=== sarnold_ is now known as sarnold === duflu_ is now known as duflu === Zachary_DuBois|A is now known as Zachary_DuBois === Zachary_DuBois is now known as Zachary_DuBois|A [07:48] Good morning [07:49] hey pitti! [07:52] bonjour didrocks, ça va ? [07:53] uff, I slept looong today [07:53] on fait aller, et toi? :) [07:53] waking up at 3 one day, at 9 at the other [07:54] didrocks: otherwise, quite okay [07:54] urgh, indeed, not even time… === larsu_ is now known as larsu [08:46] moin [08:52] Trevinho: impressive stuff with hidpi btw! I'm glad it could be achieved to some extent, it's better than tell "well we'll have this Unity 8 in 16.04 LTS" [08:53] seb128, good morning [08:53] Mirv: indeed... I know it was feasible, and well... it's a reality now... unfortunately not all the apps scale as they should but the shell will be at least [08:53] https://code.launchpad.net/~hikiko/unity-control-center/unity-control-center.bug-warning-fixes/+merge/208162 that's ready [08:57] Trevinho: yep, better something than nothing. I guess it becomes more hacky the further one tries to solve it though, like installing an extension to Firefox and making it listen to the scale factor :) [08:58] anyhow, nice [09:00] good morning desktopers [09:00] hey hikiko, Trevinho, Mirv, Sweetshark [09:01] hahaha we were all waiting for you seb128 !!! [09:01] lol [09:01] hikiko, thanks for the branches, adding that to my list for the day [09:02] seb128, I will submit another with the builder but since it's not urgent (it's just a coding style fix) I will do it a bit later [09:02] seb128: ;) -- do you have the libreoffice-dicts still on the map? [09:02] morning [09:03] Sweetshark, yeah, sorry yesterday was busy (and we are beta frozen), going to do that today [09:03] Laney, hey, how are you? [09:04] hikiko, ok [09:04] seb128: np [09:05] hey seb128 [09:06] seb128: good thank you! [09:06] you? [09:07] I'm good, thanks [09:10] you got a hidpi laptop right? [09:10] no [09:10] oh ok [09:10] those were too expensive, we got a bunch of cheap touch ones [09:11] why? [09:11] https://bugs.launchpad.net/unity/+bug/1282804/comments/2 [09:11] Launchpad bug 1282804 in Unity "[FFe] Move DPI settings over to use u-c-c settings." [Medium,In progress] [09:11] Even GTK-based applications become pretty much unusable if it is set to anything but 1 [09:11] seems curious to me [09:11] you can "fake" hidpi with xrandr --output -- scale x [09:12] hum [09:12] let's talk with bregma when he gets online [09:13] or maybe hikiko or Trevinho knows what he means there [09:13] I just acked that ffe anyway [09:13] can do other work there later on if needed [09:14] right [09:14] thanks for that [09:14] oh yes [09:14] seb128, and Laney [09:14] I had a string gsetting in unity [09:14] before the requirements changed [09:14] in unityshell gschema [09:14] Laney: that setting for now only controls unity, but I plan to add scale gtk apps also... based on our settings [09:15] and then I replaced that with the GVariant dict we had in control center [09:15] and didnt remove the old one [09:15] it's fixed now [09:15] Trevinho: that comment basically says that the gtk scaling is bad though? [09:15] hikiko, I think the topic discussed there is different, it's basically making the slider writes the GTK setting as well [09:15] oh sorry new bug [09:16] Laney, I think Trevinho wants to "teach GTK about their non-int-based scaled" [09:16] -d [09:16] Trevinho, still set on trying to do that? ;-) [09:16] good luck convincing upstream [09:16] well, I don't think upstream needs "convincing", it's more a technical issue there [09:17] Laney: well, it's not bad, but to get best results we need to mix the scaling value with text scaling... [09:17] the int-based scale is suboptimal but there are probably non-trivial technical reasons to it [09:18] yeah, there are, but still we can get a scaling by using both ui