=== vibhav is now known as Guest80328 [00:36] UserError: libsound2 isn't part of cloudimg, so its not an issue for ec2, azure, openvz, openstack. [00:36] is there an easy way to download "newest" build logs for a source package for all supported releases? [00:38] jrwren, i'm sorry I ruined your worldview then :(. It does indeed exist in many of those locations for sound processing support by default. [00:39] as do many unneeded modules even in the minimal VM kernel versions [00:42] https://coderwall.com/p/a56j3w -> https://gist.github.com/steakknife/6086608 [00:52] so I'm trying to debug using valgrind and I am being hit with what I can only call an idiocy of Xwindows... while displaying a pop up window, input is locked and you can't alt tab so I can't switch to the terminal running valgrind to tell it yes, I do want to attach gdb now that you noticed the error.. is there a workaround for that? [00:52] ( other than switching to a tty and kill -9ing valgrind ) [00:58] psusi: Welcome to the joy of X11 grabs. Run valgrind from a TTY. [00:59] RAOF, can't you disable or break that stupid lock? also I just noticed that menus in thunderbird don't seem to do this... I wonder why? [01:00] constantly switching back to a tty to tell valgrind to continue a dozen times before it gets to the real problem does not seem appealing [01:00] Oh, yes. You can break the lock by enabling the XKB option which allows you to break locks. [01:00] xkb? [01:01] http://who-t.blogspot.com.au/2012/01/xkb-breaking-grabs-cve-2012-0064.html [01:01] xkeyboard-config before 2.5 in X.Org before 7.6 enables certain XKB debugging functions by default, which allows physically proximate attackers to bypass an X screen lock via keyboard combinations that break the input grab. (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2012-0064) [01:02] how have I never heard of control+alt+numpad /* ?? [01:02] Because we disable it, because it kills your screen lock? [01:03] exactly, that would have been a pile of fun back at school [01:06] lol [01:07] I remember using that to insert ascii 255 characters into file names in high school [01:07] in DOS it looked like a blank, so you could add it to the end of a file and not be able to tell it was there [01:07] but then you couldn't touch the file or directory if you didn't add the character [01:07] confused the hell out of the teacher when he couldn't see in the directory or delete the file [01:08] then Windows 95 came out and things got even funnier... it would show an _ in place of the 255... and yet still puke when you tried to actually touch the file [01:08] haha [01:09] I'm sure at one point we had no less than 25 copies of doom on the netware server protected using that little trick [01:11] awesome :) [01:12] though that wasn't nearly as cool as when a class mate figured out how to use the dos debugger to disassemble login.exe and modify it to connect as a print server object instead of a user... print servers were supervisor equivalent and had no password ;) [01:13] heh, I never heard of that, either, that also would have been handy to know.. [01:15] RAOF, I don't suppose there's a way to make menus NOT grab in the first bloody place? ;) [01:19] RAOF, bah... breaking the lock seems to kill the application [01:23] odd, why does indicator-application-service refuse to start up after i restart compiz? [01:26] psusi: No, you can't make menus not grab input; otherwise, they couldn't close when you clicked somewhere else. [01:27] RAOF, why not? they lose focus when you click elsewhere, and close.. the menus in thunderbird don't grab [01:28] psusi: Then I suspect the menus in thunderbird are hilariously broken under focus-follows-mouse? [01:29] oh wait they're back O_O [01:29] wait... so this grab idiocy is all to avoid focus-follows-mouse closing the menu when you move the cursor outside the menu? [01:30] No, this grab idiocy is the only way guaranteed by the X11 protocol to work. [01:30] for *what* to work? [01:30] For menus to be