[07:08] good morning desktoppers === jose_ is now known as jose [09:29] goood morning IRC [09:35] hello seb128 [09:39] hey ricotz, how are you doing? [09:42] seb128, sore muscles after some volleyball yesterday, otherwise all good :) [09:42] how are you? [09:43] * ricotz will have to convince sil2100 to accept failing s390x autopkgtests for libreoffice/kinetic :\ [09:51] ricotz, we don't support desktop on s390x so should be easy convincing :) [09:52] ricotz, I'm alright, I recovered from covid, might be able to go play tennis tonight if the rain stops early enough [09:52] seb128, hehe [09:53] seb128, oh, still playing outdoors [09:54] yes, the club I'm going to as very few indoor courts so most of the classes and games are outside even in winter... [09:57] I see, great that this still works out, would be quite cold to do so here ;) [09:58] :) [10:17] ricotz, https://ubuntu-release.kpi.ubuntu.com shows the queues got smaller over the course of the week but they seem to move slowly at this point :/ [10:26] So, app switching is broken in lunar [10:26] Alt+Tab can't switch back to an open gnome terminal [10:27] The terminal app always appears on the left side of the Alt+Tab popup [10:27] In fact, Alt+Tab just tries to switch somewhere but always switches back to the same window [10:28] you need to keep it pressed, wait for the switcher to pop up and then select [10:32] i'm gonna try the proposed gnome-shell :D [10:34] hmm closing the terminal windows made it work again, how odd [10:34] still need to investigate why chrome is not picked up [10:34] possibly a chrome window manager notification issue [10:35] I have multiple PWAs installed in Chrome so when it opens a normal chrome window I guess it forgets to tell the WM about it? [10:36] hmm also on a fresh Chrome start [10:36] the window is open but you don't see it in alt+tab [10:36] for another 5s or so while there's the loading cursor [10:56] hmm [10:56] firefox supports a --class argument, and gnome-shell recognizes correct wmclass but always groups it under firefox_firefox.desktop when launched via snap launcher; whereas if you launch the firefox binary in the snap directly it gets grouped correctly [11:34] juliank, mind filing a bug at https://bugzilla.mozilla.org/enter_bug.cgi?product=Firefox for that? [11:35] let me find my bugzilla account from 15 years ago sometime later :D [11:49] thanks :) [12:41] juliank, there's a bug already about the launcher thing [12:41] old one [12:41] I thought it was fixed [12:42] https://bugzilla.mozilla.org/show_bug.cgi?id=1751153 [12:42] -ubottu:#ubuntu-desktop- Mozilla bug 1751153 in Core "[meta] No Firefox Nightly launcher in Gnome (Wayland apps can't set their icon themselves, a proper Nightly.desktop file is also required for non-packaged builds)" [--, Reopened] [12:56] lissyx: different issue afaict [12:56] are you sure? [12:56] people mentions similar things [12:57] what happens there afaict is that StartupWMClass is missing from the desktop files [12:57] But the matching of StartupWMClass to desktop only works when not invoked by snap [12:58] which sounds a lot like the last comments on that bug [12:59] the last comment talks about firefox-beta appearing as firefox because it sets the wrong class which is different [12:59] so they hide their firefox.desktop because they don't use it [13:00] What happens to me is that basically everything works fine as long as you don't put snapd in between [13:00] Which I don't even think is a firefox issue, but a GNOME one [13:01] (also I'm still on X) [13:02] due to libva accel in google chrome :D [13:03] well I never had any issue here [13:03] but I'm forcing wayland [13:06] lissyx: so you have a custom firefox-foo.desktop with like /snap/bin/firefox -P foo --class firefox-foo --name firefox-foo and you can run that and the windows appear as Firefox (foo) instead of firefox? [13:06] wait wait [13:06] you're adding new preconditions now [13:06] I'm not, no [13:07] well you are [13:07] Not really no [13:07] changing the class name [13:07] I was referring to the default setup [13:07] there are still bugs open around handling of the class name [13:08] some are linnked from this bug [13:08] > firefox supports a --class argument, and gnome-shell recognizes correct wmclass but always groups it under firefox_firefox.desktop when launched via snap launcher [13:08] so it could also just be this [13:08] literally talking about --class the whole time [13:08] good morning [13:08] --name I added to maybe help but it doesn't [13:09] the desktop file we produce makes use of StartupWMClass [13:09] Also if you open looking class, the wmclass is recognized correctly, but it matched it to the wrong desktop file (earlier?) [13:09] Right so my firefox-foo.desktop