scale and text scale... So basically we use our integer scaling part for the GTK scaling, and then we set the text-scaling so that multiplied for the integer scaling equals our float scaling.. [09:19] http://lwn.net/Articles/562287/ is what I read [09:19] well, what I was thinking was to just use a float scale (slider) and when the value is selected, just save 1 float gsetting for unity 7 as it was before and 1 int (rounded and x8) for unity8 [09:19] seems a reasonable approach to me [09:19] and each app can use the more convenient [09:23] Laney, btw unity with hidpi on is in https://launchpad.net/~ci-train-ppa-service/+archive/landing-012/ if you want to test [09:23] * seb128 going to test that in a bit [09:23] can do, to check everything stays the same [09:26] Trevinho, btw did you see https://bugs.launchpad.net/ubuntu/+source/unity/+bug/1055166/comments/6 ? [09:26] 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] [09:27] Trevinho, seems like the unity decorations make llvmpipe unhappy, that has a pointer where a similar bug that was fixed in compiz decorations before [09:27] seb128: mhhm [09:28] Trevinho, https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/1284536 is the same issue reported by jibel, seems to be hit often by the QA tests [09:28] Launchpad bug 1284536 in compiz (Ubuntu) "compiz crashed with SIGSEGV in two_way_long_needle()" [High,Confirmed] [09:30] Trevinho, mlankhorst started having a look to drisw yesterday and said there is a bug there, but the code his an error when doing a XGetGeometry call or something like [09:31] mlankhorst, still looking at that issue btw? [09:31] seb128: mhmh... I see... I could try to use the same way was used before... [09:31] one way or another we need to fix it [09:31] if mlankhorst can fix the bug in drisw, great [09:32] otherwise doing the same workaround than compiz had would be nice [09:32] seb128: yeah but I got a maxwell card now, playing with it a little on nouveau :P [09:32] mlankhorst, important bug fixes first, playing then, please ;-) [09:33] hwe is an important bug ;D [09:33] hehe [09:33] well, I'm sure the QA guys would appreciate if you could fix the compiz/drisw issue [09:40] yay [09:41] with the latest fixes u-s-s updating works [09:41] nice! [09:44] just one small bug [09:54] well finally that gvariantbuilder was a really small change so I pushed the fix [09:54] hikiko, great! [09:58] pitti, hey, have you seen https://code.launchpad.net/~psusi/ubuntu/trusty/udisks2/fix-standby/+merge/206951 (just checking if you did, it's in the sponsoring queue and looks like something for you) [09:59] seb128: I did, it's sitting in my mailbox; it's just utterly big, so certainly not something which we want to carry as a patch [10:00] ah no, that's just the .pc madness, go UDD [10:00] pitti, I was going to say [10:00] it's a few liners [10:01] seb128: either way, it's on my TODO list; I'll verify and apply it upstream, and sponsor [10:01] pitti, danke! [10:23] seb128: replied on the ML and in the MP/bug [10:23] pitti, danke [10:32] ugh that error doesn't make sense.. [10:32] why would I get BadRequest on XGetGeometry [10:41] unity8-desktop-session-X starts in phone resolution, instead of full screen. [10:41] unity8-desktop-session-mir doesn't start at all for me at the moment, investigating. [10:45] happyaron, hey, what happened to the ibus-pinyin update? [10:45] happyaron, also https://bugs.launchpad.net/ubuntu/+source/ibus-anthy/+bug/1279845 [10:45] Launchpad bug 1279845 in ibus-anthy (Ubuntu) "ibus-anthy sets to jp keyboard layout forcibly." [Undecided,New] [11:20] ugh [11:20] any idea why XGetGeometry could fail with BadRequest? [11:21] no idea sorry, maybe try asking on some xorg channel? [11:22] shrug, that seems to be the real issue here [11:24] oops nm :P [11:48] xnox, if you're trying to run unity8-desktop-session-mir you'll need the patched mir and