=== duflu_ is now known as duflu [05:08] morning desktopers [05:40] Good morning [08:18] cyphermox: (wpa) "1.x branch is considered to have reached itsend-of-life. If you are still using 1.x (or any older version), you arerecommended to update to the 2.1 release" (released two weeks ago) [08:19] not that myself saw problems with 2.0 either, but I knew Debian bug claimed there were many bugs with that [08:28] I'm now rolling with the 2.1 from Anders' PPA https://launchpad.net/~anders-kaseorg/+archive/ppa/+packages [09:03] morning! [09:03] Hey Laney [09:03] good morning desktopers! [09:03] how are you? [09:03] hey seb128 [09:04] i'm good thanks [09:04] you? how was the curling? [09:04] I'm good thanks [09:04] "curling" was good :p [09:05] curling those yellow balls over the net ... [09:05] the stone was just yellow and round and we had devices to handle it :p [09:05] how was your evening? did you go climbing? [09:06] yeah but I hurt my finger a little bit so I didn't stay that long to not make it worse [09:07] so came back and played games ;-) [09:07] utch [09:07] good thinking to not insist on it though [09:07] yeah, things happen [09:37] xnox, hey, saw the most recent comment on https://code.launchpad.net/~xnox/unity/fix-packaging/+merge/207920 ? [09:38] seb128: yeah [09:38] xnox, seems like the CI error are your fault and not in trunk ;-) [09:39] seb128: yeah [09:39] can you fix? ;-) [09:39] seb128: i'm on my laptop however, with my desktop having memtest errors =/ [09:39] :-( [09:41] time to go to buy some new ram I guess? [09:42] seb128: well, it's started on monday, and on sunday i did "clean" my computer.... [09:43] the perfect is the enemy of good? [09:43] well, joke aside, good luck getting your desktop back in shape [09:43] still would be nice if you could fix that mp from your laptop meanwhile ;-) [09:45] Trevinho, hey, do you know what's the status of unity's landing? [09:46] bregma put a landing ask with 3 vcses but didn't validate it [10:01] larsu, hey, how is the nautilus hacking going? do you have any estimate how much work that's going to be to add a traditionnal menu to it? [10:01] seb128: fixed https://code.launchpad.net/~xnox/unity/no-debug-symbols/+merge/208093 [10:01] seb128: the rest is spurious changes, that I'll keep on working. [10:02] xnox, ok, thanks [10:02] seb128: since e.g. i think it's redundant to have make-check-headless, and e.g. xvfb-run should be sufficient to run all tests at all times. [10:02] (without compromising on any GL/NUX textures) [10:11] mlankhorst, hey, could you have a look at bug 1284536, seb128 suggested it might be an xorg bug. [10:11] Launchpad bug 1284536 in compiz (Ubuntu) "compiz crashed with SIGSEGV in two_way_long_needle()" [High,Confirmed] https://launchpad.net/bugs/1284536 [10:12] jibel, mlankhorst: it's segfaulting in swrast_dri.so === tkamppeter_ is now known as tkamppeter [10:13] so either it's a bug there, or unity using wrong calls/not fallbacking properly === vrruiz_ is now known as rvr [10:27] seb128: today I learned that I should reconnect to irssi after suspending :( [10:28] * larsu was wondering why it is so quiet in irc this morning [10:28] larsu, "why is the world so calm today" ;-) [10:28] haha [10:28] larsu, good morning! [10:28] seb128: hi! [10:28] how are you? [10:28] I'm good thanks! [10:28] how are you? [10:28] very good [10:28] I'm making good progress on nautilus [10:29] nice [10:29] it seems like there are no crazy actions [10:29] if I'm right about that, I'll probably have a patch before eod [10:29] did you see my question about that? [10:29] not sure if you have backlog or not [10:29] (a first version of course) [10:29] seb128: yes, I did [10:29] I wanted to tell you last night, but had to rush [10:31] excellent [10:31] I was after an estimate [10:31] like if that's 3 weeks worth of work that's not worth it for the this cycle [10:31] if you plan for eod, double for budget and say it's going to be ready this week [10:31] which is good enough for me ;-) [10:55] * xnox loves seb128 planning skillz =) [10:55] ;-) [11:06] * didrocks hugs seb128 and larsu on unity-settings-daemon [11:06] was waiting for that for so long! [11:06] * seb128 hugs didrocks ;-) [11:07] Trevinho: are you around now that I guess it's a quieter time for you for some debugging on this decorator? :) [11:07] * larsu doesn't have anything to do with that but hugs didrocks anyway [11:07] oh? [11:07] [ Lars Uebernickel ] [11:07] * media-keys: add support for allow-amplified-volume When that key is [11:07] larsu, it's your work dude :p [11:07] set, the maximum volume is PA_VOLUME_UI_MAX instead of [11:07] PA_VOLUME_NORM. This makes it consistent with the sound indicator [11:07] and system settings. [11:07] it was denouncing you! [11:07] :) [11:07] didrocks: oh, I thought you meant u-s-d itself [11:07] ya I did that one. Hug accepted and appreciated! [11:07] ;-) [11:07] ah, it's in u-s-d, I was just speaking about the last upload :) [11:07] ahah [11:08] seb128: so, just to be clear, you told me you didn't get that bug [11:08] which is: [11:08] 1. ensure you have a terminal window not maximized [11:08] click on a menu item in the decorator [11:08] then, press left [11:08] (well, 2. and 3. ;)) [11:08] larsu: \o/ finally full-max-sound instead of cripple one =) [11:09] -> for me, the menu opened is then the global menu [11:09] didrocks, it cycles in the local menu for me [11:09] I really have a weird version of the decoration :p [11:09] xnox: testing appreciated :) [11:09] didrocks, does it work in a guest session? ;-) [11:09] seems I got the low-cost one :p [11:09] let me see [11:10] larsu: as soon as my desktop bisects which of the 4 RAM modules is busted. [11:10] =) [11:10] with memtest [11:10] haha, have fun! [11:12] and of course, it's working in a guest session [11:12] I don't even have global menu shown at all [11:12] when the app isn't maximized [11:12] (and local menu was enabled by default on my session) [11:12] not sure what happens, let me see u-c-c [11:12] so… [11:12] I have lim [11:12] where the option in appearance is set to global menu [11:14] and nothing in dconf… [11:14] ok, now that I opened u-c-c, no more global menu (I think it set to global) [11:14] and after selecting "in windows titlebar", I don't have that bug anymore [11:15] do you still have the wrong buttons? [11:16] seems like your config was in a weird state [11:16] yeah, I have the wrong buttons though [11:16] it's weird, I removed the compiz config though… [11:17] dconf config or .compiz directory? [11:17] both… [11:17] brb, restarting with unity-lock-screen [11:35] seb128: probably a llvmpipe bug then [11:36] mlankhorst, can you help getting it fixed, or at least reported to the right bug tracker? [11:37] some way to reproduce would be nice [11:38] drisw_update_tex_buffer (drawable=, ctx=, res=) at ../../../../../../../src/gallium/state_trackers/dri/sw/drisw.c:320 [11:38] look like garbage [11:38] does https://bugs.launchpad.net/ubuntu/+source/unity/+bug/1055166 looks similar to you? [11:38] Launchpad bug 1055166 in unity (Ubuntu) "compiz crashed with SIGSEGV in memmove() from drisw_update_tex_buffer() from dri_set_tex_buffer2() from operator() from compiz::opengl::bindTexImageGLX() from ... from unity::UnityWindow::DrawWindowDecoration" [High,Triaged] [11:38] yeah [11:38] same backtrace [11:39] duflu wrote "This one is coming from Unity trying to hand-draw decorations." [11:39] Trevinho, ^ is that an issue you are aware of/working on? [11:39] some way to reproduce would be nice, even if it means running things in valgrind [11:39] mlankhorst, seems like the way to reproduce is "try to run unity from trusty in software rendering mode" [11:40] seb128: yeah but LIBGL_ALWAYS_SOFTWARE=1 compiz is probably not enough [11:40] jibel, ^ do you have a way to reproduce the issue? [11:41] that should be the way to force sw rendering at least [11:46] ok, found it [11:46] it's a form-factor=desktop [11:46] which triggered this [11:46] Trevinho: ^ [11:48] where is that form-factor setting? [11:48] what are the other values? [11:48] it was an old key, not sure where, I'm trying to get back to a working setup first :) [11:49] (it was in unity code for sure) [11:49] * didrocks lost ctrl + alt + T, even if it's set as shortcut [11:49] (and yes, it's working in a guest session) [12:00] didrocks, it's nice to know that the unity team has all those "if