/srv/irclogs.ubuntu.com/2015/03/05/#ubuntu-desktop.txt

=== duflu_ is now known as duflu
=== mfisch is now known as Guest63892
pittiGood morning06:27
didrocksmorning07:29
happyarondidrocks: hey07:32
happyaronattente_: qt5 version is uploaded to archive07:32
didrockshey happyaron07:34
pittibonjour didrocks, c,a va /07:53
pittieek, what happened to my compose key?07:53
didrockspitti: ça peut aller, et toi ?07:53
pittididrocks: je suis fatigue', comme tous le jeudi apre`s basketball07:54
* pitti switches keyboard layout twice, and voilà - compose key !07:54
pittididrocks: and I'm fighting some issues with the latest cloud-init07:54
didrockspitti: oh, still following up from yesterday?07:56
pittididrocks: yeah, I just filed a bug for and worked around broken ssh07:57
pittididrocks: and I'm now looking at why rsyslog stopped working (systemd's tests fail on that)07:57
pitti(on the cloud images)07:57
didrocksurgh07:58
didrocksyeah, I imagine for cloud-init, rsyslog forwarding may be important :)07:58
seb128hey didrocks pitti08:00
pittibonjour seb12808:00
didrockshey seb12808:00
pittigo rsyslogd --  no manpage, no useful --help08:00
pittioh sorry, it does have a manpage08:00
pittiopen error 13, file '/var/log/syslog': Permission denied08:01
pittiah-haa08:01
pittithat's fun, that should affect every new installation where /var/log/syslog doesn't already exist..08:02
didrocksthe acl changes?08:02
didrocksoh?08:02
pittihah, seb128 filed that as bug 1401984, and we added a tmpfiles.d for it ages ago08:03
ubot5bug 1401984 in rsyslog (Ubuntu) "non persistent logging after cleaning log files on disk" [High,Fix released] https://launchpad.net/bugs/140198408:03
pittididrocks: no, the ACL is still there08:05
didrocksinteresting08:06
didrockspitti: btw, now, I'll use any opportunity to check if systemd-timesyncd is activated or not08:06
didrocksjust to know if that random starts condition only happens in a vm or not08:07
seb128woot, a systemd upload where the changelog is not the usual 3 screens long summary of the ubuntu delta :-)08:07
pittiseb128: oh, it's usually just one screen :)08:07
pittithe other two are the changes in Debian :)08:07
seb128hehe08:07
didrockspitti: it does try to start it on my system, and but as I already have /tmp as tmpfs, I didn't notice…08:08
pittididrocks: yeah, same for me I guess; but timesyncd seems to start reliably here08:08
didrocksthe condition failed is funny08:08
didrocks           ConditionFileIsExecutable=!/usr/sbin/ntpd was not met08:08
didrocks-rwxr-xr-x 1 root root 672624 févr.  9 19:12 /usr/sbin/ntpd08:08
didrockspitti: I don't understand yet why it becames a RequiresBy=tmp.mount though08:09
didrockswaow, here, it's required even by more things than in the vm08:10
didrocksRequiredBy=colord.service systemd-timesyncd.service local-fs.target rtkit-daemon08:10
didrockslocal-fs.target is because I guess it's in my fstab08:10
didrocksseb128: do you have /tmp on tmpfs?08:10
seb128didrocks, that would show in "mount" right?08:11
seb128I guess not08:11
pittididrocks: oh, so if you have ntpd it should indeed *not* start08:12
pittidoes it?08:12
didrocksseb128: yeah, it should. So systemctl status tmp.mount says it's not active? (and can you paste the RequiredBy line of systemctl show tmp.mount?)08:12
didrockspitti: right08:13
seb128didrocks, inactive indeed08:13
seb128● tmp.mount - Temporary Directory08:13
seb128   Loaded: loaded (/lib/systemd/system/tmp.mount; disabled; vendor preset: enabled)08:13
seb128   Active: inactive08:13
seb128RequiredBy=rtkit-daemon.service systemd-timesyncd.service colord.service08:13
didrocksso colord isn't running for you I guess08:14
didrocksneither rtkit-daemon08:14
seb128$ systemctl status colord08:14
seb128● colord.service - Manage, Install and Generate Color Profiles08:14
seb128   Loaded: loaded (/lib/systemd/system/colord.service; static; vendor preset: enabled)08:14
seb128   Active: active (running) since mer. 2015-03-04 07:52:18 CET; 1 day 1h ago08:14
didrocksok, I don't understand it seems :/08:14
didrockspitti: any idea? ^08:14
didrocks1. it's not an explicit Requirement08:15
pittididrocks: no idea, sorry08:15
didrocks2. it's listed as RequiredBy and the same layout starts sometimes tmp.mount in a vm (I was thinking via systemd-timesyncd.service)08:15
didrockspitti: yeah, I smell something really fishy…08:15
didrocksI guess we'll keep the "fail on no disk space" this cycle anyway08:16
didrocksthanks seb128 :)08:16
seb128yw08:16
pittididrocks: but timesyncd should never start in a  VM, btw (ConditionVirtualization=no); if you want to test it that way, drop that line08:17
didrockspitti: it never starts, but it tries to activate (and show, brings the RequiredBy=) at some boot08:18
didrocksso basically, it's either "inactive" or "condition failed"08:18
didrocksin the second case, it started tmp.mount08:18
didrocksand tmp.mount shows it in RequiredBy= even if we don't see it in the unit file08:18
