[00:15] hi guys :) === dragonfyre13 is now known as dragonfyre_away [05:46] good morning [05:49] didrocks: good morning! [05:52] hey nhaines === API is now known as Guest32644 [07:29] hey folks === kim0|holiday is now known as kim0 [08:25] kamstrup: hey, do you have a second for playing with some cjk foo? [08:25] in xapian (apt-xapian index) [08:26] didrocks: give me 15 mins === Guest32644 is now known as apinheiro [09:30] njpatel: it's likely not one of those indicators: https://launchpadlibrarian.net/77655516/log.png [09:30] the jump at the beginning was caused by opening some programs iirc [09:30] htorque, ah, excellent [09:30] htorque, I'll try and take a look today [09:30] but i see that the appmenu is missing, so that's something left to test [09:32] njpatel: however, i don't quite understand your directions: "probably from metacity, without an exsiting global menu" - should i run metacity --replace and run the command? with valgrind? or just look at mem consumption? [09:33] htorque, you can initially just look at memory consumption of the indicator-loader with appmenu, to confirm/deny if it's leaking when you start an app [09:33] htorque, valgrind will give you a deeper analysis if it is, I guess [09:33] but i should replace compiz & co with metacity, right? [09:34] htorque, the reason for running it in metacity is that you can't have two appmenus running, so you can't run in unity [09:34] htorque, right, exactly [09:34] just metacity --replace from a terminal will do [09:35] so now i don't have a panel at all - is it still useful then? [09:35] yep [09:36] htorque, with indicator-loader, you're just testing a single indicator at a time, which is good to isolate the issue [09:36] if it doesnt' leak, most likely something is wrong in the service itself, or maybe another indicator [10:11] kamstrup: I didn't touch that ;) [10:11] didrocks: in a nutshell what I mean: python -c "import os; os.environ['FOOBAR'] = 1; print os.environ['FOOBAR'] == 1" [10:11] njpatel, ok thanks, although he is not here ... [10:11] just in case [10:11] throws an error because you can only pass strings into environ[] [10:11] apinheiro, yeah, he's in US so her ein a few hours [10:11] njpatel, these messages from panel-service: [10:11] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Child is realized, but doesn't have a GTK Widget! [10:11] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Got new menuitem signal, which means they want something [10:11] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: that I simply don't have [10:12] do you know if those are normal messages [10:12] no idea [10:12] I actually have the same question for him [10:12] didrocks: and if you change 1 to '1' you get False: python -c "import os; os.environ['FOOBAR'] = '1'; print os.environ['FOOBAR'] == 1" [10:12] kamstrup: oh, I clearly know that, I'm not the one making the patch once again :-) [10:14] njpatel, ok, is because I'm working on that panel-service crash with a11y enabled, and that "he doesn't have a GTK widget" sounds something to look at [10:14] didrocks: the reason I bang on about this is that afaics this means the db is reindexed on *every* run [10:15] didrocks: is the cjk patch applied to xapian by default now in Oneiric? [10:15] kamstrup: yeah, and the testsuite doesn't pass anyway, so I think nobody tested it [10:15] kamstrup: yes, it is :) [10:15] didrocks: how can that possibly go wrong?!?!?!11 ;-) [10:16] kamstrup: touching data! never be afraid :-) [10:16] apinheiro, heh, yeah :) [10:16] didrocks: i'll let you touch my data [10:17] kamstrup: that's so kind! I let you spy my computer with zg already, so, in some kind of mutual touch ;) [10:17] didrocks: kinky [10:31] njpatel: the amount of leakage is definitely influenced by the menu (eg. with 180 simple menu items in the menubar, u-p-s mem consumption raises by 15mb, with one it's just a couple of kb) [10:32] htorque, right, the reason why I thought it must be appmenu is because the panel service only extracts info about the toplevel menu items (it has no idea about the actual menus), and hence I felt that appmenu is leaking something in constructing the actual menu tree [10:32] good, because putting more items in a menu didn't change anything [10:50] didrocks: attached a fixed patch to https://bugs.launchpad.net/ubuntu/+source/apt-xapian-index/+bug/838274 [10:50] Ubuntu bug 838274 in apt-xapian-index (Ubuntu) "The indexer should use the new CJK tokenizer" [High,Fix