qtubuntu packages from ppa:unity8-desktop-session-team/custom until the respective maintainers decide to land the required bugfixes [11:50] unity8-desktop-session-X runs in the phone form factor because that's it's default size without a desktop shell to run under on X11 [11:55] compiz (core) - Debug: stacks are out of sync [11:55] what's this? [11:58] bregma: can you please get all of those into qt5.2 ppa ? as i'm testing qt5.2. [11:59] bregma: also for a -DESKTOP- session, on the desktop, full screen / tablet makes more sense as a default. Note that for 14.04 we will not ship both phone/touch/tablet and desktop modes together, thus the two should be able to have different defaults. [11:59] bregma: and even when it starts in phone shell by default, there is no window manager to resize it into full screen. [12:00] (and or provide a config file or something) [12:04] xnox, there is no way to tell it to resize to fullscreen without it getting the size of the screen from the desktop shell it's running on -- blame the QPA maintainers I guess, they're trying to run a phone emulator or something [12:05] bregma: what does QPA stand for again? [12:05] xnox, somthing about Qt platform abstraction [12:07] bregma: so at the moment we have 3 hacks that I know of, with respect to screen size, in autopilot there is hard-coded list of screen resolutions, or "$ fbset -s" is queried to get the resolution. [12:07] bregma: there is also mirout tool to get the screen resolution. [12:07] bregma: i presume both of these are low level API. [12:10] xnox, they could also use xrandr, since it's running on X, but there's going to be a lot of code to write to get something we're not really planning to support long-term [12:11] we only plan to support unity8 on Mir long-term === brainwash_ is now known as brainwash [12:12] bregma: right, but i don't think it would be a lot of code. Does unity8 at all supports command line args? [12:13] bregma: and how does unity8 on the phone ends up with correct resolution? (e.g. mako vs nexus 10) [12:13] seb128: heh, not sure if it's a mesa problem any more, adding a xsync(dpy, true) before the failing call fixes it.. [12:13] mlankhorst, so unity bug you think? [12:14] Trevinho said he can add the same workaround that was in the old decorator I think [12:14] Trevinho, ^ right? [12:15] seb128: I don't know where the bug is, but I don't think it's in mesa [12:16] k [12:16] thanks for work you did investigating === alan_g is now known as alan_g|afk [12:19] hmm, when I try to select a gnome-fallback guest session, I still get a unity guest session. known issue? [12:19] yes [12:20] bregma: also, it doesn't appear to be upstart managed desktop session, which is odd. [12:20] Sweetshark, though I don't find a bug report, so if you want to open one, feel free [12:21] Sweetshark, I mentioned it to robert_ancell in London, but a bug would be better for tracking [12:21] ... and curiously the patch I applied on trusty/LO 4.2 packaging doesnt work. But it when I backported it to precise/LO 3.5 it works. :/ [12:21] seb128: what target package? lightdm? [12:22] Sweetshark, yeah, https://launchpad.net/ubuntu/+source/lightdm/+filebug [12:22] bregma: if i login into normal unity session, and do "unity8 -fullscreen" it looks nice and big. [12:22] xnox, if you log in to Unity7 and run Unity8, you have a desktop shell to tell it the screensize [12:23] probably the best solution is to remove the unity8-desktop-session-x11 package, reduce confusion [12:23] bregma: right. I guess i don't understand the concept of "shell" then. [12:23] bregma: well -x11 one at least works at the moment for me, where as -mir one fails to start. [12:23] bregma: is "shell" ~= window manager? [12:26] seb128: https://bugs.launchpad.net/ubuntu/+source/lightdm/+bug/1285132 <- done [12:26] Launchpad bug 1285132 in lightdm (Ubuntu) "guest session ignores session type