didrocksbut in seb's case, it seems that even tmp.mount shows RequiredBy=colord.service (for also on unknown reason), having colord.service started didn't activate tmp.mount08:19
didrocksmy only bet is that colord is started via something else08:21
seb128it's of type dbus08:21
seb128so it's maybe dbus activated?08:21
seb128does dbus activiate respects the requirements?08:21
didrocksI guess so08:22
didrocksand that would explain at least this unmatched dep08:22
didrocksno08:22
didrocksit doesn't respect, I guess systemd just see a process and say "oh, it's that unit"08:23
didrocksso mark it as running08:23
didrocksI still don't know why those appears in RequiredBy=, there is nothing explicit, particuarly for systemd-timesync08:23
didrocks(we could think of PrivateTmp=yes for colord would transition to tmp.mount)08:24
seb128didrocks, what appears in what requiredby?08:24
didrocksseb128: $ systemctl show tmp.mount | grep Required08:24
didrocksRequiredBy=colord.service systemd-timesyncd.service local-fs.target rtkit-daemon.service08:24
didrocksif you systemctl cat (or open) unit files, they don't have a Requires=tmp.mount08:25
didrocksif you grep in /lib/systemd/systemd/*service, no sign of tmp.mount being mentioned as a requirements in any unit08:25
pittiman, git bisect run is so awesome08:39
larsuya. Feels like magic whenever I use it08:41
larsumorning pitti :)08:41
pittihey larsu!08:41
pittiyeah; start it, let your computer burn and smoke, and enjoy a Piña Colada :)08:42
pittior, in my case, work on NFS instead :)08:42
seb128is that different from bisect?08:42
seb128does it auto-flag good/bad and how?08:42
pittiseb128: it is a bisect, but in fully automatic mode08:42
seb128how do it determine the outcome?08:42
pittiseb128: you toss it a script, and looks at its exit code08:42
seb128k08:42
pitti0 -> good, < 125 -> bad, 125 -> untestable (e. g. doesn't build)08:42
pitti> 128: abort08:43
pittiso as soon as you have an automatic test case you can let it run08:43
pittiseb128: I did that for the boot hang: my test case was to run an autopkgtest in QEMU with rebooting 30 times08:43
larsu... and let it find your bug for you08:43
pittiso it went through ~ 1100 commits with that test case in about 2 hours08:43
pitti(now running it for the /var/log issue)08:44
larsupitti: watch out, you're dangerously close to replacing yourself with a script08:44
pittiI wish I could -- one of these days I *actually* want a Piña Colada!08:44
seb128lol, I wonder if https://plus.google.com/u/1/+MathieuTrudelLapierre/posts/bFdphmbgAz8 is real :-)08:44
seb128pitti, nice!08:44
pitticyphermox: ♥08:45
ochosilarsu: i vaguely recall you said you don't wanna work on that just yet, but i just patched a panel plugin for xfce to use symbolic icons, it's fairly easy and straightforward (if you'd be interested in doing that for indicators, i mean)08:45
larsuochosi: ah cool. It's not that much work, is it?08:45
larsuochosi: we'd need to merge ubuntu-mono-dark and -light first08:46
larsuwhich is something I've been looking at again yesterday when I added window decoration buttons for both08:46
ochosilarsu: this is pretty much it: http://git.xfce.org/panel-plugins/xfce4-pulseaudio-plugin/commit/?id=314286b6f7da37472dcc69adfe64b66c07d580ec08:46
ochosiand no, not necessarily08:46
ochosijust ship the symbolic variants in the parent theme08:46
ochosithe -dark variant will inherit automatically08:47
larsuochosi: the parent theme for both is humanity iirc08:47
larsuhm, why do you preload icons?08:47
ochosilarsu: oh, that's just how that particular plugin was written08:48
ochosii didn't want to change anything there apart from the icons08:48
ochosiit preloads all icons at once and then uses the appropriate one per volume08:48
ochosinot sure whether that's more efficient than loading what you need at the time08:48
larsushouldn't make that big of a difference08:49
ochosiprolly not08:49
larsuicons are mmapped anyway08:49
larsuall you do is waste some memory08:49
larsubut it's not a lot, so whatevs08:49
ochosiyeah, it's fairly little08:49
ochosiso basically what i'd propose is start with indicator-sound as a testcase08:50
ochosiuse the symbolic icons for the panel button and the menu icons08:50
ochosi(which will be nice because the icons will finally *always* be recolored correctly with all themes)08:51
larsuochosi: indeed08:51
larsubut!08:51
larsuindicator-sound colors it's icon sometimes08:51
ochosiyeah08:51
larsuwe'd need to add a css class in that case and add that to the theme08:51
ochosisymbolic icons obviously account for that08:51
larsu*themes08:51
ochosinope08:51
ochosiwell, at least not really08:51
ochosias long as the symbolic icons have the error state, you should be fine by just setting that state08:52
larsuthis only works for one state, though, and I'm not sure that e.g. Ambiance and Radiance use the same color there08:52
larsuor is that state colorable from the theme as well?08:52
ochosiit's "success_color", "warning_color", "error_color"08:53
ochosihttps://developer.gnome.org/gtk3/stable/GtkIconTheme.html#gtk-icon-info-load-symbolic08:54
