[04:31] Good morning [06:48] morning [06:53] bonjour didrocks [07:00] ça va pitti ? [07:00] * didrocks digs into ubuntu archives about the hibernate removal discussion :p [07:01] didrocks: ça va bien, et toi ? [07:01] pitti: ça va bien ! Il fait encore un peu trop nuit pour aller faire du vélo avant la pluie :) [07:02] oui, c'est encore sombre ici aussi [07:34] Hello, world! [07:37] hey mlankhorst! [07:39] bonjour [07:40] ca va? [07:43] ça va bien ! et toi ? :) [07:43] ca va :p [07:47] hmm, did anyone try ubuntu-desktop-next? I just tried the amd64 one on my thinkpad, and I don't get very far [07:48] I see the "welcome to your phone" wizard coming up (using only a small part of my external screen, and only rendering halfway on the internal one) [07:48] I can move the mouse cursor, but that's it -- no clicking, no "continue", no changing lanugage, keyboard doesn't do anything, etc. [07:48] Laney: ^ [07:51] pitti: did you try only with one monitor? [07:51] I know multiple monitors are not supported yet, so maybe that's what puzzle Mir? [07:52] only external one would be ok too :) [07:52] no, I didn't try yet, I'll do now [07:53] pitti: maybe it can't just decide between your two shiny monitors ;) [08:03] didrocks, Laney: ok, with just the internal monitor it worked better [08:04] the terminal app doesn't install, though :/ [08:04] (install button doesn't do anything) === happyaro1 is now known as happyaron [08:04] anyway, I could do what I wanted on VT1 after finding out how to log in :) [08:07] can we pretty please pre-install terminal-app? [08:07] pitti: that would be for Laney's seed I guess, but yeah :) [08:07] pitti: did you log in to sso by any chance? [08:07] didrocks: yes, I had to [08:07] otherwise you can't install apps [08:07] ok, so termina app doesn't install? weird [08:08] terminal* [08:08] ok, going for some cycling before the rain [08:08] ttyl guys :) [08:08] have fun [08:08] thx [08:41] cold cold cold [08:47] morning [09:04] pitti: it's click only and we can't/don't install them [09:04] and it's not "my" seed [09:04] also hi [09:05] :) [09:24] ooh, greece spam [09:25] helpfully with a english translation below it === lifeless_ is now known as lifeless [10:12] didrocks, thank you for you email re: hibernate - very interesting [10:12] I agree, we shouldn't/couldn't use it [10:18] willcooke: yeah, the feature is great under really specific circumstances [10:18] but not widely applicable unfortunately [10:18] yeah [12:04] Laney: https://jenkins.qa.ubuntu.com/job/vivid-adt-glib2.0/28 regression (email notifications are still broken) [12:04] it also seems to have regressed deja-dup [12:08] pitti: blag, ok [12:09] Test glib/network-monitor.test failed: Child process killed by signal 11 [12:11] interesting that it worked at build time [12:11] let's see [12:12] Laney: I can retry the glib one [12:12] it's probably real [12:14] well bah, it works in a chroot [12:14] * Laney does the adt dance [12:26] oh man, accidentally hit ctrl-c after it failed [12:26] at least it did fail [13:25] larsu: gdbus question [13:25] It's possible to create a GDBusProxy for any name if you give G_DBUS_PROXY_FLAGS_DO_NOT_AUTOSTART isn't it? [13:27] Laney: you can also create it for any name if you don't pass that flag :) [13:27] won't it try to start it and give an error then? [13:28] I mean a name which you don't have a .service for on the system [13:28] right, it will error out with "no such name" [13:28] larsu: not sure whether you're seeing this in unity too, but indicator sound's scale is reacting a bit odd to click-events with gtk3.14 [13:29] Laney: but the same happens if the name doesn't exist and you pass DO_NOT_AUTOSTART [13:29] run this test program [13:29] ochosi: indeed... [13:30] http://paste.ubuntu.com/9550482/ [13:30] larsu: ok, good to know. i was afraid it would only be in xubuntu... [13:30] ochosi: thanks for pointing it out :) [13:31] I get "yes" [13:31] larsu: no worries, i guess that's the whole point of testing the gtk3.14 PPA ;) [13:35] Laney: not sure right now what the semantics are. The proxy definitely calls GetNameOwner(), so "g-name-owner" should be NULL for you [13:35] Laney: but it doesn't seem to watch the name. Strange. [13:35] right I'm wondering if you need to do a test like this before using it [13:35] this is coming from glib itself https://git.gnome.org/browse/glib/tree/gio/gnetworkmonitornm.c#n248 [13:36] this code crashes if nm isn't installed [13:36] wow [13:36] do you have a bt? [13:36] over there -> [13:36] line 222 props is NULL [13:36] from line 260 [13:36] it should definitely not crash, even if you call a method on a non-connected proxy [13:37] I think 256 probably just wants to check the owner as well? [13:37] Laney: g_dbus_proxy_get_cached_property_names() is documented to return NULL in some cases [13:38] so that code should check for that [13:38] Laney: not sure, let me check when this even happens [13:39] and probably check for NULL there too I guess, for safety's sake [13:39] Laney: ah, it only returns NULL if it cannot connect to the bus at all [13:39] nod [13:40] it should probably check for name-owner [13:40] and also for NULL in 222 [13:40] but that won't ever be a problem I think: properties are always loaded when the name exists [13:41] might as well, or an assert ... [13:41] right [13:42] this code also doesn't deal with nm restarting [13:42] not sure if that is ever an issue though [14:02] larsu: https://bugzilla.gnome.org/show_bug.cgi?id=741653 [14:02] Gnome bug 741653 in gio "gnetworkmonitornm: Check if network-manager is running" [Normal,Unconfirmed] [14:03] something weird happened to the bt [14:07] * Laney lunch [14:12] soo, people onboard UAL28 seem to be doing sightseeing over the Isle of Wright and the Channel for 2-3 hours now ... thats not exactly on the way from Heathrow to NY. === Zdra is now known as xclaesse === jdstrand_ is now known as jdstrand [14:53] didrocks, qengho, bregma - got a familly emergency, have to cancel our meetings. WIll catch up with you later. [14:53] mlankhorst, ^ [14:54] rgr === hikiko-lpt is now known as hikiko === hikiko is now known as hikiko|bbiab === hikiko|bbiab is now known as hikiko [15:05] ok bb [15:09] good luck [15:14] Laney, hi [15:18] Laney, about the syncing of cups-filters, I asked OdyX already some weeks ago and there he already told me that he had started on the ippusbxd package and so I did not do the same, already done work. He told me that he will put the ippusbxd package into experimental for now, due to the Debian freeze. [15:20] Laney: just pushed two commits which to ~larsu/ubuntu-themes/gtk-314 which fix all button issues I was aware of [15:21] and updated the pad [15:29] tkamppeter: I see some recent commits there actually, would be nice to find out what's left to do and get it uploaded :) [15:29] larsu: cool, care to make a MP? [15:30] Laney: sure [15:30] does anyone normally review those? [15:31] https://code.launchpad.net/~larsu/ubuntu-themes/gtk-314/+merge/245004 [15:31] Laney: seb does [15:31] when I ask him too :) [15:32] cheers [15:48] Laney: are you subscribed to unity-control-center MRs? [15:49] nope [15:49] Laney: this trivial fix would be nice-to-have: https://code.launchpad.net/~larsu/unity-control-center/primary-toolbar/+merge/245005 [15:50] I joined the team but it's not enough to get notifications [15:51] yeah looks good, thanks [15:52] can you approve though? [15:52] yep [15:52] thanks [15:52] trade for a review of my glib patch [15:54] deal [15:55] is it just me having funny issue logging in vivid/lightdm today? it claims I can just "log in" (no password), but it fails to do so (obviously). if I login from tty1 and press enter in lightdm, it logs