able to determine when someone has clicked outside them. [01:31] Yes, this sucks. [01:31] I don't understand why you would care.... [01:31] Because you want to close the menu when someone clicks outside it? [01:31] so close it when you loose focus? [01:31] That's standard menu behaviour. [01:31] There is no guarantee that clicking somewhere else robs you of focus [01:32] does it not? [01:32] how doesn't it? [01:32] It's entirely up to the window manager. [01:32] hmm i suppose it's not directly, yeah [01:33] psusi: kill compiz repeatedly until it doesn't come back, then try switching focus by clicking =p [01:33] oy ve [01:33] more X11 nicompoopedness [01:33] Which is why X11 toolkits use the only mechanism guaranteed by the X11 protocol to work :) [01:33] * hyperair lols [01:33] Yeah, this sucks. [01:34] so... what would be so wrong about not closing the menu if your wm is dead and you can't switch to another window? [01:34] RAOF: is there any window manager around that allows you to click on another window without switching focus? [01:34] hyperair: Almost certainly. [01:34] =\ [01:35] so how does it switch focus then? [01:35] hotkey? [01:35] Oh, it'd probably switch focus in *some* situation; maybe you need to click on the titlebar or something. [01:35] and X doesn't have a soft grab where you can capture clicks outside your window, but not bloody prevent switching windows? [01:36] psusi: Correct. “Soft grabs” are “give me a full-blooded grab when $CONDITION”. [01:36] For $CONDITION that doesn't include “an input event happens outside my window ☺” [01:37] so.. what would be so bad about the menu not closing when you click outside the window and don't loose focus? [01:37] That's not how a menu is supposed to act? [01:37] so? [01:37] it's also not supposed ot disable alt-tab ;) [01:37] So toolkits want to provide a menu that provides certain behaviours. [01:38] Also note that your “lose focus => close menu” breaks under raw X11, because the default behaviour is focus-follows-mouse. [01:38] derp [01:38] I don't see a problem with the menu closing if you move the mouse to another window [01:39] and have focus-follows-mouse on [01:39] psusi: what about those times you accidentally move your mouse out of the menu window? [01:39] Then you're perfectly welcome to write your own toolkit. [01:39] That doesn't have this behaviour. [01:39] GTK and Qt both want their menus to behave like menus :P [01:39] how about I just patch gtk to not be stupid? ;) [01:40] I'd like them to behave like menus too... which means not disabling alt-tab [01:40] eh well at least esc works [01:40] hyperair, yea... unless the app is hung, for instance, by the debugger ;) [01:40] though you're kinda screwed if the application hangs while a menu is open [01:40] :) [01:40] UserError: lol. I promise, you did nothing to my world view. [01:41] * psusi wonders when he will be able to swtich to wayland [01:41] The day you install E19 [01:42] RAOF, there should at the very least be a setting to disable that behavior in the toolkit... if I wanted to add a patch to do that, what should I search for? what is the api to grab? [01:43] if only just for debugging ;) [01:43] psusi: just go look inside the GtkMenu widget code? [01:43] hrm... good idea [01:43] psusi: You're looking for XGrabCursor, likely. [01:44] export GTK_MENU_NO_GRAB=1 [01:45] nah doesn't look like it [01:45] grepping shows no XGrabCursor [01:46] but i see a bunch of XGrabKeys [01:46] oh i see gdk_device_grab in gtkmenu.c [01:51] psusi: Would this act as the basis of a workaround, so you can start valgrind/gdb at the correct point? "while ! xprop -name 'New Tab - Mozilla Firefox' | grep -q _NET_WM_STATE_MODAL; do sleep 1; done; xdotool -window id-of-valgrind-window " [01:53] TJ-, I don't follow... I'm actually running gparted under valgrind and noticed some invalid memory references that sometimes are followed by an actual crash... I can kind of sort of reproduce it by right clicking up a menu at the right time [01:56] funny though... I can't get it to crash when not running under valgrind [01:56] I was thinking if gparted has a 'modal' window keeping focus, and you need to send input when that appears to the valgrind window to fire up gdb, then a shell fragment like I showed in another terminal can watch for the modal window and when it sees it, can send the key-strokes needed directly to the valgrind/gdb terminal window (thus avoiding the modal focus lock) [01:57] Obviously, you can tweak the basic idea to trigger on your specific circumstances and windows [01:59] Riddell: Looks like eigen2 was completely removed from Debian in favour of eigen3, but a few things in Ubuntu (almost entirely KDE) still {build-,}depend on it. Any chance that'll be fixed? [01:59] shadeslayer_: ^ [02:03] shadeslayer_: Looks like merging our kdeplasma-addons packaging with Debian (or, at least, pulling eigen3.patch) would sort that package, and probably get it building on ppc64el again (obviously also dropping the arch-restricted build-dep) [02:28] * hyperair thinks a gdb session in byobu/screen/tmux would be much more helpful [02:29] that way you could just connect to it via ssh or from a tty [02:37] hyperair: ++, also gdbserver [02:37] oh okay, i didn't know about gdbserver [02:45] whats the best way to capture a full kernel oops when it happens during cloud-config ? [02:49] serial connection? [02:49] netconsole? [02:56] hyperair: good idea, I'll add a serial connection to the vm [02:57] :) [02:57] we were having lockups on that though [02:57] hopefully I can have cake and eat it [02:57] lockups on serial? o_O [03:03] yeah [03:04] 'fun' and didn't have time to debug... === FJKong_AFK is now known as FJKong [04:25] ScottK: Any comments on my eigen2/eigen3 observations above to Riddell and shadeslayer_? [04:26] infinity: About to go to bed and haven't read them yet. I'll weigh in tomorrow. [04:26] ScottK: Mmkay. [05:11] Laney: I may not understand the cross-compiling magickry you're trying, but it's possible you need to give qtchooser a specific configuration that understands where to get binaries + libraries [05:12] or binaries only if libraries are already fine [05:30] Good morning [05:32] morning pitti, thanks for kicking the retracers yesterday :) [05:32] hey sarnold; thanks for the poke [05:46] sarnold: I adjusted apport to use a per-run temporary launchpadlib cache now, so this shouldn't happen again [05:47] pitti: oh, cool! :) thanks === vibhav is now known as Guest86949 [07:52] good morning [08:07] hi, I proposed the code merge for fix bug LP: #1284447, may someone approve the merge and make the update available for users on precise? https://code.launchpad.net/~taihsiangho/ubuntu/precise/pcmanx-gtk2/fix-for-1284447/+merge/211676 [08:07] Launchpad bug 1284447 in pcmanx-gtk2 (Ubuntu Trusty) "middle click could not close the tab you selected by middle click" [Low,In progress] https://launchpad.net/bugs/1284447 [08:14] huh, trying to dist-upgrade a machine and I get [08:14] The following packages have unmet dependencies: indicator-network : Depends: unity8 (>= 7.82) but it is not going to be installed [08:15] claims some pkgs are held.. [08:34] tjaalton: you don't happen to have -proposed enabled? [08:40] nope.. === Ursinha is now known as Ursinha-afk [09:11] hrm, jenkins output files don't have a proper mimetype so ffox offers to just save them [09:11] shows them as BIN === Ursinha-afk is now known as Ursinha === davidcalle_ is now known as davidcalle [09:17] xnox, around? we seem to be hitting issues with MongoDB on ppc64el due to the change in PAGESIZE [09:17] xnox, wondered if you could help me understand/fixup [09:43] infinity: will look into eigen today [09:46] Mirv: likewise I don't understand qtchooser, but extending an existing section in cmake fixes it for me - https://bugs.launchpad.net/ubuntu/+source/cmake/+bug/1294186 [09:46] Launchpad bug 1294186 in qtbase-opensource-src (Ubuntu) "Cross-building broken in qt 5.2" [High,Triaged] [09:46] ;-) [10:21] jamespage: i'd redirect you to apw =) ppc64el has large pagesizes than any other platform. [10:21] jamespage: db5.3 now does it right, but it is a bit vodoo to me. [10:30] *sigh* https://bugs.launchpad.net/launchpad/+bug/336439 [10:30] Launchpad bug 336439 in Launchpad itself "No notification on conflicts in merge proposals" [Low,Triaged] [10:54] Mirv: Sorry, looks like I forgot to bzr add some patches when updating that branch. [10:55] I will now push a commit that adds them. [10:57] mitya57: you could just re-merge with the _510 branch too where I added those [10:57] either way, it's already uploaded to the https://launchpad.net/~ci-train-ppa-service/+archive/landing-017/+packages [10:57] Ah, I see, you already pushed that === MacSlow is now known as MacSlow|lunch [11:37] I just setup a bzr mirror of the python-apt git repository (debian/sid) branch at lp:~juliank/python-apt/debian-sid, maybe that's useful for someone. Note that it does not contain up-to-date mirror lists, those are created by running pre-build.sh [11:43] hey tkamppeter_ - how are you doing? [11:44] tkamppeter_, a friend in the office where I go asked me about https://bugs.launchpad.net/ubuntu/+source/poppler/+bug/537970?comments=all - it seems like his company has a customer who would benefit from the bug being fixed - do you know what the state of things is there? [11:44] Launchpad bug 537970 in poppler (Ubuntu) "Evince does not print images in some pdf files." [High,Confirmed] [11:47] hi everyone, how can I make plasma widget Deb package? it's cmake based, and I don't know how to add cmake parameter in debian/rules file === tkamppeter_ is now known as tkamppeter === alai_afk is now known as alai [12:56] dholbach, hi [12:57] dholbach, you go to an office? Do you have a second job? [12:58] dholbach, is Rouven Sacha (rouvensacha) your contact concerning this bug? [13:05] Did the behavior of grep change between 13.10 and Trusty? I used to use “ps aux | grep thunder” to tell if Thunderbird is still running, but now that command just hangs [13:05] And it’s not the ps part that’s hanging [13:06] mpt: that is so crazy that it makes me think you got rooted and the kit replaced grep. [13:06] “touch ~/test && grep ~/test foo” also hangs [13:06] no changes at that level in grep [13:06] strace grep? [13:06] mpt: that second example is backwards. you mean grep foo ~/test [13:07] Aha! [13:07] but it shouldn't hang anyway, you should've got an error [13:07] Yeah, “grep foo ~/test” exits immediately [13:08] and that doesn't pertain to your first example [13:08] what does 'type grep' say? [13:08] grep is hashed (/bin/grep) [13:08] ok, so no funny business with aliases [13:09] I agree with sladen, the next thing I'd reach for would be ps aux | strace grep thunder [13:10] open("/usr/share/locale-langpack/en/LC_MESSAGES/grep.mo", O_RDONLY) = -1 ENOENT (No such file or directory) [13:10] fstat64(0, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0 [13:10] ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or TCGETS, 0xbfdb2a28) = -1 ENOTTY (Inappropriate ioctl for device) [13:10] read(0, [13:10] mpt: what's 'type ps' as well? [13:10] ps is hashed (/bin/ps) [13:10] "(Inappropriate ioctl for device)"? [13:11] mpt, do you use lxc? I wonder if you have an issue similar to the one Laney had recently [13:11] random guessing [13:11] seb128: mpt does use lxc a lot. [13:11] Where “a lot” = “1 hour per week” [13:11] seb128: no, successful runs have that same thing [13:12] that's just grep checking for the terminal size and finding it's not a terminal [13:12] cjwatson, ok [13:12] "read(0," means it's trying to read from stdin and getting no input [13:12] so I think I dispute the assertion that ps is not hanging [13:12] or at least would like to see further evidence there :) [13:13] “ps aux” by itself completes in real 0m0.028s [13:13] no input> more precisely, read on its standard input blocks, which means that the other end hasn't written anything (or not enough to fill a PIPE_BUF) and