ochosiafaik that's all in the icon, nothing required in the theme08:54
larsuochosi: ah, those *are* css styles08:56
larsuans we already do style them08:56
larsu*and08:56
ochosiyeah, sry, but in fact you have to set a class within the icon08:56
larsuright08:57
ochosiso it's both, the theme sets the color and the icon has to have the class set on the paths that should be recolored08:57
ochosiso basically for the blocked stuff you just copy an existing symbolic icon and set the class of all paths to "warning"08:57
larsuso we'll still need to set the corresponding class on the gtkimage I guess08:57
ochosinot sure, no08:58
larsuhow else would it know when to recolor?08:58
ochosiit automatically looks inside the svg and sees the class08:58
ochosiat least that's how i think it works08:58
larsuyeah but it needs to know _when_ to draw that part08:58
ochosiso you have "audio-volume-blocked-symbolic" or something08:58
larsuoh08:59
larsuso we'd need to change the icon then08:59
larsuya, could do08:59
ochosiand that has the "warning" class set08:59
ochosiso all themes would color the icon accordingly08:59
ochosiso it's not about re-using existing icons, although that might be possible as well08:59
ochosisimply by adding a class08:59
ochosiand removing it conditionally from the style contexxt08:59
larsuyeah what you're describing makes more sense to me08:59
ochosiso the icons remain the same, i think the -blocked is already a separate icon09:00
larsuochosi: this is early-next cycle work though - don't want to mess with that now09:00
ochosiyeah, sure09:00
ochosiif you wanna start the transition and need some help lemme know09:00
ochosiwould be good to do that for all indicators09:01
larsuindeed, it would09:01
larsuochosi: thanks! Will do09:01
ochosiokeydokey :)09:01
larsuthanks for pointing me towards this09:01
ochosinp, btw, you might want to talk to gnome upstream and see whether they see any reason to include a "blocked" icon in adwaita-icon-theme09:02
ochosithat way this would sort of become a standard09:02
ochosi(since nobody seems to care about the fd.o icon spec anymore these days and gnome is the defacto standard for icon-names)09:02
ochosi(nobody loves fd.o anymore :'( )09:02
Laneyyo09:04
larsuochosi: there's quite a lot of stuff hosted on fd.o and we've been adding new stuff to the specs09:06
seb128hey Laney09:06
larsuochosi: icons is a bit failed...09:06
larsumorning Laney!09:06
seb128Laney, wants to update gtksourceview3 in debian/sync the update? ;-)09:06
larsuyes!09:07
larsuhe wants!09:07
seb128that should fix bug https://bugs.launchpad.net/gedit/+bug/142833309:07
ubot5Ubuntu bug 1428333 in gedit (Ubuntu) "gedit crashes on autocomplete with SIGSEGV in gtk_text_iter_get_buffer()" [High,Confirmed]09:07
ochosilarsu: yeah, i was specifically referring to that spec, not all of them are as outdated as the one for icons09:07
seb128well combined with https://git.gnome.org/browse/gedit/commit/?id=d48cd10bf0bb99ead7b1af1ead52f1e554ad239d09:07
didrocksmorning Laney09:08
Laneyhey seb128 didrocks larsu09:10
Laneyumm yeah okay09:10
seb128danke09:10
larsuLaney: that commit depends on gtk 3.16, so please upload that as well09:10
seb128lol09:10
* larsu <-- on troll day. Won't be here tomorrow09:11
seb128which depends on bluez5, so while you are at it...09:11
larsuHAHA09:11
larsunice one09:11
ochosilarsu: btw, how did you fix the scale grabbing in indicator-sound?09:11
seb128:-)09:11
seb128larsu, want to review https://code.launchpad.net/~3v1n0/libdbusmenu/custom-stock-item-label/+merge/251840 ? it's a one liner09:12
larsuseb128: yes, a long time ago?!09:12
seb128larsu, what a long time ago?09:12
larsuseb128: oops, this was supposed to go to ochosi09:12
seb128larsu, was that for ochosi? he asked "how"09:12
* larsu can't read, and can't type09:13
ochosilarsu: wait, but that happened during this cycle, no? i remember not being able to click the volume scale correctly09:13
larsuochosi: it did. I think the issue was that something reset the volume while you were dragging it09:14
seb128larsu, ochosi, http://launchpadlibrarian.net/195548750/ido_13.10.0%2B15.04.20141103-0ubuntu1_13.10.0%2B15.04.20150122-0ubuntu1.diff.gz09:14
seb128that one?09:14
ochosithat looks like it could be it09:15
seb128https://code.launchpad.net/~larsu/ido/fix-volume-slider/+merge/24719809:15
seb128is the corresponding mr09:15
larsuindeed, thanks seb12809:16
seb128yw09:16
ochosiawesome, thanks seb12809:16
seb128yw09:16
larsuthe volume resetting stuff was about the notification popping up09:16
ochosiso is that a change that is gtk3 version specific or anything?09:16
ochosicause i remember it broke with gtk3.14 in the indicator09:17
larsuochosi: yes. I remember not bothering to figure out the actual change that caused the issue, but there was quite some work done in gtk on event handling09:18
larsuall I did was not propagating events from the menu item to the scale in all cases09:18
ochosiyeah, i wonder whether that works on older gtk3 releases as well09:19
larsuochosi: don't try?!09:19