committed] [10:51] kamstrup: did yo utest it, you should get a missing "import os" [10:51] in update-apt-xapian-index [10:52] kamstrup: tell me if the testsuite still run after that :) [10:52] didrocks: ah, wait... [10:57] didrocks: needs some more fixing I can see :-) [10:57] kamstrup: indeed :) [11:51] didrocks: ok v3 of the patch is running with all tests - I even added 3 new test cases so the diff has grown somewhat. But this should be upstreamable now i think [11:52] kamstrup: excellent news, do you think it's that part which is used by software-center or we should force a reindexing as well before reopening the database as well? [11:53] didrocks: i don't have the faintest clue what S-C is doing :-) [11:54] didrocks: S-C has it's own index in addition to the axi one [11:54] so S-C prolly needs patching in a similar spirit [11:54] kamstrup: ah, because I was thinking that it was using the axi one [11:54] so we could have activate this env var unconditionnaly [11:54] didrocks: it is using 2 indexes [11:55] i think we should, but still, the code needs to work regardless as other distros might not set it [11:55] (mostly thinking about the zg bits here though) [11:55] kamstrup: indeed [11:56] didrocks: is upstream watching these bugs, or should we ping someone? [11:57] kamstrup: I think we should ping them, but I won't block on that for now [11:59] didrocks: i'll move on to the zg-fts bits cjk enablement now, is mvo on top of the S-C issue or should I take that later? [11:59] kamstrup: mvo is on vacation, not sure if you have the time to look at it, otherwise tremolux will (but he doesn't really have a clue as well) [12:00] hehe :-) [12:00] i'll see if I get around to it === MacSlow is now known as MacSlow|lunch [12:01] JohnLea, Hi could you confirm if bug 838759 is a design decision or a bug? [12:01] Launchpad bug 838759 in unity "window title should not appear until cursor leaves the menubar" [Undecided,New] https://launchpad.net/bugs/838759 [12:02] kamstrup: great ;) [12:02] om26er: I know it's strange but it seems something the design want... No menu unitil you move again! :o [12:02] oh twitter people are saying its by design :/ andyrock Trevinho [12:03] yes.. [12:03] Trevinho, :( [12:03] om26er, have you said my pastebin? [12:03] it should be easy to fix but the comment let me think that it's by design [12:04] andyrock, the window button problem stands, doesnt it? [12:04] on second click on the menus window buttons dont show [12:04] unless cursor is moved [12:05] om26er: are you still confirming the bug of the VP switch (move to another VP and select the launcher button of the focused window in another vp => Nothing happens after the click) ? [12:05] If you do please report, I can confirm :P [12:05] (and work on it) [12:06] Trevinho, i confirmed it yesterday, i am reporting it now :) [12:06] thanks :) [12:06] since DBO can't reproduce it... [12:06] om26er, which window button problem? [12:07] om26er: check these issues using my branch please, I guess I've solved them... [12:07] andyrock, see in the video, when i click on the menu again window controls dont appear the faded title does [12:07] om26er: Ah, That's true [12:07] I'll fix that. [12:08] Trevinho, i'll report that and assign you as well ;) [12:08] ok [12:08] launcher stuck hard this time, it wont hide :/ [12:08] These panel issues are really annoying [12:08] om26er, wanted behavior too? [12:08] but I guess we're in the right way... [12:09] i mean the window buttons appear only when the menu appears... [12:09] andyrock: it could be... But it's better to have a reply by JohnLea (or design, in general) about these things. [12:09] so if the menu is not shown why we should show the window buttons? [12:10] Trevinho, mpt too (I guess) [12:10] of course... [12:11] no, andyrock... The problem is on the second click [12:11] On the second click only the menus are shown [12:11] and this is not so consistent [12:11] ah i got it... [12:11] also if that is enough (as you reclick without moving, we guess you want to see the menu) [12:12] Trevinho, btw i guess that design info are needed [12:13] *is [12:36] Trevinho; andyrock; just about to grab lunch, will get back to you in 30min [12:36] JohnLea, np [13:03] om26er, Trevinho; re. bug #838759, good catch! Yes the current behaviour wrong and needs fixing. I've confirmed the bug [13:03] Launchpad bug 838759 in unity "window title should not appear until cursor leaves the menubar" [Undecided,New] https://launchpad.net/bugs/838759 [13:07] om26er, Trevinho; any idea how people got the idea that bug 838759 behaviour was by design? !! :/ [13:07] Launchpad bug 838759 in unity "window title should not appear until cursor leaves the menubar" [Undecided,New] https://launchpad.net/bugs/838759 [13:16] JohnLea: this is the code that implement that http://pastebin.ubuntu.com/679655/ [13:16] I also found a bug somewhere mentioning it [13:18] Trevinho; cool, I've confirmed bug https://bugs.launchpad.net/unity/+bug/838759 if you could fix it that would be great ;-) [13:18] Ubuntu bug 838759 in unity "window title should not appear until cursor leaves the menubar" [Undecided,New] [13:21] ok JohnLea I'll work on it [13:24] JohnLea: and what about the other thing we discussed before? (as you can see in https://launchpadlibrarian.net/78696225/5ii.ogg on second click no window buttons are shown) [13:34] Trevinho; yes that's a bug, would that not be fixed with the solution om26er outlined in bug #838759? [13:34] Launchpad bug 838759 in unity "window title should not appear until cursor leaves the menubar" [Undecided,New] https://launchpad.net/bugs/838759 [13:35] I guess it will JohnLea. I've to check btw [13:36] JohnLea: however about the #838759 I'm pretty sure that I've read about that behaviour in an unity bug (maybe related to ayatana design too).... [13:36] Otherwise I would have fixed for a long time :D [13:38] Trevinho; it is definitely a bug ;-) If any other items like this come up please ping me and I'll try to help clarify [13:48] No problem... However I'm happy it's a bug :P I didn't like that behavior! [13:51] hi guys :) [13:52] Kaleo: I'll have to upgrade my vm to Oneiric before being able to work to that bug ;) doing right now [13:55] Andy80: good luck ;) === MacSlow|lunch is now known as MacSlow [14:05] tedg, hi, you here? [14:06] apinheiro, Yes, but I'm testing something where my machine might shutdown (or I get a good backtrace)... give me a second. === smstudy is now known as smspillaz [14:18] apinheiro: the one who work at Igalia? [14:19] Andy80, yes [14:20] I usually use API as my IRC nickname, but someone already registered it [14:20] apinheiro: hi :) don't know if you remember me.... I did a summer internship 2 years ago in your company :) que tal? [14:21] tedg, btw, my previous question, Im trying to fix that crash on panel-service when a11y is on, and I have these messages: [14:21] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Child is realized, but doesn't have a GTK Widget! [14:21] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Got new menuitem signal, which means they want something [14:21] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: that I simply don't have [14:21] it is normal? [14:21] Andy80, hi, yes I remember you ;), fine thanks [14:22] apinheiro, Uhm, I don't think that those should be happening. Do they happen without a11y? [14:24] well, I asked that in the morning [14:24] and [14:24] * apinheiro looking backlog [14:24] andyrock: I'm upgrading my VM to Oneiric. I will need your help with that Nautilus-dbus stuff later if you can [14:25] of course [14:25] apinheiro, yeah, he's in US so her ein a few hours [14:25] njpatel, these messages from panel-service: [14:25] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Child is realized, but doesn't have a GTK Widget! [14:25] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Got new menuitem signal, which means they want something [14:25] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: that I simply don't have [14:25] do you know if those are normal messages [14:25] no idea [14:25] I actually have the same question for him [14:25] tedg, ^ [14:25] tedg, I also see messages like: [14:25] "IndicatorObject class does not have an accessible description." [14:26] Andy80, btw Kaleo told me that in QT there is already a dbus helper class... [14:26] and finally, I have a question about indicators, although not sure if they are related with dbusmenu [14:26] on indicator.h: [14:26] @get_menu: Gets the image for this object. Should be set [14:26] to #NULL if @get_entries is set. Should NOT ref the [14:26] object. [14:26] apinheiro, The not having an accessible description is just to make us do the work, though it shouldn't harm anything. [14:27] so i think you should use it instead of UnityCore::DBusWrapper [14:27] andyrock: I know that class, but I don't need help for using the QDbus class, I need to know which parameters to