 [04:46] that's not what I have... [04:47] I have ef7a6ac... [04:47] Set UNRELEASED; this isn't yet releaseable [04:47] You may have pulled before I pushed the update to the master snapshot? [04:47] Ah, yeah. That's it. [04:47] oh, ok [04:47] Hm. To make it easy I should import the orig into the pristine-tar branch. [04:49] There you go. Now with pristine-tar goodness. [04:51] ok, hopefully I will finally have a tree that will build [04:57] RAOF, I'm still hitting that doxygen bug [04:57] how are you able to build it? [04:57] If you use git-buildpackage it should pull the .orig.tar.gz out of the pristine-tar branch? [04:58] so I need to git fetch origin [04:58] to get the new branch [04:58] Yeah. [04:58] then do I need any options for git-buildpackage? [04:59] I've never used it before [04:59] It should just work. [04:59] Although you might want to pass --git-ignore-new to it, otherwise it gets narky if you're not in an absolutely pristine git tree. [05:02] How's that going? [05:03] well, it's still building [05:03] but the doxygen isn't till the end [05:03] Ah. You're not building in a chroot? [05:04] That *might* be why it built for me, then. [05:04] That'll be annoying if it is the case. [05:04] no, I don't usually bother [05:04] and it might explain it [05:05] I have a tmpfs chroot, which makes it nice and fast to sbuild, so that's what I tend to do. [05:05] well, I can work around it for tonight [05:05] so I can test and push [05:05] Yeah. [05:18] * cnd is turning off his alarm tomorrow [05:18] Sorry for pulling you into a firedrill :( [05:18] heh, this is basically payback [05:19] remember when we pulled your out of your sleep for maverick feature freeze? [05:19] Yeah. [05:19] really sorry bout that one [05:20] :) [05:20] got some wrong info and it caused a big mess [05:20] That's ok. [05:20] It didn't end up being *that* late for me. [05:20] I can pull through 2am once a cycle :) [05:21] heh [05:21] so, ubuntu doesn't use udebs anymore [05:21] what would you think about patching out building them [05:21] ? [05:21] We don't use udebs? When? [05:21] Or do you mean -dbg packages. [05:22] our installer isn't based on debian-installer [05:22] hasn't been for quite some time [05:22] Even our alternate-cd? [05:22] according to cjwatson [05:22] hmm, don't know, but he said we didn't actually need utouch udebs [05:22] and without those, you'd get no X evdev either [05:23] Maybe we don't use *X* in our alternate installer, and so don't need *X* udebs? [05:23] That would make sense. [05:23]