ochosiheh09:20
ochosiwell, tbh for the plugin that i linked earlier we copied the indicator's scale menu item09:20
larsuochosi: port to popovers!09:20
larsuit's the future09:21
ochosiyeah, we haven't decided about those things yet09:21
ochosibasically that needs a global xfce decision09:21
larsuright, makes sense09:21
ochosiand we've only just released 4.12 last weekend09:22
ochosiso my work is more or less exploratory at this stage09:22
larsuseb128: the bug linked to this MR is WontFix by both tedg and mpt...09:23
larsu(the dbusmenu issue)09:24
larsuochosi: yeah I saw. Congrats on 4.12 :)09:24
larsuseems people like it09:24
seb128larsu, feel free to reject the change09:24
ochosilarsu: hah, that is a very diplomatic statement "some ppl like it" :) i'm just happy we could finally move out of the stalling and now there's a clear goal (port to gtk3)09:25
larsuochosi: I didn't say "some" ;)09:26
ochosioh ooops :)09:26
* ochosi wonders why/how he managed to misread that09:26
larsuI read some reviews and they were mostly positive09:26
larsuthat's what I meant09:26
larsuhaven't tried it myself yet09:26
ochosiyeah, no dramatic changes, which is generally an xfce policy09:27
larsubut yeah, I'm glad you guys are releasing again :)09:27
ochosiyeah, took a while and a lot of personal effort09:27
larsuoh, mpt reopened that bug...09:27
larsuman, such a non-issue09:27
larsuochosi: I can imagine09:27
ochosilarsu: btw, i'm not sure you're interested in this sort of redesign, but i had an idea for indicator-sound that would rid you of the mute menuicon09:37
ochosilarsu: basically, only have a volume icon on the left (the one on the right is a bit superfluous really, i mean what's that scale gonna do? decrease the volume?) and make that a button instead of an icon which on click mutes09:38
ochosilarsu: well un/mutes obviously09:38
Sweet5harkmoin09:39
larsuochosi: both icons already do something (move the slider all the way in that direction)09:39
larsuochosi: I personally dislike having mute separately from volume, but this is an issue to take up with mpt. He designed the sound menu09:40
ochosilarsu: oh right, i never even noticed that09:40
larsuochosi: https://wiki.ubuntu.com/Sound09:40
ochosiguess some hover-effect on the icons would help users to notice that09:40
Sweet5hark... feeling trollish today, I have to note that larsus Berlin and my home of Hamburg both moved up in the 2015 imercer quality of living ranking, kicking out Canadians and kiwis.09:41
ochosilarsu: so it's actually less of a departure from the current state, mostly adding a feature where clicking that icon again will restore the previously set volume...09:41
larsuochosi: feel free to file a bug for i-sound and assign mpt (or the ux team)09:41
larsuyeah, that would be weird though09:41
larsuSweet5hark: economist ranked toronto the most livable city09:42
ochosinot if there's only one icon09:42
mptDon’t assign teams to bugs, real people don’t see them09:42
larsuSweet5hark: and really, I trust those guys more :)09:42
larsumpt: sorry. And hi!09:42
mptGood morning :-)09:43
ochosimpt: does that idea even make sense to you? i guess there's no use in writing something up when you already know that's not what you want09:43
Sweet5harklarsu: you should move to Hamburg then, economist ranks them first in Germany (Top 14 globally) followed by Frankfurst as next german city ;)09:44
larsuSweet5hark: haha, you're funny.09:45
larsu:O09:45
larsuSweet5hark: seriously though, I totally believe that Hamburg is more livable than Berlin09:45
larsuthey look at all kinds of factors09:45
Sweet5harklarsu: this is why our rents and real estate prices are where yours are only slowly going to ...09:46
larsuSweet5hark: actually, not so slowly these days...09:46
larsuSweet5hark: but then, we have quite some catching up to do compared to HH09:47
* larsu should visit there properly09:47
larsunever really been to the city09:47
mptochosi, sorry, I just read back 60 minutes but don’t see any bug or MR reference … What’s the idea?09:47
larsumpt: remove mute in favor of the buttons next to the volume slider09:48
ochosimpt: yeah, no MR/bugreport yet, i basically proposed dropping the mute menuitem, only using one sound icon to the left of the scale and using that as mute-button. adding a hover-style to that icon would help users see what it does09:48
flexiondotorg_The MATE team have found a bug in upower that affects mate-power-manager.09:58
flexiondotorg_Fedora has a simple patch that is confirmed as working.09:58
flexiondotorg_https://bugs.launchpad.net/ubuntu-mate/+bug/142833709:58
ubot5Ubuntu bug 1428337 in ubuntu-mate "MATE power statistics crashed" [Undecided,Triaged]09:58
flexiondotorg_Could someone take a peek?09:58
larsuflexiondotorg_: sure. Please add upower to the bug and link an upstream bug (create one if it doesn't exist)09:59
flexiondotorg_larsu, Thanks.09:59
pittior rather, move it to upower, doesn't seem to be m-p-m's fault10:00
pittiflexiondotorg_: I'll have a look later and commit it upstream (but still debugging some stuff)10:00