give it, which strings to use ecc.... [14:27] tedg, ok, anyway I'm just setting "" in this case [14:27] but about that doc [14:27] do you know what would happen if you ref that menu? [14:27] Andy80, no problem then :) [14:27] apinheiro, Memory leak [14:27] andyrock: thanks :) [14:28] tedg, why, shouldn't the indicator be freed on the last unref? [14:28] apinheiro, But, in general, it should be "does not" instead of "should not" [14:28] Andy80, yw... [14:28] apinheiro, The indicator should, but it's saying that by calling "get_menu" you are not incrementing the ref count. [14:28] tedg, ah ok [14:28] but you could increment the ref count [14:28] right? [14:29] I mean to ensure that the object will be alive while you are working with it [14:29] apinheiro, Yes, and you should if you're working with it. Just some of the callers just pass it to GTK which does it's own increment, so I didn't want to have one stuck in the middle. [14:30] apinheiro, This way users can treat it more like a floating count. [14:30] tedg, anyway don't worry, just asking, on the a11y side there are a weak reaf, so if the object is freed, the atk object should be notified [14:30] tedg, so going back to my "weird messages questions" [14:30] do you know if I'm doing something wrong? [14:30] I mean if there are something missing in my system? [14:31] apinheiro, No, it's basically saying that it thinks the menu item should have a GTK Menu Item associated with it, but it doesn't. [14:31] apinheiro, It could be a sign of some sort of race condition. [14:31] apinheiro, I'd be curious if the a11y slows down the construction of the GTK+ widgets in some way that we're not taking into account. [14:32] working on a bug and found a race condition ... this can be tricky [14:32] hmm [14:32] tedg, well, in that case I will disable a11y and check it [14:32] but anyway, njpatel said that he also wanted to ask [14:32] apinheiro, Yeah, that'd be best, just to see. [14:32] so I suppose that this also happens without a11y enabled [14:32] apinheiro, Yeah, njpatel is like that ;-) [14:33] I like to ask things [14:33] tedg: njpatel asked me to run indicator-loader to check memory consumption of an indicator, but that seemed to fail and i think i got the very same warnings [14:33] could this be connected? [14:33] * tedg keeps njpatel's secret that he's really 5 years old. [14:33] htorque, UBUNTU_MENU_PROXY="" before using indicator-loader apparently! [14:33] tedg, but WHY? [14:33] tried that [14:33] htorque, Getting the same messages or no indicator at all? [14:34] the indicator-loader just showed "File" (appmenu loaded) [14:34] njpatel, we were talking about this: [14:34] apinheiro, yeah, he's in US so her ein a few hours [14:34] njpatel, these messages from panel-service: [14:34] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Child is realized, but doesn't have a GTK Widget! [14:34] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: Got new menuitem signal, which means they want something [14:34] (unity-panel-service:5672): LIBDBUSMENU-GTK-WARNING **: that I simply don't have [14:34] do you know if those are normal messages [14:34] htorque, Did you set UBUNTU_MENUPROXY= or the wrong env like njpatel said ;-) [14:34] no idea [14:34] yeah, thanks apinheiro [14:34] I actually have the same question for him [14:34] and your "I actually have the same question for him" [14:35] tedg: the right one ;) [14:35] htorque, Hmm, that works for me... [14:35] htorque, Which indicator? [14:35] i posted the output of indicator-loader here but that did get logged and i'm on a different machine now :( [14:35] appmenu [14:36] btw, when I get the panel-service working without a crash [14:36] I can open the menus with the mouse pointer [14:36] but I'm not able to navigate or give the focus with F10 [14:37] is this only happening to me? [14:37] njpatel, htorque ? ^ [14:37] F10 doesn't do anything here [14:38] ?