=== jono is now known as Guest99797 [05:51] ...that point where you realise there's *more* umockdev work to be done :/ [05:55] RAOF: Sounds less than awesome [05:55] At least from your perspective, having worked on it for so long already [05:57] I think what it actually means is that I'll do something else for a moment. [07:31] Hey Mir's surface class hierarchy appears to be simpler and nicer than last time I mapped it out. Nice [07:41] Eek. I just managed to delete 620 lines of code, and everything still works :) [08:04] duflu: as long as they are not 620 lines of tests... ;) [08:04] alf__: No, mostly interface definitions not used within Mir itself. Seems unity-mir uses it still so can't quite propose removing it yet [08:06] But it's fun to see a large diff of red that doesn't break anything locally :) [09:48] duflu: are we intentionally delaying https://code.launchpad.net/~vanvugt/mir/server-abi-16/+merge/206842? or shall we top-approve? [09:49] alan_g: Nothing has landed since 0.1.5 requiring the bump yet. Still all under review [09:49] ok [09:49] It would be nice to have a point release without ABI bumps [09:50] What about https://code.launchpad.net/~andreas-pokorny/mir/report-namespace/+merge/205577 [09:52] alan_g: Yeah that would probably qualify ( bzr status -r1413..1414) === dandrader is now known as dandrader|afk [09:57] alan_g: Given we will almost certainly have to change unity-mir multiple times for various SDK cleanups, it's probably a good idea to spread the changes out, without dumping too many of them in a single point release. [09:57] Although that's not necessarily something we can completely plan [09:59] It seems every time I find something that makes no sense at all in the Mir source, unity-mir is using it. It's just a matter of whether we can design a better way each time :P [09:59] duflu: Yep [10:03] * duflu goes to find dinner, with a newfound aversion to spaghetti ;) === dandrader|afk is now known as dandrader [10:35] alf__: can this really be true? https://code.launchpad.net/~vanvugt/mir/fix-1281938/+merge/207123/comments/486466 [10:37] alan_g: no, replying [10:38] thanks [10:40] BRB (reboot) [10:52] alf__: are we happy with https://code.launchpad.net/~afrantzis/mir/valgrind-suppressions-armhf/+merge/207188? [10:53] alan_g: I am waiting fo fginther to schedule the pre-merge tests [12:17] anpok_: alf__ - [not urgent] any preferences on the way forward? https://code.launchpad.net/~alan-griffiths/mir/make-ConfigurationOptions-a-dependency-of-DefaultServerConfiguration/+merge/206996 [12:40] hm hm [13:02] anpok_: alan_g alf__ so who wants to have some mwc fun ? :) [13:02] ok, so seems like we might have a freeze/crash caused by the snapshotting interaction on the phone [13:03] phone is special due to the new right edge navigation [13:03] https://bugs.launchpad.net/ubuntu/+source/unity8/+bug/1281728 [13:03] kgunn: what? (can it wait until after lunch?) [13:03] Launchpad bug 1281728 in unity8 (Ubuntu) "Unity8 random freeze on demo image" [Critical,In progress] [13:03] alan_g: yes...eating still allowed :) === alan_g is now known as alan_g|lunch [13:03] looking at it, i wonder if alf__ might be the right person ? [13:05] hey, I'm able to reproduce https://bugs.launchpad.net/ubuntu/+source/unity8/+bug/1281728 [13:05] Launchpad bug 1281728 in unity8 (Ubuntu) "Unity8 random freeze on demo image" [Critical,In progress] [13:05] mzanetti: i think alf__ might be the right person...but he hasn't responded yet....but seems all the repro instructions are there [13:05] I believe its the screenshotting functionality in mir [13:05] mzanetti: yep...they got prepped [13:05] in case a screenshot is requested shortly after an app has crashed [13:05] stepping away for a.m. duties.... [13:06] who can help me with this? [13:06] mzanetti: he might be at lunch...he'll come back soon...as alan_g|lunch just left [13:06] kgunn: I am investigating https://bugs.launchpad.net/mir/+bug/1282248 , but I can switch if it's more urgent [13:06] Launchpad bug 1282248 in Mir "Unity8 crashes when using latest mir/android backend with system-compositor" [High,In progress] [13:06] ah [13:06] alf__: yeah....this one is a race on time unfortunately [13:06] thank you [13:08] alf__: steps to reproduce are in the bug report, 3rd last comment [13:09] mzanetti: which image should I get? [13:09] alf__: in theory it should happen with trunk too. however, we've been using the MWC demo image. its easiest to reproduce there [13:09] alf__: unitymirteam ML has instructions how to flash it [13:10] comes down to: [13:10] sudo add-apt-repository ppa:cwayne18/demo [13:10] sudo apt-get update && sudo apt-get install mwcinstall [13:10] mwc-flash [13:10] alf__: ^ [13:10] mzanetti: ok, nexus 4 or 10? [13:10] 4 [13:11] mzanetti: ok, starting process I will let you know if I need anything... don't leave town :) [13:11] heh, ok === dandrader is now known as dandrader|afk [13:18] mzanetti: so, the problem is that unity-mir is requesting a screenshot for a crashed app, and mir can't handle that? [13:18] alf__: yes, afaics [13:19] alf__: there is a time gap in between the app crashes and unity-mir gets informed about that [13:19] alf__: and if unity-mir tries to update the screenshot in that timeframe it'll crash [13:20] alf__: given that this should behave better for MWC I'd probably be ok if Mir just doesn't crash and returns an empty screenshot or whatever [13:20] the proper solution would probably be to handle crashes faster [13:20] mzanetti: ok [13:20] not sure what's involved there tho [13:23] mzanetti: do you know if this is pure mir 0.1.5 or whether it's mir/devel or ...? [13:23] alf__: hmm... afaik it's whatever is in the regular image + this ppa: https://code.launchpad.net/~unity-team/+archive/phone-right-edge/+packages [13:24] mzanetti: ok, thanks [13:27] anyone else using nexus7 find if shell restarts, input breaks? [13:28] greyback: I think I've seen this on the Nex4 today [13:28] greyback: does it recover if you press the power button twice? [13:29] mzanetti: will try that next time [13:30] mzanetti: ah that helped [13:30] greyback: so yeah. seen this quite often already today with the MWC image [13:30] on the Nexus 4 [13:30] mzanetti: well shell just has to never fall over and we'll be fine :) [13:31] heh, ack [13:31] greyback: there are more reasons to that :D. While hunting another issue I found that there are cases where unity8 restarts, but leaves a previously running app around, eating 100% cpu [13:32] have seen it a couple of times, but not way to intentionally reproduce it yet [13:32] mzanetti: yeah that I see often too. When it looses the Mir server, the client sometimes does that [13:38] alf__: leaving town for ~1h, ok? [13:38] anything you need from me before? [13:40] mzanetti: Do the instructions to reproduce this still apply? [13:40] mzanetti: I read that you have a partial fix for this? [13:40] alf__: yes, they still apply. especially if you didn't upgrade after flashing you don't even have the partial fix ye [13:40] t [13:41] alf__: if you did upgrade, just make sure to be fast enough after the app crashes. [13:41] but its about 2 secs, so plenty of time [13:41] mzanetti: ok === dandrader|afk is now known as dandrader [14:12] mzanetti: kgunn: I can't get a crash! I open Calculator, swipe it away => back to apps screen, open Media Player, click on OK button. Then I see something both a "blank" app (white) and the Calculator app trying to get to the screen. I have tried swiping to the right both while this is happening and after things have settled, and everything works ok, I am back to apps screen. [14:14] alf_, I attached test results to your MP: https://code.launchpad.net/~afrantzis/mir/valgrind-suppressions-armhf/+merge/207188 [14:15] alf_, the armhf build passed, the amd64 did not [14:15] alf_: re [14:15] fginther`: great thanks [14:15] alf_: hmm... strange. crashes reliable here [14:15] alf_: did you use the mwc-flash script to set it up? [14:15] alf_: so just to be sure...you're using the mwc-flash image ? [14:16] mzanetti: kgunn: yes, perhaps because I got the image so recently, something has changed? [14:16] mzanetti: video maybe ? [14:16] of your methodology [14:16] ack [14:17] * kgunn is updating his n4 [14:17] alf_: its possible...or we just buy you and your n4 a ticket to barcelona :) [14:18] mzanetti: kgunn: unity8 7.84+14.04.20140218-1rightedge647~659~ubuntu14.04.1, libunity-mir1 0.2+14.04.20140214-1rightedge176~180~ubuntu14.04.1 [14:18] * mzanetti compares === alan_g|lunch is now known as alan_g [14:18] alf_: yep, that's it [14:20] kgunn: I take it alf_ has taken