username === didrocks" codepaths, make you feel like special ;-) [12:01] yeah, it's so painful :p [12:01] I clearly don't understand what is eating ctrl + alt + T [12:01] if I reassign to ctrl + alt + u, no issue [12:02] and nothing in dconf dump / [12:02] seb128: btw, it's not the form factor that failed before [12:02] seb128: it was before of the decor plugin was loaded [12:02] becauseù [12:02] because* [12:03] didrocks, ctrl-alt-t is handled by compiz itself [12:03] so the migration didn't work [12:03] it's in the special keybindings there [12:03] yeah, in system settings [12:03] right? [12:03] control center [12:03] no [12:03] ctrl-alt-t is handled in a special way in compiz [12:03] not through the normal keybindings handling [12:04] it was before handled as a normal keybinding with the gnome compat [12:04] so, where is it supposed to be defined now? [12:04] that UI works for me to configure it [12:05] same here, if I set to anything else than Ctrl + alt + t :/ [12:06] didrocks, ctrl-alt-t is set specifically in "compat with GNOME "in cssm [12:06] try to unset it from there? [12:07] seb128: well, it disabled it in u-c-c at the same time [12:07] (as it's the same key) [12:08] and yeah, Ctrl + alt + T doesn't work [12:08] setting ctrl + alt + U works :/ [12:09] weird :/ [12:09] I tried creating a custom command as well… [12:10] but back to the bug, how do I reproduce it? [12:11] I wonder if compiz/u-s-d fight for the key grabiing [12:11] seb128: yeah, that's the most plausible one [12:11] mlankhorst, the compiz/software rendering one? [12:11] yeah [12:11] mlankhorst, did you try to run compiz under LIBGL_ALWAYS_SOFTWARE=1? [12:12] mlankhorst, the reports seem to be along the line of "boot trusty daily iso in a virtualbox without 3d support" [12:13] ah misunderstood it :P [12:13] seb128: hum [12:13] if I restart compiz [12:13] it works [12:13] let me try to logout/back [12:13] * mlankhorst adds a bunch of dbgsyms [12:14] * seb128 shakes fist at upstart and ssh-agent [12:14] which is that thing so unreliable? [12:15] we should move to systemd [12:15] ok, working now [12:15] so yeah, probably a race [12:15] still worried that the decor plugin wasn't removed on upgrade though [12:17] it was for me [12:17] it's weird because I even tried to readd it for testing and it seemed like unity would force unload it if it was loaded [12:27] bah, "Not touching package due to block request by freeze (contact #ubuntu-release if update is needed) " [12:27] beta week! [12:28] Laney, how comes it affects ubuntu-ui-toolkit? [12:28] I saw we blocked only seeds [12:28] seeded-in-ubuntu ubuntu-ui-toolkit [12:28] webbrowser-app is on the image now remember [12:29] oh, right [12:29] our turn to get back to the touch guys by blocking their landing due to desktop freezes? :p === jhernand1z is now known as jhernandez [12:29] didrocks, no UI toolkit fixes for you! :p [12:30] seb128: I don't really care/wanted that change anyway :p === MacSlow is now known as MacSlow|lunch [12:30] didrocks, no pulseaudio fix for you! (better? :p) [12:31] seb128: yeah, that one will maybe be discussed a little bit more :p [12:31] * seb128 though we had shorter freezes nowadays [12:31] this is shorter [12:32] Laney, is it worth asking for u-c-c u-s-d to be moved in, or should I just wait for the end of week? [12:32] * seb128 hates beta freezes [12:33] it'll be thursday [12:33] * seb128 uses the "merge back in trunk even if not in destination" flag from didrocks [12:33] depends how urgently you think they should go in [12:34] I'm annoyed by things not being merged back in trunk more than by them not going to release [12:34] but CI train link both [12:34] and you can thank me for implementing overriding when used responsibly :) [12:34] didrocks, indeed, thanks! ;-) [12:35] I'm just going to do that [12:35] heh [12:35] I've already been abusing it while u-c-c was blocked in proposed some days ago [12:35] as long as you care for it moving it and know the version isn't busted, that's fine :) [12:36] yeah, that's fine [12:36] * seb128 looks at excuses and sees all those nices bugfixes being hostage of beta freeze [12:36] oh well, it's only a few days [12:41] mlankhorst, bug 1284588 