mptochosi, larsu: This is going back far enough that I need to refer to what I wrote at the time to remember why I did stuff. <https://wiki.ubuntu.com/Sound#A.2BIBw-Mute.2BIB0-> says ‘The first item, “Mute”, makes it easy to urgently silence unwanted sound using a pointing device.’ Just the tiny button to the left of the slider wouldn’t be clickable nearly as quickly.10:04
mptEspecially (but not only) while bug 552920 remains unfixed.10:05
ubot5bug 552920 in unity (Ubuntu) "Moving diagonally from narrow menu title often opens adjacent menu" [Medium,Triaged] https://launchpad.net/bugs/55292010:05
flexiondotorg_pitti Also affects gnome-power-manager10:06
pittiwhy?10:07
pittiis there another bug which needs to be fixed there?10:07
pittilooks like a crash in libupower-glib10:07
larsumpt: there are no submenus involved there...10:08
mptochosi, larsu: The only reason the volume slider icons do *anything* when clicked is that I thought, “well if someone *did* click them, what would they be expecting to happen?” — not because I expected anyone to use them as the primary way of doing something.10:08
flexiondotorg_pitti, I am just linking to where that bug has been experienced.10:08
mptlarsu, eh? I didn’t mention submenus :-)10:08
larsumpt: oh! I thought the bug you mentioned was about that, but it's about menut titles. Sorry10:09
larsumpt: makes more sense now :)10:09
larsumpt: I like https://wiki.ubuntu.com/Sound#Mute, which is basically what the left menu item now does10:10
larsubut I don't think it would be well received if we switched this now10:10
larsuespecially since many laptops have a "mute" key with an led which is on while the sound is muted10:11
mptIs that software-controllable?10:12
mptFor convergence!!!!1 I guess we’ll need to change one or the other eventually :-)10:13
larsuhaha, indeed10:13
larsuI don't know, it's probably not controllable on all machines10:13
ochosimpt: yeah, understandable, and i wouldn't propose this as a change for this cycle anyway10:25
ochosimpt: and i guess i'd only have one icon but bump the size of that to 22 or 24px, so it's more easily clickable10:27
mptochosi, that would increase the height of the item as a whole, and I’m not a fan of menu items of slightly varying heights — it looks accidental10:30
ochosiwhat about the playback controls then? :]10:31
mptThose (a) are massively different, not slightly, and (b) have borders around them clearly indicating their non-standard-ness :-)10:32
ochosisure, makes sense10:34
Laneydid anyone check the gnome-terminal wrapper?10:41
ochosilarsu: there is something wrong with your commit or the idoscalemenuitem still. if you try to hover the scale-button, it isn't drawn in hover style. however, if you hover approx 50px to the left of the scale-button it gets the hover-style applied10:47
ochosinoticed when porting your change to the plugin i'm working on10:47
larsuindeed :(10:48
* larsu makes a note of fixing this10:48
* Laney gets a keyboard layout debconf prompt on latest dist-upgrade10:49
Laneycccccccyyyyyphhhhhhhherrrrrrrmooooooooooooxxxxxxxxxxxxxxxxxxx10:49
didrocksLaney: opportunity for you to switch to a french layout!10:51
LaneyI 'ate french!10:51
didrocks:p10:51
larsuLaney: don't be so 'ateful10:52
* Laney fluffles all French #ubuntu-desktoppers10:52
didrocks /msg seb128 when did we discuss last time we would lay off a non french person in the desktop team?10:53
didrocksoups ;)10:53
* Laney looks at mlankhorst 10:53
didrocksroh, he left, that doesn't count!10:53
larsuhaha lol10:54
* Laney slinks away :(10:54
mlankhorstyou can still fire me :P10:54
Laneyhey mlankhorst10:54
Laneydo you have to move somewhere for intel?10:54
mlankhorstno, still working from home10:54
Laneynice, didn't know that they did this10:54
mlankhorstthey only do it when you were already working on something related from home10:54
didrocksseems Laney is getting infos for his next job :p10:55
mlankhorstnothing wrong with that10:55
Laneyheheh10:55
didrocksthis is sad, even no willcooke for teasing!10:55
Laneywhen the cat is away10:55
mlankhorsthttp://thedailywtf.com/articles/Up-or-Out-Solving-the-IT-Turnover-Crisis :P10:56
Laneyhaha10:56
LaneyI see what you're calling yourself10:56
mlankhorstnaww but it's going to be hard to top Xmir, and X is on the way out :(10:57
mlankhorstI'm investigating on how to make GLES2 work on X in arm, it's harder than desktop :P10:58
mlankhorstafk a bit11:00
didrockspitti: would you suggest to use ID_FS_TYPE to get a file system type (I don't use udev in the script, so I will just build a new udev_device_new_from_devnum for this), or is there any other way that is easier?11:24
pittididrocks: if you don't use udev already, calling blkid might be easier11:24
didrocksok, by directly getting the TYPE tag11:25
pitti$ blkid -s TYPE -o value /dev/sda311:25
* didrocks looks11:25
pittibtrfs11:25
didrocksthanks pitti :)11:25
pittididrocks: ah, Lennart's recent answers clarify that quite a bit?11:31
pittie. g. colord also has PrivateTmp=yes11:31