the "fun"? [14:20] alan_g: yes, but I cannot reproduce the issue... [14:20] alan_g: yes :) of course there's always fun to go around if you want some [14:21] ...uh, i think i need to reboot...somethings wonky...brb [14:26] alf_: http://notyetthere.org/data/14020002.mp4 [14:26] mzanetti: ok, I think I can get it now (I rebooted) [14:26] mzanetti: thanks [14:29] alf_: not always, but most of the time you need to reboot after reproducing because the media player won't start at all any more [14:36] mzanetti: an interesting observation... removing the apport package "fixes" the problem for me, try it and in the worst case we can use this as a workaround [14:37] alf_: now that's quite weird... [14:37] * mzanetti tries [14:40] alf_: indeed... confirming you observation [14:40] kgunn: managed to reproduce (needed a reboot for some reason), also found that removing the apport package """fixes""" this [14:40] so I guess apport tracing the media player delays some threads in unity/mir to increase chances of this happening [14:41] alf_: ack...weird.. [14:41] what made you try that ? [14:41] hunch ? [14:42] kgunn: no, just I saw apport delaying the phone's startup :) [14:44] is this one of those apps get the wrong session object bugs? [14:47] anpok_: hmm, I don't know, but may be related somehow [14:54] if the media player closes itself it might go through ApplicationManager::onSessionStopping - and it could then terminate the calculater of it was in "Starting" mode while the media player came up [15:09] kgunn: mzanetti: anpok_: After installing the debug packages I am getting a different problem. *Every* time I try to swipe an app away, I get: http://paste.ubuntu.com/6965862/ [15:10] greyback: dandrader ^ [15:10] kgunn: mzanetti: anpok_: perhaps that's related to what anpok_ suggested, i.e., apps get the wrong session object somehow? [15:12] there would be a fix for that one.. [15:13] but if it happens everytime you swipe away it might be that there is just no session yet [15:13] not sure [15:13] where is updateScreenshot? [15:13] thats a different branch? [15:13] alf_: this the nexus4? [15:13] anpok_: I mean every time I swipe away from an app, so there's at least that app's session [15:13] greyback: yes, with mwc image [15:14] ok, let me flash & see [15:14] greyback: plus some -dbg packages [15:14] I don't understand how -dbg packages would change anything [15:14] timing [15:14] ....or worse [15:15] * kgunn remembers finding new code path in gpu vendor (who'll remain nameless) debug drivers [15:19] kgunn: greyback: I will reflash to get a clean environment... [15:30] alan_g, can you take a look at this job and let me know if it's running the tests as you expect? http://s-jenkins.ubuntu-ci:8080/job/mir-clang-trusty-amd64-build-fjg/6/ [15:30] fginther`: sure... [15:32] greyback: just fyi...i'm gonna land the 4 mp's approved on unity-mir [15:32] https://code.launchpad.net/unity-mir/+activereviews [15:32] kgunn: thanks [15:33] fginther`: it is running the tests (which is what I asked for) but not under valgrind (which is much less important as we do that with the other builds). So I'd be happy with this [15:33] alf_: which version of unity-mir do you have when you see the "new" issue? [15:33] anpok_: ah thanks for https://code.launchpad.net/~andreas-pokorny/unity-mir/fix-1281075/+merge/207375 ! I had it done myself, but no test. [15:34] alan_g, just for the record, adding the build tests increased the build time from 16 minutes to 75 minutes [15:35] fginther`: that's weird: "Total Test time (real) = 12.33 sec" [15:37] alan_g, ugh, I looked at the wrong field [15:37] build time was under 9 minutes [15:37] alan_g, sorry for the alarm [15:37] fginther`: np [15:38] kgunn: ^^ === alan_g is now known as alan_g|tea [15:41] cool, times are coming down [15:45] mzanetti: I don't know, I reflashed and I can't reproduce anymore... [15:45] mzanetti: I will check if I may have updated to a newer version [15:46] alf_: well, you need the ppa stuff as that one changes when screenshots are requested [15:46] might well be that trunk never triggers this [15:52] greyback: after refactoring it got tempting simple to add the tests === alan_g|tea is now known as alan_g [15:53] mzanetti: just to be sure...the recipe uses unity-mir trunk & then merges your right edge changes