selection" [Undecided,New] [12:26] xnox, yeah, window manager [12:26] Sweetshark, thanks === alan_g|afk is now known as alan_g [12:30] xnox, where's the qt5.2 PPA? I can try uploading the patched qtubuntu and mir packages you need (or you could grab the code from https://code.launchpad.net/~unity8-desktop-session-team) [12:31] bregma: and in -mir session window manager / shell is mir. Hence in the -x11 something like compiz needs to be started as well. Which would be trivial to do if the whole -x11 is upstart managed (we have compiz user-session upstart job to start things) [12:32] bregma: https://lists.launchpad.net/ubuntu-phone/msg05681.html https://launchpad.net/~canonical-qt5-edgers/+archive/qt5-beta2/+packages [12:32] bregma: that also has qtubuntu patches for qt 5.2 [12:33] bregma: not sure if timo idles here or not, cause coordinating the two would be good. [12:33] xnox, Mir is the display server: when Unity8 is running on the mirserver QPA, it is the desktop shell [12:34] bregma: oh, i see. ok. [12:45] Sweetshark, hum [12:45] Sweetshark, http://ftp.rezopole.net/tdf/libreoffice/src/4.2.1/libreoffice-dictionaries-4.2.1.1.tar.xz gives me an error [12:45] Sweetshark, well, http://download.documentfoundation.org/libreoffice/src/4.2.1/libreoffice-dictionaries-4.2.1.1.tar.xz [12:45] that redirects to that url that doesn't work [12:47] seb128: hmm, using that link from the webpage seems to work. maybe just retry to get another mirror? [12:47] Sweetshark, I tried 15 times in wget and firefox [12:47] I always get the same mirror [12:47] Sweetshark, can you give an url for another mirror? ;-) [12:47] seb128: I get http://ftp.rz.tu-bs.de/pub/mirror/tdf/tdf-pub/libreoffice/src/4.2.1/libreoffice-dictionaries-4.2.1.1.tar.xz which works [12:48] that indeed works [12:48] seb128: Ill forward that to upstream infra ... [12:48] thanks [12:50] done [12:50] danke [12:52] seb128, Laney, the HUD in system-settings has a quit option now ... but thats only available in the top level, if i have any of the pages open it isnt there, is that wanted (very confusing imho) [12:52] mpt, ^^^ [12:53] ogra, talk to pete, we nothing to do with the HUD, system-settings doesn't do anything there [12:53] that's coming from the toolkit or from unity8 [12:53] ah, i thought it has to implement the quit option [12:53] we have nothing* [12:53] nope [12:53] no [12:53] k === alan_g is now known as alan_g|lunch [13:25] larsu, desrt: can g_settings_get_value() return NULL? [13:25] no. never. [13:25] i was just about to mention this fact in a review :) [13:26] /* this code will never be reached because desrt made it crash */ [13:27] damn straight [13:27] * desrt has been taking the blame for 5+ years so that people don't have to check for NULL :) [13:27] trying to make sense of https://launchpadlibrarian.net/167529008/Stacktrace.txt [13:27] sources = 0x0 [13:27] that code is [13:27] sources = g_settings_get_value (settings, KEY_INPUT_SOURCES); [13:27] g_variant_builder_init (&builder, G_VARIANT_TYPE ("aa{ss}")); [13:27] g_variant_iter_init (&iter, sources); [13:28] larsu, "that bug is not possible", right? [13:28] so when you say NEVER... [13:28] it might violate its interface if you violate preconditions [13:28] in particular: [13:28] g_return_val_if_fail (G_IS_SETTINGS (settings), NULL); [13:28] g_return_val_if_fail (key != NULL, NULL); [13:28] desrt: I guess I should acknowledge that in between giving you shit for it: thanks! ;) [13:28] lol [13:29] so it's possible that 'settings' was invalid going into this [13:29] since i'm pretty sure KEY_INPUT_SOURCES is just a constant... so that's sure to have been OK [13:29] right [13:29] settings = 0x9c7c530 [13:30] we should get some warning if that's not a valie settings though? [13:30] non-NULL but maybe (a) pointing to something else; or (b) already freed [13:30] seb128: ya.. you'd see critical output about