didrockspitti: yeah, that was what I was infering here11:34
didrockspitti: but so, if any of those runs, it means that users will have tmpfs by default on vivid11:35
pittioh, what a disaster :)11:35
didrockspitti: so, either was mask the unit by default, either we enable it for everyone, not feeling well to have different experience here in case of bugs, wdyt?11:36
pittididrocks: can we discuss that in #d-systemd?11:37
didrockssure11:37
MacSlowTrevinho, seb128: hey folks... have you perhaps seen a glitch with current vivid (desktop) like this yet: unity/compiz reacts sluggish once qmlscene is started to run a .qml file?12:17
MacSlowTrevinho, seb128: just wondering if it's just on my system or a general issue12:18
=== alan_g is now known as alan_g|lunch
=== MacSlow is now known as MacSlow|lunch
Laneyblah12:56
Laneylost my compose key12:56
Laneynow I can't type my gpg passphrase!12:56
Laneyrestarting u-s-d fixed it13:07
=== alan_g|lunch is now known as alan_g
Trevinhomlankhorst: mh not saw that..13:57
Trevinhoerr... sorry, I meant MacSlow|lunch ^13:57
TrevinhoMacSlow|lunch: nothing outside the qmlscene itself, which is something qt related13:58
MacSlow|lunchTrevinho, I'll still have to try it on my laptop (intel gfx) to have a comparison to my desktop (nvidia, binary blob driver), where I've seen this13:59
MacSlow|lunchTrevinho, unity's app-switcher was also affected... being very slow and sluggish to respond to a first invocation14:00
Trevinhomh14:00
=== MacSlow|lunch is now known as MacSlow
* Trevinho has no nvidia to check with14:00
MacSlowTrevinho, I'll get back to you after I tried it on my laptop14:00
Trevinhoandyrock has one btw ^14:01
MacSlowTrevinho, andyrock: hm... no issues on my laptop (intel gfx)14:07
cyphermoxseb128: pitti: yes, we can "Spock" our old five dollar bills by drawing hair and such on John A. Macdonald. Legal, but not very appreciated ;)14:08
pittihaha14:09
cyphermoxpitti: btw thanks for fixing NM autopilot :)14:09
pitticyphermox: it's autopkgtest, but you're welcome :)14:09
cyphermoxuh, right14:10
Laneymardy: hi, are you planning to drop windows live (msn) messenger support from $uoa_component?14:46
mardyLaney: mmm... no, should I?15:05
Laneymardy: it's gone now15:06
mardyLaney: you mean that Live doesn't exist anymore?15:06
LaneyThere's some dodgy patch to use a not shut down server, but basically yes15:09
Laneyhttp://ismsndeadyet.com/15:15
* didrocks doesn't understand the open() behavior on an overlay fs15:44
Laneymardy: ↑ forgot to highlight you15:45
desrtdidrocks: which overlay fs?15:54
desrtaufs? overlayfs? other?15:54
desrtit changes depending on the FS15:54
didrocksdesrt: overlayfs15:54
desrtoverlayfs is the weirdest one15:54
didrocksdesrt: basically, I try to detect on which fs type a file is (for a systemd patch)15:54
desrtif you open() a file and there is nothing in the overlay then it will open() the file in the underlay15:54
desrtwhich is weird15:54
didrocksyeah, that's what I see15:55
desrtsince if you then open() and trunc the file from another process, the person who opened it will still see the underlay file15:55
desrtthey do it in the name of performance15:55
desrtsince the 99% case is that there is no overlay file15:55
didrocksyeah, my issue is different and probably in the open options I'm using15:55
mardyLaney: thanks; I guess it's a matter or editing the rdepends of account-plugin-windows-live, and remove it as a dependency?15:55
desrtand keeping around all of those extra files the kernel (and constantly doing checks to see if the file has been created each time you access it) is very expensive15:55
didrocksdesrt: for using blkid to get the fs type, I first need to have the device name, and so, I'm getting the devnum from the fd of that file (and so, I open() it)15:56
desrthah.15:56
didrocksdesrt: until I touch the file, open() works well15:56
Laneymardy: I don't know exactly what'd need changing, but yes I'm sure it's mostly deleting15:56
didrocksand I get the squashfs type on the live15:56
desrtst_dev is completely screwed up on overlays...15:56
Laneymardy: We'd also want to SRU this to 14.04 at least I think15:56
didrocksdesrt: once I modify the file, open() returns an < 0 fd15:56
desrteh?15:57
desrtyou mean to say, it fails15:57
desrtpresumably with some helpful hint in errno15:57
didrocksyeah15:57
didrocksell15:57
didrockswell15:57
didrockshelpful being "No such file or directory"15:57
desrtare you using openat() on the dir or something?15:58
didrocksdesrt: oh screw that, it's the probe failing15:58
desrtheh15:58
didrocksdesrt: I was on the original script to check the error message15:58
didrocksso yeah, due to the bad st_dev15:58
didrockshum…15:58
didrockshow can I know then I'm on an overlayfs system?15:59
desrtthe 'usual tricks' go out the window here15:59
desrtie: traversing the chain up and comparing st_dev at each point15:59
desrti think the best mechanism it to consult /proc/mount15:59
didrocksurgh, ok :p15:59
desrti'm planning to make similar changes in glib, fwiw...16:00