also hasn't closed its end [13:13] maybe pastebin strace -s1024 -f sh -c "ps aux | grep thunder" [13:14] “ps aux > ~/test && grep thunder ~/test” also works fine [13:14] so we can see the whole pipe setup [13:14] possibly better bash -c there [13:14] assuming you use bash as your interactive shell [13:14] cjwatson, sorry, where would the “bash -c” go? [13:14] at the start? [13:14] strace -s1024 -f bash -c "ps aux | grep thunder" [13:16] That’s more than my Terminal buffer [13:17] ok, got it now [13:19] cjwatson, http://paste.ubuntu.com/7119748/ [13:19] * cjwatson looks [13:21] pipe([3, 4]) = 0 [13:21] ... [13:21] [pid 26509] close(4 [13:21] [pid 26509] <... close resumed> ) = -1 EBADF (Bad file descriptor) [13:21] wtf === _salem is now known as salem_ [13:21] oh, no, it had already closed that one [13:23] mpt: so that seems to work under strace ... [13:24] and nothing seems out of order [13:24] Meanwhile, it has been running in another terminal for 7 minutes now and hasn’t exited [13:25] tkamppeter, yes, he is - it's a coworking space - I just call it "the office" :) [13:25] infinity: at the very least I'd like KDE upstream to weigh in on eigen3 migration [13:25] mpt: gdb -p pid_of_running_process and then ctrl-c inside GDB (if needed) [13:25] maybe find the pid of that running grep, ls -l /proc/THAT-PID/fd/0, and try to find that in other /proc/*/fd/ [13:25] mpt: You could attach strace to the running grep. "strace -p " [13:26] see what other process is holding the same pipe open [13:26] dholbach, so you do not work at home but rented a place in an office (the "coworking space")? [13:26] tkamppeter, sometimes like this and sometimes like that [13:26] :) [13:26] shadeslayer_: I'd suggest just checking to see if Debian's patch is merged upstream and if it, we go for it. [13:26] soren, forgive me, the only way I know how to find the pid of a process is by running grep on ps in the first place [13:26] ofcourse, if it's merged then it's alright [13:27] Wait, that’s not true, I can use Find in a text editor [13:27] ScottK: infinity I also recall seeing some projects switching to eigen3 in 4.12.80, though I can't recall the name [13:27] mpt: LOL. Right, of course. My bad :) [13:27] shadeslayer_: If it's not merged, then ask the Debian people why they haven't upstreamed it. We ought to get it in 4.13. [13:27] dholbach, the bug seems to be fixed in the current version of Ubuntu. Is your co-worker sticking to an older LTS? [13:27] *nod* [13:27] ScottK: I'll have a look at it [13:27] tkamppeter, precise is the current LTS [13:27] infinity: ^^^ my opinion [13:28] Process 26532 attached [13:28] read(0, [13:28] soren, ^ well, that was unexciting [13:29] dholbach, the problem seems to be Poppler when I look at the bug report but it is not very clear herewhich change on Poppler fixed it. [13:30] mpt: Does this happen only with ps as the input? [13:31] cjwatson, “lr-x------ 1 mpt mpt 64 Mar 19 13:29 /proc/26532/fd/0 -> pipe:[9579047]” … And I don’t understand the second part of your suggestion, sorry [13:31] mpt: ls -l /proc/*/fd/* | grep 9579047 [13:31] tkamppeter, ok - so it still needs further investigation, I guess? [13:32] soren, “ps aux > ~/test && grep thunder ~/test” also works fine [13:32] mpt: Hm... There I go, using grep again. [13:33] I would redirect it to a file anyway, because you're going to want to track down the owning process [13:33] so ls -l /proc/*/fd/* >all-fds 2>/dev/null; less all-fds and then look for 9579047 [13:33] cjwatson: ls -l /proc/*fd/* will show the full path names, so the owning process hsould be readily available. [13:34] soren, “ls -l /proc/*/fd/* | grep 9579047” has not exited after a minute (I assume it’s quicker than that) [13:34] It should be nearly instantaneuos. [13:34] I've found this https://wiki.ubuntu.com/Packaging/Training/Logs/2009-06-18 and it's what I want to do, but I can't find the debian/* example files [13:34] (It's all memory operations) [13:35] mpt: redirect to a file to avoid running into