is the same compiz crash with symbols for libgl1-mesa-dri, but retracing failed. so not very useful [12:41] Launchpad bug 1284588 in compiz (Ubuntu) "compiz crashed with SIGSEGV in _wordcopy_fwd_aligned()" [Undecided,New] https://launchpad.net/bugs/1284588 [12:43] dup it :p [13:00] hm looks like a drisw bug http://paste.debian.net/83976/ === alan_g is now known as alan_g|lunch [13:07] mlankhorst, did you manage to reproduce then? [13:18] seb128: did you win? :) [13:19] desrt, oh, yesterday was training, no win or loosing ;-) [13:19] desrt, good morning! [13:19] everybody wins! [13:19] good morning :) [13:19] indeed [13:23] desrt: morning, how's it going? [13:23] good [13:23] my gmain stuff is already starting to land, a cycle early :) [13:24] still a lot of work there, though... turns out the relationship between GMainContext and GSource is slightly complicated. go figure. [13:55] hikiko: hey... i was looking at that add_dict_entry() code again and i actually found a very serious problem. i'd recommend replacing it with that new code ASAP. i don't think that it's doing what you think it's doing, at all [13:55] hikiko: in particular, you will probably notice that you can get duplicated dictionary entries... [13:56] char str[512]; [13:56] g_variant_get_child (tmp, 0, "s", &str); [13:57] this will not put the string into that buffer... rather, it will put a pointer to a string into the buffer -- which you will then interpret as a string [13:57] the only reason this code doesn't crash outright is because the only thing you do with it is a strcmp and at some point out of 512 it is sure to contain a '\0'... === alan_g|lunch is now known as alan_g [14:00] desrt, thanks, I ll check it in a while and change it to use the builder, I didnt get any duplicated entries though [14:00] I had a debug function that was printing the dict [14:00] interesting... [14:00] and everything looked fine [14:01] I'll be back in 15 minutes (lunch) to see what's wrong and give you the code that prints the entries [14:01] to check [14:01] if you like [14:01] hikiko: would be interesting if you added a printf() for 'str' each time around that loop, just before the strcmp -- i think you'd be surprised at what's in there [14:02] yes I had one let me show you [14:02] I just deleted them in the final branch [14:02] ohhhh... wait. [14:02] i take it back [14:02] this is EXTREMELY confusing: [14:02] char str[512]; [14:02] later... [14:02] Mirv: I have 2.0 in my own PPA, been testing it for a while [14:02] { [14:02] int val = 0; char *str; [14:03] i misunderstood it because of that [14:03] desrt, could you have a look to https://bugs.launchpad.net/ubuntu/+source/glib2.0/+bug/1284647? [14:03] Launchpad bug 1284647 in glib2.0 (Ubuntu) "gsettings crashed with SIGSEGV in g_settings_backend_dispatch_signal()" [Medium,New] [14:04] yup [14:04] thanks [14:04] cyphermox: yes I know, I got it from there during the spring and now updated to 2.1 which continues to work fine (+ probably better) [14:04] seb128: i've seen this bug from time to time... [14:04] desrt, likely https://bugzilla.gnome.org/show_bug.cgi?id=710367 ? [14:04] Gnome bug 710367 in gsettings backend "Crash in g_settings_backend_dispatch_signal()" [Major,Unconfirmed] [14:04] but always somewhat rare and difficult to trace [14:05] seb128: i'm pretty sure it's an invalid use of gsettings, but it's a use that i plan to make valid... [14:05] since it's common and (honestly) not that much to expect that it would already work this way [14:05] desrt, the launchpad report is from the "gsettings" command line... [14:05] that's new. [14:05] do you have invalid gsettings use in there? ;-) [14:05] seb128: it's possible :) [14:05] * desrt wants to see this stack trace :) [14:06] seb128: now you have my attention :) [14:06] desrt, it's on the bug [14:06] seb128: looking [14:13] https://code.launchpad.net/~hikiko/unity-control-center/unity-control-center.per-monitor-fonts-scale-factor-slider/+merge/205227 what happened to the history? :s :/ [14:13] seb128: unity8-desktop-session is stuck in proposed =( [14:13] seb128: * arm64: unity8-desktop-session-mir [14:14] hikiko, you pushed over the wrong location yesterday, that nuked it [14:14] can't wait to try it on my convertable [14:14] oh nooooo [14:14] https://code.launchpad.net/~hikiko/unity-control-center/unity-control-center.per-monitor-fonts-scale-factor-slider/+merge/205227/comments/484957 <- I am looking for that func btw [14:14] well be back in 15 minutes to either find it or rewrite it [14:15] because I was printing everything before/after get/set [14:15] and the values were ok [14:15] desrt, ^ brb to find it [14:16] hikiko: don't worry about printing it out. i believe you now :) [14:16] seb128: hm the failling crash reported by valgrind call maps to a BadRequest for XGetGeometry [14:17] hikiko: the way you declared a variable in one way globally and then shadowed it with a variable of the same name declared inside of a block on the same line as a variable of a different type confused me. you should not do that :) [14:21] xnox, is that because it's not installable on that arch? [14:21] seb128: yeah. [14:22] seb128: which is shocking, given that it's a new package.... [14:22] xnox, should be add an hackish build-depends on unity8 to depwait on e.g arm64? [14:22] seb128: i see, it does increase uninstall coun by one =) [14:22] seb128: dep-wait is usually the best way, yes. [14:24] xnox, I don't understand briney [14:24] seb128: i hate gobject [14:24] britney [14:24] why do we care about installability of things that have never been installable? [14:24] (read: i figured out the bug) [14:25] it sucks to have to b-d on unity8 [14:25] desrt, found the issue? [14:25] "good" (is it easy to fix/workaround?) [14:25] no [14:26] what's the issue? [14:26] seb128: oh, it cares about total number only. So e.g. it will hapily trade: unity uninstallable, for unity8-desktop-session, such that total number of uninstallable packages on the architecture stays same or goes down. [14:26] basically, if a signal arrives just as someone drops the last ref on a gsettings (irrespective of if they have a signal connected or not) then a race occurs -- and if the signal handler grabs the mutex first, it will increase the reference count on the object that is in the middle of destruction.... but in the other thread, we are waiting on the mutex acquire in the remove-from-list function.... after we unlock, we have a ref on the dead object, but [14:26] seb128: this is to prevent arch-scew on first time a package is uploaded. [14:27] seb128: cause it can't know if it's uninstallable.... or it's just arch:all etc packages weren't built yet. [14:27] seb128: it has to do with how our weak reference notifies really have nothing at all to do with object lifecycle, but rather with disposing [14:27] desrt, irc cut at "but" [14:27] but it will still be removed from the list.... and then we access the list and crash. [14:28] seb128: at least this is what i gather. that's why removing packages can be dangerous, cause it will go "oh look uninstall count went up, and i can shove random crap in and uninstall count will stay the same" [14:28] there were two 'but' there. hope that was the right one :) [14:28] hum, "fun" [14:28] seb128: fun and very difficult to fix. [14:28] it was [14:28] good luck :/ [14:28] * desrt has a complicated plan that just moved a few notches up his todo list [14:29] xnox, I guess that makes sense, in practice we just got bitten often there because of qt5 missing on those arch, we need 5.2 to land [14:29] seb128: oh, are you landing it now, then?! =) [14:29] *shiny* [14:30] I wish we would, I'm not the one deciding [14:30] it's a negative cost-saving at this point, the longer we delay the less time we have to resolve things. [14:31] bregma, can you make unity8-desktop-session b-d on unity8? cf backlog for details, basically britney doesn't like the fact it builds on arm64 but can't be installed there because unity8 isn't available on it [14:31] xnox, yeah, you don't need to convince me, but the touch team has that 0 regression rule [14:33] seb128, sure, I guess, but wouldn't specifying [!arm64] be cleaner? [14:38] back [14:38] ok desrt :) [14:39] desrt, if you agree, I ll leave it like that for the moment to fix more urgent errors (I mean since this is not a real bug but a better coding style) and come back to it at a later