desrtoverlays are kind shit for this reason16:00
* desrt can't type today16:00
didrocksdesrt: basically systemd got tricked because to check if a file is a mount point, it checks the file device block, then the dir device blocks16:00
didrocksand of course, they are different (as for dir, they are 0)16:00
desrtya... that's what you're supposed to do16:00
desrtat least in the old world :)16:00
desrtnow i think you'll just have to check /proc/mounts16:01
didrocks"nice" :)16:01
desrtthis is sort of what led me to my recent work on the mount monitor code in glib16:01
desrtwe have some serious pain coming in the next couple of years as overlays-in-containers get more popular16:02
desrtand there is still a lot of work to do there16:02
didrocksright, I bet this is not the only "exception" due to overlays16:02
desrtanyway... i hope you're enjoying yourself :)16:03
didrocksdesrt: totally, it's yet-another-supposively-10-minutes-hack-which-won't-be-one :)16:03
desrtheh.  welcome to my life :p16:04
desrtnothing can ever be easy -- otherwise someone else would have done it by now :p16:04
didrockspitti: FYI btw, (context: bug #1411140) ^16:04
ubot5bug 1411140 in systemd (Ubuntu) "systemd-machine-id-commit.service fails on live system" [Low,Triaged] https://launchpad.net/bugs/141114016:04
didrocksdesrt: heh, yeah, I see. Didn't I tell you I prefer using glib than having to code at the same low level without it? :)16:04
pittiglib! glib! glib!16:05
desrtdidrocks: i'm afraid today's glib won't help you a lot here16:05
desrtwell, the mount monitor stuff would help, i guess16:05
didrocksdesrt: yeah, in that particular case, agreed :)16:05
desrtsince it provides a nice wrapping around /proc/mounts16:05
desrtbut this isn't saving you the 'hard work' really16:05
desrtalso: it's not very efficient today16:05
didrocksoh, like with separators or structured?16:05
desrthttps://developer.gnome.org/gio/stable/gio-Unix-Mounts.html16:06
desrt_get_mount_path() _get_device_path() _get_fs_type() etc16:06
desrtall it really saves you, though, is the parsing16:07
desrtwhich is not difficult to do16:07
desrtthis is meant more as an abstraction layer to help on platforms that don't have /proc/mounts16:07
desrtmacos has some weird syscalls for that info, for example16:07
desrtthere is no efficient mechanism for doing a lookup, as an example.  that's something i want to add.16:08
didrocksinteresting, yeah, lookup part would avoid looping over manually16:08
didrocksanyway, no glib for me :p16:08
didrocksonly my eyes to cry16:08
desrt:)16:09
didrocks(<-- insert sad violon music here)16:09
* desrt isn't sure what he'd do without didrocks around to provide a constant injection of sadness16:09
didrocksrohhh :)16:09
desrtno.  it's good :)16:10
desrtgotta have balance, you know16:10
didrocksahah16:10
desrtanyway... all of this chatting has reminded me that i was going to solve a bug for you16:11
desrtand i should get back to that16:11
desrtbut, as with all things, ... it won't be easy :p16:11
desrt(the race between installing a new desktop file and file monitors being slow)16:12
didrocksdesrt: well, no urgency yet, I have the workaround :)16:12
didrocksdesrt: but would be nice to get it right, indeed!16:12
didrocksdid you find any clever idea on it?16:13
desrtdidrocks: i dreamed up an elaborate system for solving this problem16:13
didrocks"elaborate" ;)16:13
didrocksand "system"16:13
didrocksso 10 minutes hack? :)16:13
desrthah16:13
desrti gotta land this massive filemonitor branch first16:13
desrtthen it will be a "10 minute hack" ;)16:13
desrtbasically, providing a mechanism for file monitors to reach into the backend and say "run your queue RIGHT NOW"16:14
desrtand then check if any events are pending on themselves16:14
didrocksoh, so this will just flush the queue16:14
desrtwell16:14
didrocksnot afraid of massive things being queued at that moment?16:14
desrtit will poll the kernel too16:14
desrtwhich is part of my concern.....16:15
didrocksyeah, I see :/16:15
desrtthe way desktop directories work, this could be annoying16:15
Sweet5harkYour build system is doing madness, should you a/ fix one problem and hope it does sane things elsewhere b/not use a build system everyone know to be brittle, whose only feature is fixing portability issues with 1990ies unixes, which your software doesnt run on and never will? Discuss: https://harald.hoyer.xyz/2015/03/05/libtool-getting-rid-of-180000-sed-forks/16:15
desrtsince we have a file monitor at each level -- ~/.local/share, /usr/local/share, /usr/share, the config dirs, any extra XDG_DATA_DIRS, etc.16:15
desrtso that would be like 6 poll() calls maybe16:15
didrocksyeah, it can be expensive polling16:15
desrtwell, the call would be really really fast, i'm sure16:16
desrtbut it's still 6 syscalls, 5 of which are pointless16:16
didrocksindeed16:16
desrtso figuring out some way to consolidate that .....16:16
didrocksat least, happy that I gave you a challenging use case!16:16