whatever the same issue is [13:35] mpt: So yeah, what cjwatson said. Pipe it to a file, and use lss. [13:35] this is sounding like a broken shell to me, but it's hard to tell ... [13:35] cjwatson, ok, there are three lines with that [13:35] you'd get this if the shell failed to close its ends of the pipe it's constructed [13:36] cjwatson, http://paste.ubuntu.com/7119825/ [13:36] mpt: ps -p 26530 -o pid,args [13:37] PID COMMAND [13:37] 26530 bash [13:37] ok, so as I thought [13:38] not that this gets us to a solution, since I'm running current bash in trusty and am not seeing this [13:38] mpt: Do you have another user on this system? [13:38] soren, no [13:39] mpt: How about the guest account? [13:39] soren, it exists but nobody is logged in to it [13:40] mpt: Can you see if it happens if you log in to it? [13:40] (according to indicator-session, at least, and indicator-session never ever lies) [13:40] ok [13:41] can someone remove mesa from proposed? didn't mean to get uploaded to archive [13:43] seb128: ^? [13:44] Laney, mlankhorst: is that a "need more testing but might be ok" thing, e.g proposed block would be enough? [13:44] no [13:44] ok, deleting it then [13:45] it's something that should have gone to a ppa so I could run some piglit tests [13:47] soren, it works fine in the guest session. [13:47] does ubuntu auto reject versions with ~ppa in the name? [13:48] Meanwhile, I can no longer move my mouse pointer in *this* session. [13:48] Laney, mlankhorst: done [13:49] ty [13:51] mpt: Interesting. [13:53] I've done: dpkg-depcheck cmake -DCMAKE_INSTALL_PREFIX=`kde4-config --prefix` -DQT_QMAKE_EXECUTABLE=/usr/bin/qmake-qt4 [13:54] and it printed several depend package [13:54] * mpt tries to log out and crashes unity-panel-service [13:54] Stop this operating system I want to get off [13:55] please, how to convert those command in debian/rules ? [13:57] miraiE: List those packages in Build-Depends field of debian/control [13:58] mitya57: okay, I'll try === marcoceppi is now known as marcoceppi-mobil === MacSlow|lunch is now known as MacSlow [14:06] cjwatson, so should I report a bug on bash? [14:09] mpt: that's my best guess [14:30] dholbach, I have asked the person for more info on the bug. [14:30] cjwatson, ok, reported bug 1294678, thanks for all the instructions [14:30] bug 1294678 in bash (Ubuntu) "grep waits forever for piped input" [Undecided,New] https://launchpad.net/bugs/1294678 [14:33] infinity: looks like debian has a load of patches to port kde bits to eigen 3 only some of which have been sent upstream and they're still under discussion for issues in the patches. is there a reason you want to switch or just don't like having two versions of same library? [14:35] (and thanks soren and sladen too) [14:36] jamespage, oh goody === retoaded_afk is now known as retoaded [14:38] pitti, would there be a chance to get rid of the xvfb dep in ofono-phonesim-autostart ? it cause quite some havoc in the touch tests (since the initial test setup installs it, so there is always an xserver running) [14:38] (we are just discussiong it in #ubuntu-ci-eng) [14:38] ogra_: ofono-phonesim is a graphical program, so it needs some X server [14:39] ogra_: it's Qt, so presumably at some day it could also use some "framebuffer Mir", but we don't have that yet AFAIK? [14:39] on touch ? [14:39] afaik we only use the sim emulation of it [14:39] and it uses Qt 4 still :/ [14:39] ogra_: yes, it's just one program [14:39] and use the dialer and messaging apps as frontends [14:40] ogra_: how does it interfere, OOI? It's running as a separate user on a private D-BUS with a private Xvfb, I thought it was shielded quite well [14:41] pitti, well, the constantly running xvfb kind of causes issues ... i would like that ofono-phonesim-autostart only gets installed for the two tests that actually use it, but apparently the current setup doesnt offer this ... so xvfb is contantly running [14:42] *constantly [14:42] ogra_: right; the whole -autostart packaging