that. [13:31] shrug [13:32] pitti, did we ever update apport to collect warnings/errors in upstart log in addition to ~/.xsession-errors? [13:32] since that's where the logs are nowadays === alan_g|lunch is now known as alan_g === psivaa is now known as psivaa-lunch [13:44] ritz: I did build with 4.2 on trusty and 3.5 on precise for the drag'n drop issue. I started off with the patch that you added to the bug. [13:44] Sweetshark, hi [13:45] seb128: yes, I thought I added that in saucy already, hang on [13:45] ritz: Funny thing is: For me it doesnt work on trusty yet, but the backport to precise seems to be fine. [13:45] * ritz checking for new builds [13:45] ritz: builds are currently still here: https://launchpad.net/~bjoern-michaelsen/+archive/libreoffice-staging [13:45] seb128: data/general-hooks/ubuntu.py calls apport.hookutils.attach_upstart_logs(), so in theory it ought to work [13:46] pitti, https://launchpad.net/ubuntu/+source/apport/2.11-0ubuntu1 [13:47] seb128: apport-bug update-notifier includes the log for me === om26er is now known as om26er|doc [13:49] pitti, hum, dunno why there is no log on https://bugs.launchpad.net/ubuntu/+source/unity-settings-daemon/+bug/1284521 then [13:49] Launchpad bug 1284521 in unity-settings-daemon (Ubuntu) "unity-settings-daemon crashed with SIGSEGV in g_bit_lock()" [Medium,Confirmed] [13:51] Sweetshark, downloading lo on trusty, will check and post an update [13:51] pitti, is the codepath for segfaults different? [13:52] seb128: no, it could just be that ubunty.py crashes before it gets there [13:52] * pitti tries to kill unity-settings-daemon and see what apport does [13:53] hm, I still get the log in apport [13:53] seb128: could it be that the file was simply empty for the reporter? [13:53] process:21802): GLib-CRITICAL **: g_slice_set_config: assertion 'sys_page_size == 0' failed [13:53] I have a few instances of this ^ in the log, but nothing else [13:53] perhaps that's specific to my system and it doesn't log anything else? [13:53] we don't run it with G_DEBUG or anything [13:54] those warnings are firefox I think [13:54] seb128: oh, so perhaps a chromium user? [13:54] yeah, mine is mostly empty (some color profile warnings) [13:54] but why do firefox warnings appear in the u-s-d log? [13:54] pitti, well, the segfault shouldn't be possible without having glib warnings [13:54] I don't have color profile warnings [13:55] seb128: why wouldn't it? [13:55] glib would usually yell on assertions (and then abort), but that's a NULL pointer deref [13:56] pitti, because g_settings_get_value() can't return NULL [13:56] it's not quite 0, it's 10 [13:56] so when you say NEVER... [13:56] it might violate its interface if you violate preconditions [13:56] in particular: [13:56] g_return_val_if_fail (G_IS_SETTINGS (settings), NULL); [13:56] g_return_val_if_fail (key != NULL, NULL); [13:56] looks like something passes an 0x10 int as a pointer value or so [13:56] pitti, we should hit one of those g_return_val_if_fail to get a sources=NULL [13:57] seb128: if we had fatal criticals, we wouldn't have to play this guessing game right now ;) [13:57] well, segfault != critical [13:57] lol [13:57] let's not have that conversation today ;-) [13:57] in particular, segfaults are distinct from assertions (SIGABRT) [13:57] pitti: just lodging my usual complaint about how annoying it is to have to trace issues backwards to the original cause [13:58] instead of aborting as soon as we detect any issue at all [13:58] pitti, I think desrt is saying that if we had an sigabrt we would have a bt and not have to parse logs for warnings [13:58] but as said, we are not going to do that, so let's not discuss it ;-) [13:59] pitti, anyway, thanks for checking the apport side, I'm going to keep an eye on reports, it's weird that we don't have one in this case [13:59] yes, but we don't have an assertion failure here [13:59] so this is moot :) [13:59] something derefs the value 0x00000010 [14:00] pitti: i'm just saying that we could have had a nice assert failure with the backtrace showing the exact cause of the problem, but instead we get a segfault some time later... [14:00] pitti: the deref of 0x10 is pretty simple: someone gives a NULL pointer for a struct and we try to read something at offset 0x10 inside that struct [14:00] desrt: ah, now I understand what you mean [14:00] pitti, the issue on frame 4 " sources = 0x0" [14:01] pitti, that's a 'sources = g_settings_get_value (settings, KEY_INPUT_SOURCES);' [14:01] pitti: pretty sure seb is right about the cause being that the 'settings' object is wrong, earlier [14:01] pitti, that can't return 0x0 [14:01] except if the settings object is invalid [14:01] which should trigger a warning in the log [14:01] *nod* [14:01] but I found a condition where it can be buggy [14:01] so I'm just going to patch that [14:02] the init code does [14:02] xkb_init (manager); [14:02] set_devicepresence_handler (manager); [14:02] manager->priv->input_sources_settings = g_settings_new (GNOME_DESKTOP_INPUT_SOURCES_DIR); [14:02] [14:02] or set_devicepresence_handler() connect a signal with a callback that uses input_sources_settings [14:02] just going to put the g_settings_new before the handler [14:03] desrt, pitti: thanks [14:03] I wonder if the log gets flushed properly after each line [14:03] seb128: i'm suspicious. [14:03] i. e. could it be that it forgets to write the pending buffers after a crash? [14:03] unless ->priv was allocated in some strange way, it should be NULL from the start [14:03] and the bt shows a non-NULL value in 'settings' [14:04] so unless the settings are being created/destroyed/created/destroyed during the life of the object, ... [14:05] clearly we need fatal criticals [14:05] oh wait, you already seem to have discussed that... [14:07] desrt, https://git.gnome.org/browse/gnome-settings-daemon/tree/plugins/keyboard/gsd-keyboard-manager.c#n1765 [14:07] hm, how do I get back my settings after killing and rstarting unity-settings-daemon.. [14:07] pitti, which ones? [14:08] seb128: theme mostly; this dark theme hurts my eye :) [14:08] well, I'll just restart my session [14:08] yeah, I don't know why theme update don't pick up :/ [14:08] that's annoying [14:10] seb128: ya... indeed the settings is being created/destroyed [14:11] seb128: it's a bit weird, though -- they seem to be managing the reference quite properly -- g_clear_object() is really a good thing to be using here [14:11] so i have no idea :) [14:12] how is that possible though, it only seems to be created once? [14:12] desrt, is something supposed to do init to NULL when the object is created there? [14:13] seb128: the private structure gets zero-filled by g_type_create_instance [14:13] k, so that's not the order in the init function I guess [14:14] we would have null there otherwise [14:14] is it a problem that it's created in an idle? [14:14] * seb128 gives up, that bug is not important enough to justify spending hours on it [14:14] maybe that callback comes before the idle is run [14:14] what callback? [14:15] user_notify_is_loaded_cb [14:15] I don't think it's possible [14:16] also, the settings object would be NULL in that case, no? [14:16] none of the callers should be triggerable before the start_keyboard_idle_cb() is called [14:16] e.g set_devicepresence_handler (manager); is called in start_keyboard_idle_cb() === om26er|doc is now known as om26er [14:22] seb128: btw: i hate you for that gsettings bug yesterday [14:22] heh, don't blame the messenger! [14:22] just thought i'd mention that ;) [14:22] ;-) [14:23] shrug [14:23] ricotz's ppa is creating issues for unity-settings-daemon [14:24] desrt, is https://errors.ubuntu.com/problem/4272a581c30b062869c18b356abcf37fa51adc65 the same gsettings bug? [14:24] that's the