in.. [16:00] Laney: bah, that g_free() in the middle of the function is quite ugly [16:00] I see why you did that though [16:01] desrt: do you have a strong policy for this? https://bugzilla.gnome.org/show_bug.cgi?id=741653 [16:01] Gnome bug 741653 in gio "gnetworkmonitornm: Check if network-manager is running" [Normal,Unconfirmed] [16:01] man, why does _get_name_owner() even return a copy? [16:01] oh, threading [16:02] larsu: looks like a danw bug [16:02] (this is where the "Dan W." problem gets extra confusing) [16:03] mh? [16:03] Dan Williams = NetworkManager maintainer [16:03] Dan Winship = GIO Networking maintainer [16:03] desrt: I asked if Laney's patch is ok for you (g_free() in the middle of the function) [16:03] NetworkManager issues in GIO = probably you want Dan Winship [16:04] this is more a gdbusproxy issue [16:04] the patch appears to be in gnetworkmonitornm.c [16:04] which I hear is maintained by you now :P [16:05] true, but that part of the question was more general [16:05] meh, whatever. I won't accept_commit-now for this part anyway [16:05] imho the correct thing to do here is to create the proxy with only unique names [16:05] proxies on well-known names have always been a mistake in my opinion [16:06] and deprecate g-name-owner? [16:06] the trouble is that david one thought it would be reasonable to try to support this.. [16:06] *once [16:06] but seriously... what do you do when people query properties on a dead service? [16:06] ya, clearly [16:07] but you could still create a proxy for a non-existent unique name [16:07] or the unique name you created it for disappears [16:07] lame [16:07] why? Same problem, no? [16:07] the idea is that you'd use watch_name_owner() to uncreate the proxy as soon as the name vanishes [16:08] of course, but you'd still need to return something from _get_properties() [16:08] well, the thing is... with the unique-name-disappears case we can reasonably assume that someone will be cleaning up the mess soon [16:08] so we can just return the old values [16:09] GetProperties() call might not succeed when you first create the object [16:09] with the well-known name, we need to pretend [16:09] larsu: isn't GDBusProxy init failable? [16:09] i think that's what happens in that case [16:09] no, it only fails when it can't connect to the bus [16:09] aka never [16:09] i'm pretty sure you're wrong [16:10] g_dbus_proxy_new_sync() takes a GDBusConnection and returns GError [16:10] I checked earlier - but maybe I missed something [16:10] that file is a bit convoluted [16:10] ya.... [16:10] i really hate that class [16:10] never use it myself [16:10] * didrocks should stop writing g_debug in systemd code… [16:10] deprecate it! [16:10] didrocks: s_debug! [16:10] "just use gdbusconnection, you wimps" [16:10] right [16:11] larsu: almost, log_debug ;) [16:11] desrt: anyhow, is danw actively reviewing patches? [16:11] larsu: trouble is that we can't really change this code to introduce new ways to fail [16:11] larsu: on this sort of stuff? yes. [16:11] ok, I'll wait for him to comment then [16:11] Laney: sorry :/ [16:12] desrt: why not? [16:12] shrug [16:12] separately, want to push https://bugzilla.gnome.org/show_bug.cgi?id=741259 ? [16:12] Gnome bug 741259 in .General "gtkmodelmenuitem: force icon scaling" [Normal,New] [16:12] * Laney will cherry-pick it for the upload [16:12] Laney: ah, mclasen ok'ed it. I missed that. I'll push it when I'm back home [16:13] larsu: this code makes me want to cry [16:13] desrt: stop reading it ;) [16:13] but anyway you're right [16:13] explicitly: [16:13] if (result == NULL) [16:13] { [16:13] /* We just ignore if GetAll() is failing. Because this might happen [16:14] :( [16:15] is gdbusproxy actually crashing, or what? [16:16] or just returning null strings? [16:16] no, gnetworkmonitornm is, because it doesn't check the return value