trick is because tests run as user and don't have root, so they can't set up phonesim by themselves [14:42] there are apps and tests that are running on desktop too ... so the running xserver can confuse them [14:42] tvoss, https://launchpad.net/ubuntu/+source/dbus-cpp/1.0.0+14.04.20140123.1-0ubuntu1 if this has to be GCC specific, please make it conditional on ppc64el, not having 4.7 [14:42] so back then, when we discussed that between fginther, awe, and me, we found that having the test depend on such an o-f-a package would be the best solution [14:42] (which then results in false negatives for failures) [14:43] tvoss, and what was the reason to fall back to 4.7? [14:43] doko, the platform-api is stuck with gcc 4.7 for the moment [14:43] pitti, right, but currently it isnt the app test that depends on it but the whole test run [14:43] ogra_: confuse how? it's a separate $DISPLAY, dbus, and everything; phonesim is only feeding ofonod, the "real" UI doesn't even know of its existance [14:43] ogra_: ah, we install all tests and their dependencies at the same time, and then run all tests? [14:43] ogra_: (for image testing, not for MPs) [14:44] tvoss, which ppc64el doesn't have [14:44] pitti, well, today we had the security test fall over because there is an active socket in /tmp/.X11-unix/ for example ... we had other issues before [14:44] doko, yup. So what do you want me to do exactly? [14:44] ogra_: so how does it confuse other tests? just because of the socket, or does it eat inordinate amounts of CPU/RAM/etc? [14:44] pitti, we run phablet-click-setup ... which installs all needed extra bits for click tests [14:44] phonesim is among them [14:45] while the fix is indeed to only have it for the tests that need it [14:45] tvoss: build depends: g++-4.7 [!ppc64el], and change the rules file to use gcc/g++ on ppc64el [14:45] i was wondering if there isnt a way to also decouple it from X completely [14:45] ogra_: I know how to make the socket invisible (we could run phonesim in an unshared namespace), if it's only that [14:45] (which is why i pinged you) [14:45] doko, ack, on my list [14:46] ogra_: right, so we'd need to re-set the device for each test, which sounds expensive [14:46] tvoss, thanks [14:46] ogra_: do you think if we hide it "more" that would be ok? then we coudl still install and run all tests at once [14:46] yeah, the tests already take 5h ... adding more reboots will only add to that [14:47] bfiller: hello, are you around? [14:47] well, it doesnt feel right to have an xseerver run on the phone at all ... but yeah, hiding might already be an improvemment [14:47] xnox: yes [14:47] thanks tkamppeter [14:47] bfiller: can we make a gallery-app branch merge / release into the archive and click? [14:48] bfiller: i'm more interested in the click into the click store with this merge proposal https://code.launchpad.net/~barry/gallery-app/xnox-pkgresources/+merge/210877 [14:48] xnox: we're working on that, see line 15 of CI Train sheet [14:48] xnox: I can include that MR in the silo so it gets released with the rest [14:48] bfiller: is that for click, or in the archive upload, or both? [14:48] ogra_: well, yes, but that's an intrinsic requirement of ofono-phonesim ATM [14:49] right, i understand [14:49] xnox: would be for both, sergiusens or someone I think has to manually do the click release after the deb gets released [14:49] ogra_: of course that could in theory be changed to not show an UI and only be a D-BUS service, but it's not doing that ATM [14:49] * jdstrand notes that this is not needed for the security tests to pass [14:49] I'm making them less brittle [14:49] pitti, would be nice if you could put that on your TODO for later :) [14:49] bfiller, yeah; is that done? [14:49] sergiusens: no [14:49] bfiller, as in, did it get to trunk? [14:49] bfiller: i see. Yes, please include that merge proposal in the landing ( https://code.launchpad.net/~barry/gallery-app/xnox-pkgresources/+merge/210877 ) [14:49] sergiusens: we found