desrtit also interacts badly with the case where the dir is not created yet16:16
desrt:)16:17
desrtthis stuff will be generally useful eventually16:17
desrtwill be nice to be able to query file monitors for this info in the generic sense16:17
desrtlike for the /proc/mounts monitor... would be really useful to know if a change event was pending or not16:17
didrocksright, so it can be quite generically reused that way16:18
=== Guest63892 is now known as mfisch
andyrockMacSlow, Trevinho sorry I was at class16:39
andyrockdo you still need my help?16:39
MacSlowandyrock, np16:39
MacSlowandyrock, well I just wanted to ask for your folks view on an issue I'm seeing here...16:39
andyrocksure I can reproduce it?16:40
andyrock*how16:40
MacSlowandyrock, from the looks of it, it's only restricted to nvidia (binary-blob) based ssystems...16:40
andyrocki got one16:40
andyrockMacSlow, ^^^16:40
MacSlowandyrock, when ever I start qmlscene on the desktop to view a .qml file... QtCreator no longer updates window-contents and the compiz/unity app-switcher gets really sluggish16:41
MacSlowandyrock, I'm on vivid with the latest updates pulled (yesterday evening)16:41
MacSlowandyrock, this issue does not happen on my all-intel laptop16:41
andyrockMacSlow, mmm you started getting it with last update?16:41
MacSlowandyrock, I can't easily test it against the nouveau-driver right now16:42
MacSlowandyrock, yeah... I recognized this late yesterday16:42
andyrockMacSlow, also can you pastebine a .qml file so I can test it?16:42
andyrockor a normal one is just fine?16:42
MacSlowandyrock, any one will do16:42
andyrockMacSlow, we pushed a fix for compiz yesterday16:45
Laneylarsu: yo, someone just pointed out out that GTK_MESSAGE_OTHER GtkMessageDialogs have an image-missing icon on Ubuntu16:45
Laneyhttp://paste.ubuntu.com/10541027/16:45
Laneydoesn't happen with vanilla gtk16:45
Laneydo you think it could be caused by the 'restore traditional look' patch?16:45
MacSlowandyrock, http://pastebin.ubuntu.com/10541037/ is a qml that will certainly trigger it16:47
MacSlowandyrock, it seems it's related to constant animations going on16:47
MacSlowandyrock, wait... you need the png that goes with it...16:48
MacSlowandyrock, http://macslow.org/foreground-particle.png http://macslow.org/background-particle.png16:49
andyrockMacSlow, i cannot reproduce it here17:06
andyrockwould be helpful if you can try to downgrade compiz17:06
MacSlowandyrock, so my compiz is at 1:0.9.12.1+15.04.20150303-0ubuntu1 right now17:08
andyrockMacSlow, same here17:08
MacSlowandyrock, so to what version do I need to get back?17:09
andyrockMacSlow, just one before17:09
MacSlowandyrock, 1:0.9.12.1+15.04.20150227-0ubuntu1 I assume then17:10
andyrockyep17:10
andyrockMacSlow, assuemd that you had regalluary updated your system17:10
MacSlowandyrock, I seem to still have 1:0.9.12.1+15.04.20150213-0ubuntu1 1:0.9.12.1+15.04.20150227-0ubuntu1 1:0.9.12.1+15.04.20150303-0ubuntu1 in my /var/cache/apt/archives17:12
andyrocktry one by one if you can17:12
andyrockyesterday we pushed a small fix that solved the black window nvidia issue17:13
andyrockbut should not be releated17:13
andyrockbut we got also another big fix for nvidia in the last few weeks17:13
MacSlowandyrock, just the compiz-package should be enough right?17:14
andyrockMacSlow, yep17:15
andyrockMacSlow, it should include compiz-plugins17:15
andyrockTrevinho, ^^^17:15
andyrockcan you confirm?17:15
MacSlowandyrock, hm... seems with compiz-plugins compiz-core and libdecoration I need to downgrade too...17:19
* MacSlow wonders how much he fucked up his system by now17:20
andyrockMacSlow, usually it easier to just build your own compiz :D17:20
MacSlowwell I'm lazy at my eod :)17:21
* MacSlow restarts...17:21
kaeaAs a single user, how can I alter my username in a clean way (where is all places my user is connected to) or am I better off to just create a new user and give same permissions? (if so how do I know my current permissions etc)17:24
Laneylarsu: http://paste.ubuntu.com/10541369/ ?17:26
didrockseod -> see you guys!17:29
=== alan_g is now known as alan_g|EOD
achiangare unity webapps still a thing? more specifically, there's a webpage i like that plays music; would love to be able to control it from sound indicator.19:53
dobeyachiang: they're sort of a thing. i don't know if unity8 has sound indicator integration though20:07
dobeybut should work under unity7 still afaik20:07
Laneyjust tried lastfm and got an aptd crash when I asked for the webapp to be installed. :D20:10
dobeynice20:16
Laneycan't report it because glib is out of date20:16
* Laney gets a trace manually20:16
LaneyTraceback (most recent call last):20:17
Laney  File "/usr/lib/python3/dist-packages/aptdaemon/pkcompat.py", line 589, in _remove_from_connection_no_raise20:17
Laney    self.pktrans.Destroy()20:17
LaneyAttributeError: 'NoneType' object has no attribute 'Destroy'20:17
Laneyhelpful20:18
=== Zdra is now known as xclaesse
=== greyback__ is now known as greyback

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!