no top [15:53] assuming that's the pkg you guys are talking about [15:54] anpok_: tests badly needed, so it's very welcome. I see your refactoring branch too, will review next week once things calm down, but again, am delighted. [16:35] Saviq: any specific question about the freeze bug ? [16:35] kgunn, no, just saw that John posted another .crash file, was wondering if anything's happening [16:36] alf_ is looking into it...but now hitting the issue of getting a different bug/result when he builds debug packages [16:37] Saviq: i thikn the guys did determine that the crashes fire apport ...which in turn causes the timing to change such that mir freezes [16:37] point being...if people are loading crash files...those are seperate bugs i think [16:56] Saviq: does building code in debug mode and calling "make install" prevent apport from kicking in? [16:56] mzanetti, it might prevent it from crashing, not from apport kicking in... [16:57] which could explain why I didn't see the issue at all when I first tried ricmm's fix manually === yofel_ is now known as yofel [16:59] alf_: anything I can do to help you? can you reproduce again? [16:59] kgunn: mzanetti: I have a workaround (solution?). I need to leave really soon, so please take care of it. It's a small patch, I will pastebin soon. [16:59] alf_: awesome! [17:02] * kgunn suspects alf_'s pastebin will say "remove apport" :P [17:02] kgunn: lol [17:09] mzanetti: kgunn: http://paste.ubuntu.com/6966425/ , it returns an empty Snapshot if we don't have a surface for some reason. It seems that ApplicationScreenshotProvider in unity-mir can handle that, i.e., a QImage(nullptr, 0,0) is a valid (although null) image [17:10] alf_: yep, should be safe from a QImage pov. I'll verify that [17:10] exactly what I hoped for :) [17:11] kgunn: what are we doing with this? merge to trunk of put into the ppa for mwc for now? I think trunk would make sense in any case. [17:12] all: bye! [17:13] byu alf [17:13] -u+e [17:14] alf_: thanks! enjoy the evening [17:14] kgunn: I need to leave too for today :/. Can you make sure this patch somehow ends up in the mir we use for MWC and first thing I'll do tomorrow morning is to verify if unity-mir can properly digest this. [17:16] mzanetti: It works for me (TM) :) [17:16] mzanetti: yeah...i'll wrestle with it [17:17] alf_: I'm thinking about generating a white surface in that case [17:17] should only take me 5x as long as you guys :) [17:17] so crashed apps hang around as white rect instead of transparent surfaces blocking mouse input [17:17] and then it'll disappear when apport is done with it [17:18] kgunn: meh... you're right... let me propose a merge for it [17:21] mzanetti: no worries... [17:21] mzanetti: i can do it... [17:21] mzanetti: i'll just create a recipe for it [17:22] mzanetti: reason being...don't quite feel comfortable promoting mir-devel right now due to some other stuff that's gone in recently [17:24] kgunn: https://code.launchpad.net/~mzanetti/mir/dont-crash-when-shooting-invalid-surface/+merge/207507 [17:24] kgunn: ah [17:24] kgunn: should I propose it against another branch? [17:25] mzanetti: propose against lp:mir/devel [17:25] ack [17:25] mzanetti: so it will get picked up and landed probably next week.... [17:25] just not today [17:26] kgunn: ok, new one: https://code.launchpad.net/~mzanetti/mir/dont-crash-when-shooting-invalid-surface/+merge/207509 [17:26] mzanetti: ta...i'll get right on the recipe [17:26] mzanetti: phone only right ? [17:27] kgunn: yeah, trunk never seems to trigger this, otherwise we'd have noticed earlier I'd say [17:27] but it shouldn't hurt (TM) [17:28] maybe it even fixes some crashes we've seen very rarely [17:29] ok. then. I'm off too [17:30] mzanetti: enjoy the evening [17:30] kgunn: meh... need to help a friend moving [17:30] carrying furniture down 4 floors [17:30] oh geeeze....please be careful!....don't end up like dandrader [17:31] heh... I'll try [17:31] alf_, any comments on https://code.launchpad.net/~afrantzis/mir/valgrind-suppressions-armhf/+merge/207188 and the modified hook script? [17:32] oh noes. I messed with the changelog.. will resubmit [17:32] fginther`: alf has gone for today [17:32] thx === alan_g is now known as alan_g|EOD === fginther` is now known as fginther === yofel_ is now known as yofel