issues [14:50] xnox: I'll include that in the silo then [14:50] xnox, bfiller \o/ [14:50] bfiller: it's no-code change, only autopilot testing code fix-up. [14:50] pitti, i'll nag CI to only have it installed for phone related tests anyway, so on low prio for you :) [14:50] xnox: ack [14:50] ogra_: hiding the socket, yes; rewriting phonesim, not that much :) (in fact, much of its functionality actually depends on Qt, like sending SMS messages) [14:50] should be fine [14:50] ok [14:50] ogra_: bug report appreciated for hiding the socket, with some details [14:51] xnox: although i'm not fond of the auto-pruning of empty files (e.g. __init__.py) [14:51] bfiller: sergiusens, barry and myself pretty-much give you 24/7 coverage to respond to any queries as we all can't wait for it to land =))))) [14:51] xnox, barry : just make sure that merges cleanly with https://code.launchpad.net/~artmello/gallery-app/gallery-app-multiple-bug-fixes/+merge/211154 [14:51] bfiller: let me confirm / check that. [14:53] bfiller: "All changes applied successfully." === marcoceppi-mobil is now known as marcoceppi [14:59] xnox: cool, I'll add it to the silo === marcoceppi is now known as marco-traveling [15:03] xnox: hi [15:03] xnox: for "dpkg-architecture -aarmhf cmake ../ && make" to work, i need to enable armhf as an arch on my machine, and then install the :armhf versions of all the build-deps? [15:05] dobey: almost, but that will most likely break a few things on your desktop. It's best to use a schroot. E.g. click chroot create / click chroot run, or mk-sbuild --target [15:05] dobey: for simply things (which don't use gl*/qt*) doing it on your desktop is safe as well. [15:05] dobey: are you already using mk-sbuild / sbuild, and what are the projects that you want to cross-compile? [15:06] xnox: i want to compile lp:unity-scope-click. i'm not using sbuild already, no (though i do have it insatlled) [15:08] dobey: test-building here, will let you know in a moment. [15:09] xnox: ok, thanks! [15:14] dobey: unfortunately automatic installation of cross-dependencies fails, as a few *accounts* *oauth* etc things are not Multi-Arch:same. [15:14] dobey: let me see if we can fix this. [15:14] ah, ok [15:14] cjwatson: have you already been "multiarchifying" accounts/oauth/signon things?! [15:14] i'll try sbuild i guess [15:15] xnox: some [15:15] xnox: I'm specifically working on making ubuntu-sdk-libs-dev:armhf installable on !armhf [15:16] dobey: well, yeah native compilation should work $ mk-sbuild --arch armhf; sbuild --arch *.dsc; will use qemu-user-static and will slowly natively compile. But i'll see if i can fix cross-compilation for that. [15:16] xnox: probably overlaps slightly with you but not much. The things I have in other people's landing queue are dee and libaccounts-glib [15:16] cjwatson: ack. thanks. I'll see which bits i can propose for landing. [15:18] xnox: eh, seems like sbuild will be the fastest option, and i need to do the building /now/, so i'll go with that [15:19] hopefully it doesn't break for some silly reason due to config differences between local sbuild, and what the launchpad builders are doing [15:19] Riddell: It's not even a shared library, it's a static (in 2) headers-only (in 3) set of stubs, so I'm not super concerned, I just ran into the "it's not even in Debian" thing when I was hunting why it was broken on ppc64el. [15:20] Riddell: Fixing it on ppc64el should be trivial, I didn't bother going that route while waiting on KDE/eigen options. [15:21] infinity: yep, I'll try and send that patch upstream today [15:21] xnox: ugh. "mk-sbuild --target armhf trusty" failed pretty horribly :( [15:23] xnox: http://pastebin.ubuntu.com/7120328/ [15:23] dobey: --skip-proposed should help cross-chroots. "--target" creates a "cross-compilation chroot". For qemu-user-static, you want "mk-sbuild --arch=armhf trusty" [15:24] dobey: not sure why it's using "archive.ubuntu.com" for you, as armhf is on "ports.ubuntu.com (us.ports.ubuntu.com). [15:24]