point [14:39] hikiko: sounds good -- but there is still a leak in there :) [14:39] so don't wait too long [14:39] leak? [14:40] ya... you call malloc() [instead of g_new...] and then don't call free() [14:40] oh that's a bug then :D [14:40] I ll fix this!! [14:40] desrt: do you mind writing these all down in a comment on the MR so that it's documented? [14:40] hikiko: if you're going to do that, then just rewrite using GVariantBuilder :) [14:41] desrt, ok! I will push in a while [14:46] hikiko: i'm happy to review when you're done [14:48] does anyone know how rtkit makes sure to get the right cgroup setup for pulse in the desktop ? [14:49] we have an issue on the phone (where pulse is just fired up by an upstart session job) where rtkit dislikes to give any realtime capabilities to pulse [14:49] seems this works flawless on my trusty laptop [14:49] ok, going for a run [14:50] (i assume that the session manager configures something we are missing in the phone setup, i just cant seem to find what/where) [14:53] tedg: hey! how's it going with the indicator-upstart transition? [14:54] tedg: i was just looking at your indicator branches and i'm slightly worried by all the "OnlyShowIn:Unity" everywhere... [14:54] bregma, well, then you would have a unity8-desktop-session-mir binary that isn't working since it wouldn't bring unity8 [14:54] bregma, you could as well change the arch: any to list the archs where unity8 is available then [14:54] which is another valid option [14:56] ali1234: indeed, that seems weird --- as those xdg autostart files aren't even needed under an upstart user session [14:57] one of them actually has "OnlyShowIn:Unity;Xfce" [14:57] my poor gnome session [14:57] so i'm mostly wondering why the others don't [14:57] elementary also uses this stuff [14:58] ogra: IIRC, it has nothing to do with cgroups. rtkit listens on the system bus. pulse requests to be bumped to realtime, rtkit adjusts priority [14:58] what's their OnlyShowIn value? [14:58] anyway, leave a review comment [14:59] ogra: you need a working policykit (does that work on the phone?) [14:59] i have absolutely no idea, but i could find out i suppose [14:59] ochosi, Okay, they're blocked on the android stuff in Ubuntu Touch right now :-( [14:59] mdeslaur, and if CONFIG_RT_GROUP_SCHED is set in the kernel and unconfigured this can only be done for root [14:59] at least according to the kernel docs [14:59] ali1234, Yeah, I just didn't want to have them start everywhere if they get installed. I put XFCE on the ones I thought you guys used. [14:59] dunno, you could just remove that line [14:59] but then people who don't want it would get it [14:59] ogra: yes, rtkit runs as root [15:00] ali1234, Happy to update or change them. [15:00] tedg: well i personally use messages, sound, sync, bluetooth, and application [15:00] mdeslaur, runs as "rtkit" here [15:00] oh and datetime [15:00] basically all of them [15:01] i don't think we have decided which ones will be default - that rather depends on which ones actually work for us, which currently is none of them [15:01] ali1234, So do you think the services should start on XFCE if they're installed? [15:01] yes [15:01] and power [15:01] ogra: the daemon drops privs buts keeps CAP_SYS_NICE [15:01] certainly if indicator-plugin is on the panel [15:01] and GNOME [15:01] Okay, I can do that. You guys can override if you want to control that. [15:02] since we use upstart, we can run them only if the plugin is loaded [15:02] the xdg autostart files are irrelevent if you use upstart [15:02] Yeah, Upstart will turn off the XDG files. [15:02] modulo the OnlyShowIn I guess [15:02] right, but one of them has a check for SESSION=unity in the upstart script [15:02] Ah, yes. [15:02] -sync iirc [15:03] We should probably then just block it from the greeter instead of checking Unity. [15:03] yeah, isn't there the greeter-mode for indicators? [15:03] sounds like a good way to block [15:03] Yes, but we were trying to ensure it didn't start at all, even in greeter mode. [15:04] ok [15:04] i do know that elementary doesn't use upstart, but they were talking about adapting the way they get launched to their own session init anyway [15:05]