most reported u-s-d issue in a week with 25 reports :/ [14:25] same backtrace, same bug [14:25] i should have a fix today [14:25] \o/ [14:25] it's one of those things that i could fix in about 50 possible ways [14:25] and i'm really not sure which is the best [14:26] but i think it's going to involve lots of hashtables [14:26] well as long as you are happy with the fix [14:26] hashtables make me happy... so it will prevent me from being too depressed while i write the patch [14:28] ;-) [14:28] attente, how do you bind keybindings to actions with the compiz/u-s-d key grabber? [14:29] attente, https://bugs.launchpad.net/ubuntu/+source/unity-settings-daemon/+bug/1284532 is a bit weird [14:29] Launchpad bug 1284532 in unity-settings-daemon (Ubuntu) "print screen button appears hard-wired to screenshot" [Low,Confirmed] [14:29] attente, the defaults settings are, "print" = screen capture, "ctrl+print" = capture to clipboard [14:30] attente, if you unset "print" from screen capture, pressing the print key triggers a capture to clipboard [14:30] which is supposed to be ctrl-print [14:30] attente, do you have any idea what's going on? [14:30] seb128, not really, but i'll look at it [14:31] attente, don't worry much about it, it's a corner case (who reassign the print screen to something else?!) [14:31] lol [14:32] desrt, hi [14:32] hikiko: good morning/afternoon [14:32] good morning :) [14:32] ooh [14:32] if you unset print and then reset it then it doesn't print screen any more [14:32] for me anyway [14:32] I'm looking at the g_settings_get_value/set_value [14:32] but there's no param for the type [14:33] hikiko: you don't need it. the type is in the schema. [14:33] Laney, "re-set" you mean? [14:33] you're always guaranteed to get the correct type of value [14:33] put it back to print [14:33] right [14:33] wfm [14:33] nfm [14:33] cool that was my question :) [14:33] did you restart u-s-d in between? [14:33] no but doing that does restore it [14:33] I think things go weird with the grabbing if you restart it [14:33] hum, k [14:33] weird [14:34] and now the grabbing is broke after another restart [14:34] I had one issue of that earlier but couldn't reproduce [14:34] woe! [14:34] isn't the issue with restart the state thing desrt, attente and larsu discussed during the meeting yesterday? [14:35] yes, I noticed that while hacking on the media-keys plugin [14:35] in fact, I'm seeing the same issue again today after unity decided to crash [14:35] (I shouldn't have used Alt-Tab!) [14:35] :( [14:35] might have to do the xdg runtime dir thing after all [14:36] ya, probably [14:36] attente, it's another "would be nice to have, but don't worry too much about it, restarting u-s-d is what we do for testing, not something most users are going to do" [14:36] attente: I still don't understand why that is necessary... [14:37] but I guess you and desrt talked about it in detail, so it'll be fine [14:38] well.. it doesn't just happen for testing unity/u-s-d, it happens if either crashes hard [14:38] right [14:38] which isn't the norm [14:38] attente: as opposed to crashing softly? [14:38] * larsu wonders what that would look like [14:39] a backtrace with lots of pillows?! [14:41] maybe put an memory eating forkbomb in the exception handling? that would softy put the machine to a halt ... [14:41] when that happens you get other issues [14:41] like g-s-d/u-s-d restarting screws your theme [14:41] ... until the OOM puts an end to it, spoils the party and wakes you up. [14:44] ritz: I think I know what the issue is on trusty ... [14:44] Sweetshark, this I would love to know [14:45] still downloading the source === psivaa-lunch is now known as psivaa [14:53] ritz: https://gist.github.com/bjoernmichaelsen/9230792/revisions <- I think I tried to cp the file before it was moved in place ... [14:54] aah [14:54] k, so the rules files is hosted on git [14:55]