¿ [14:38] F10 is supposed to give the focus to the panel [14:38] have this changed? [14:39] is not related to those memory comsuption issues, but just asking as is related with the panel [14:39] when opened with the mouse i can navigate within the toplevel menu, but i cannot switch left and right [14:39] somewhat offtopic I know [14:39] htorque, urgh [14:39] that confirms that panel key nav is broken :/ [14:39] htorque, Hmm, I'm not sure that appmenu works in the loader in general, since it's based on which window has focus. When the loader has focus, it tries to show menus from it. So you can't really click on another window and interact with its menus. [14:40] It unfortunately makes appmenu ridiculously hard to debug. [14:40] tedg it was just to confirm whether it's causing bug 835646 [14:40] Launchpad bug 835646 in unity (Ubuntu) "unity-panel-service memory leak when starting various applications" [High,Confirmed] https://launchpad.net/bugs/835646 [14:40] We should probably add a delayed mode for using the loader. We could use the loading hints to determine whether it should be on or not. [14:41] njpatel, Did you ever include a mode in unity-panel-service where you could specify which indicators it would load? [14:42] apinheiro: i think it's not broken per se, but rather that the mouse pointer keeps the selection on the toplevel menu it's pointing to - with F10 working, this wouldn't be a problem i guess [14:43] tedg, I did, but it seems like someone removed it :( [14:44] njpatel, do you know something about this F10 issue? I mean, is F10 now assigned to a different short-cut? [14:44] apinheiro, nope, need to investigate it [14:45] njpatel, ok, I will keep focus on the crash then [14:45] * apinheiro reminds that he need to check if those messages are also present without a11y enabled === davidcalle_ is now known as davidcalle [14:56] tedg: that's the output i got from indicator-loader when testing it earlier: http://paste.ubuntu.com/679567/ [14:57] will try again [14:58] htorque_, You've got a GTK2/3 conflict there. You need to use loader3 with the GTK3 indicators. [14:59] oh dear [14:59] is that in libindicator-tools? [15:00] htorque_, I think so. [15:00] libindicator3-tools... [15:00] Oh, yes [15:00] meh [15:03] tedg, forget my question, it seems that those warning only appear if I launch by hand the panel service from a terminal that it is not in the same session that unity [15:04] executing this "in a normal way" I don't get those messages [15:04] sorry for the noies [15:04] noise [15:04] Hmm, that's interesting... I wonder why that is. [15:05] \o/ loader3 works [15:06] tedg, in my environment I have a "working session" and the session where I made the "unity --replace", so I move from one to the other using ctrl+alt+F7/F8 [15:06] I was running the panel-service from a terminal on my working session, [15:06] (with a export DISPLAY) [15:06] and in this case I got those messages [15:06] njpatel: starting gimp adds lots of ram to the loader process ;-) [15:06] again, sorry for the noise [15:08] njpatel: should i now run valgrind on that and attach the output to the bug report? [15:08] htorque_, Yeah, that'd be good. [15:08] will do :) [15:08] htorque_, Firefox is also good if you have lots of bookmarks. [15:09] unfortunately no ff user [15:09] kile and gimp work fine [15:30] tedg: there are a couple of "???" entries in the log about libfontconfig/libexpat/libpango but i couldn't find debug symbols for those packages. is that a problem? also i forgot to install dbgsym for loader3 :-/ [15:30] htorque_, naw, we're not going to get those solved today :-/ [15:34] ok, it's uploaded: bug 835646 [15:34] Launchpad bug 835646 in unity (Ubuntu) "unity-panel-service memory leak when starting various applications" [High,Confirmed] https://launchpad.net/bugs/835646 [15:34] if you need anything else, let me know [15:37] tedg: hi [15:37] jml, Howdy [15:37] htorque_, Thanks! [15:38] tedg: yw :) [15:39] Unfortunately that doesn't seem to have any smoking guns in it. [15:40] * tedg quickly concludes that there's no way it's his fault, it must be njpatel's bug ;-) [15:41] tedg: I just tried the valgrind thing you suggested [15:41] tedg: but... umm... where does the output go? [15:42] bug 740382 [15:42] Launchpad bug 740382 in indicator-session (Ubuntu) "gtk-logout-helper crashed with SIGABRT in __libc_start_main()" [Low,Confirmed] https://launchpad.net/bugs/740382 [15:42] jml, You'll need to redirect it to a file. [15:42] jml, So I guess I should have included that in my command line :-) [15:43] htorque_, Did you just start GIMP, or start it and shut it down before closing the loader? [15:44] tedg: np. I'll do that now. How will I know if I've triggered the condition? [15:44] jml, Well, I'm guessing that it doesn't shutdown, no? [15:44] tedg: ok :) [15:45] jml, I'm guessing that somewhere in that log there has to be an invalid read/write. [15:46] tedg: stdout & stderr, or just one of them? [15:46] jml, Oh, grab them all. /me is feeling generous with LP's disk space. [15:48] tedg: i started it, waited for the loader to show the menu, then closed it. started it again, a.s.o. [15:48] htorque_, Okay, makes sense. Just making sure. [15:48] As much as I love blaming njpatel, I try to not do it arbitrarily, well, not all the time. [15:49] htorque, it's tedg's fault and he knows it [15:50] I don't leak. Nor does my code. [15:50] i only know that it's not my fault. opening applications seems like expected behavior. :P [15:52] Some times ago I also gave a look to this problem [15:52] JohnLea, Finally I found the time to fix this https://bugs.launchpad.net/unity/+bug/832631 [15:52] Ubuntu bug 832631 in unity (Ubuntu) "Dnd hover timeout in spread view does not work as expected" [Undecided,Confirmed] [15:52] i've add this behavior [15:52] and it seems that unity-panel-service itself is working well [15:52] according to valgrind [15:52] it seems something more related to the indicators [15:53] and libappmenu seems the one [15:53] Trevinho, Heresy I say! [15:53] * tedg sends the Spanish inquisition! [15:53] tedg: I know it's not your fault... :) [15:53] the timeout is reset if the mouse hover another window or if the mouse move more than 3px [15:54] but, the fact is that also running the appmenu-indicator in natty, the leak persists [15:54] Trevinho, Did you try running the test app indicators under valgrind and see if anything came up? [15:54] so, maybe it's not directly connected to that indicator, but to something going wrong in the framework [15:54] I don't, but there's a valgrind log attached to the mentioned bug [15:54] https://launchpadlibrarian.net/78705404/valgrind.log [15:55] there is a 6x6 rect and the mouse pointer is the center, when the mouse pointer go outside of this rect the timeout is reset [15:55] is it ok? [15:55] Trevinho, That is appmenu not application though... [15:55] Really both of them are pretty dynamic. If there's a memory leak, it'd have to be one of those. [15:57] tedg: attached [15:57] tedg: my computer doesn't have working g-s-d now. [15:57] ah, sorry tedg I'll check [15:58] applications definitely has one: bug 829961 [15:58] Launchpad bug 829961 in Unity Foundations "indicator-applications leaks memory" [High,Triaged] https://launchpad.net/bugs/829961 [15:58] tedg: so, uhh, I'll do launch that manually [15:58] but that's not affected by starting applications [16:14] andyrock; great news! Do you know if it will be landing in today's or next week's release? [16:15] it's not a unity bug but a scale plugin one [16:15] https://code.launchpad.net/~andyrock/compiz-core/fix-832631 [16:33] andyrock: poke [16:33] andyrock: what's the reason for resetting scale on mouse motion ? [16:34] (or a buffer of mouse motion) [16:34] seems like it only makes sense to do it per window [16:35] err [16:35] the scale dnd timeout [16:35] smspillaz, well if you move the mouse a lot without changing window the timeout don't reset [16:35] *doesn't [16:35] JohnLea, ^^^ [16:36] andyrock: ah, so you want it so that it it resets when you move the mouse around inside the client window a little bit, but not when you move it a lot [16:37] smspillaz, yes [16:37] andyrock: in that case it might make more sense to accumulate motion events [16:37] feel free to remove the rect... :) [16:38] andyrock: like, watch MotionNotify, and if we've moved more than X pixels then reset the timeout [16:38] (timer.stop (), timer.start ()) [16:38] during the xndd we don't get the motionnotify events [16:38] ah, crap, that's right [16:38] ehh [16:38] i think because the mouse is grabbed by the source [16:39] well hang on [16:39] but the source should do it for us [16:39] Trevinho, bug 838923 wont that be compiz? [16:39] Launchpad bug 838923 in unity (Ubuntu) "Start dragging a maximized application title causes it to re-decorate" [Undecided,In progress] https://launchpad.net/bugs/838923 [16:39] you should get xdndPosition? [16:39] om26er: the redecoration is done by unity [16:39] yes.... [16:39] andyrock: so you can use that, right ? [16:39] there are a few other bugs reported already about it though. [16:39] case ClientMessage: [16:39] if (event->xclient.message_type == Atoms::xdndPosition) [16:39] { [16:40] smspillaz, ^^^ [16:40] smspillaz, oh ok.. [16:40] andyrock: what's wrong with using xdndPosition ? [16:40] smspillaz, we alredy use xdndPosition :) [16:41] and get x,y form the message event [16:41] andyrock: right. so instead of using the rect, what you could do is accumulate the pointer movement [16:41]