of get_cached_properties() for NULL [16:16] because unless gdbusproxy is crashing in response to relatively innocent outside uses, i'm not inclined to change anything inside of it [16:16] I wonder if NULL is correct there, or if { NULL } would be better [16:16] right. so we need to fix the nm code [16:17] NULL is better [16:17] this is what Laney's patch does [16:17] does it return NULL when the name disappeared, too? [16:17] { NULL } means "i'm connected and i see that there are exactly 0 properties" [16:17] I agree [16:17] something like _is_connected() might be nice to have [16:18] so that you don't need to make a copy of the name owner [16:18] you have it. it's called g-name-owner. [16:18] ahb [16:18] this shouldn't make a copy. [16:18] yay threads [16:18] no. [16:18] yay bad api design [16:18] the proxy belongs to a particular context. it receives all notifications through dispatches on that context [16:19] if you're querying it from that context (as you must) then there is no way that it can see changes from the bus while you're querying it [16:19] you have to return to the mainloop [16:19] it's just plain old bad api [16:19] you must? [16:19] oh please tell me you must... [16:19] otherwise why would it need the property_lock? [16:20] oh. you not must. [16:20] head → desk [16:20] THIS OBJECT IS TERRIBLE [16:20] right. I had the same reaction earlier today [16:21] and here i am in kdbus land figuring out a way to forbid even touching the _connection_ from the wrong thread [16:21] and we have proxies taking locks all over the place [16:22] i can't even imagine what horror show awaits me when i try to get gdbusproxy working on a kdbus gdbusconnection [16:23] i particularly love how they use code locking here and not even data locking [16:24] larsu: my suggestion is to slowly back away from GDBusProxy and fix things in the nm code (as the patch does). danw is your go-to for that, though. [16:28] desrt: backing away from GDBusProxy towards subscribing to property change notifications manually? [16:28] backing away as in typing ":q" into any relevent vim windows [16:28] lol [16:28] done. [16:28] :) [16:32] * desrt is having lots of fun in kdbus land this morning [16:33] it turns out that pretty much everything is racy [16:34] uh oh [16:35] the cookie-based approach turns out to be kinda bad -- so we're trying futexes instead :) [16:35] uhm, yay!? [16:36] ya. i love futexes. [16:52] Laney: pushed the patch to bug 741259 [16:52] bug 741259 in transmission (Ubuntu) "Transmission hangs no matter what" [Undecided,Expired] https://launchpad.net/bugs/741259 [16:52] (gnome bug) [16:53] thanks! [16:54] * Laney screams [16:54] ci traiiinnnnn [16:55] choooo choooo [16:55] https://ci-train.ubuntu.com/job/prepare-silo/3548/console [16:57] Laney: not sure how they handle the REQUEST_ID nowdays [16:57] yeah pinged in the channel [16:58] waow, they reintroduced all the variable in prepare-silo? [16:59] instead of having to click in the spreadsheet, filing the REQUEST_ID and fetching the info from there? [16:59] (this was still available manually in the prepare-silo-manual beforehand, but that was for emergency) [18:02] right then, climbing time [18:02] see you! [18:02] larsu (everyone): gtk is in ppa:ci-train-ppa-service/landing-008 [18:02] I think it's ready to go and will do that first thing tomorrow [18:02] it's there with theme and u-c-c [18:02] there's an alpha freeze so now/then doesn't matter much anyway [18:02] darkxst: ^^^^^^^^^^^^^^ fyi [18:03] * Laney waves === alan_g is now known as alan_g|EOD [18:08] Laney: neat, thanks! [18:08] I shit myself because all of the theme fixes went away [18:09] luckily it had been downgraded by mistake :) [18:10] haha [18:13] * didrocks waves too === dpm is now known as dpm-afk [20:35] Laney, Great! === ara is now known as Guest47294 === dpm-afk is now known as dpm