has --class firefox-foo, StartupNotify=true, StartupWMClass=firefox-foo [13:10] If you replace the Exec=/snap/bin/firefox prefix with Exec=/snap/firefox/2088/usr/lib/firefox/firefox it works fine :D [13:10] like arounnd https://bugzilla.mozilla.org/show_bug.cgi?id=1530052 [13:10] -ubottu:#ubuntu-desktop- Mozilla bug 1530052 in Core "WM class usage inconsistent" [Normal, Resolved: Fixed] [13:10] juliank, have you verified the --class arguments gets into firefox itself? [13:10] I think about:support should help around that [13:11] well yes, it appears in the wmclass field in gnome's looking glass after [13:11] so it gets in and back out again [13:11] it's all communicated *at some point* [13:12] i'm checking something on my Windows VM, I should be able to hack on that soon [13:12] (command line parsing is ... what it is) [13:13] https://searchfox.org/mozilla-central/rev/05f486c76f84830041e51b7c68e3e697dc768454/widget/gtk/nsAppShell.cpp#329 [13:13] this I dont like [13:15] https://searchfox.org/mozilla-central/search?q=symbol:F_%3CT_mozilla%3A%3AXREAppData%3E_remotingName&redirect=false [13:15] quite a few places [13:16] oSoMoN, I clearly have a perf issue using multipass on core22 [13:16] build is too slow and it's failing likely because of lacking memory [13:17] lissyx, why don't you use the default lxd provider? [13:18] because I have to go through the whole setting up lxc ... [13:18] and that's inconvenient because I dont really have time to take care of that right now [13:56] juliank, ok so default stuff is working fine, let me try to repro your issue [14:01] juliank, ok, indeed 100% repro [14:04] juliank, so I click on the "Firefox Foo" in the dock on the left and it matches the default Firefox's Snap icon, is that right? [14:10] oSoMoN, ok bad news, using lxd fails with a weird python error [14:10] zip set to automatically installed. [14:10] 'NoneType' object is not subscriptable [14:10] We would appreciate it if you anonymously reported this issue. [14:10] No other data than the traceback and the version of snapcraft in use will be sent. [14:10] Would you like to send this error data? (Yes/No/Always/View) [no]: yes [14:10] (and sentry send seems to have failed) [14:19] so it's failingn there is nnothing in the logs :( [14:19] lissyx, can you share the entire log somewhere? [14:19] oSoMoN, there's really not more [14:19] that's the full log [14:19] Launching a container. [14:19] Waiting for container to be ready [14:19] Waiting for network to be ready... [14:19] 'NoneType' object is not subscriptable [14:21] meh, that's not really helpful [14:22] no and I have to leave in 10 min [15:00] lissyx: yeah I think so, if you alt+f2 and enter lg to get to looking glass, you'll notice the window has the correct WMClass but is associated with firefox_firefox.desktop [15:02] lissyx: I've read peopel who say it works in other WMs, but gnome is too stupid, the idea being that firefox starts out as firefox, and then changes the class name, but gnome already categorized it under the firefox.desktop file by then [15:03] but I've not done a lot of investigating :D === Trevinho[m] is now known as Trevinho-[irc] === Trevinho-[irc] is now known as Trevinho[m] === Trevinho[m] is now known as Trevinho[IRC] === Trevinho[IRC] is now known as Trevinho[m] [15:37] juliank, so it would definitively be a gnome-side issue? [15:37] would we have a way to confirm? [15:37] I do not know for sure, I don't do much desktop development, so I don't know if it looks at other properties that may be wrong? [15:39] do we have a way to track how gnome handles that? [15:39] the thing the snapd launching does at the very least is put the processes into a snap.firefox.firefox..scope [15:39] you said we change the class name, but I dont know exactly if it is what is happening in the code for sure [15:40] well otherwise gnome would'nt see a changed class name :) [15:40] oSoMoN, got rid of the conntainer, restarting and now I get a proper stack [15:40] like https://paste.mozilla.org/neRUPNgx/raw [15:42] https://forum.snapcraft.io/t/object-is-not-subscriptable/27894/3 :( [15:42] https://bugs.launchpad.net/snapcraft/+bug/1869506 [15:42] -ubottu:#ubuntu-desktop- Launchpad bug 1869506 in Snapcraft "snapcraft crashes if machine not connected to internet" [Undecided, Fix Released] [15:42] I think my LXD setup is somehow messed up and this breaks its networking [15:43] and from there I hit this known issue [15:46] juliank, i'm wondering if "--class" is handled by us or gtk ... [15:46] especially reading the comment https://searchfox.org/mozilla-central/rev/05f486c76f84830041e51b7c68e3e697dc768454/widget/gtk/nsAppShell.cpp#329-335 [15:54] gnome-software ubuntu/master 397f4f5 Jeremy Bicha * pushed 4807 commits (first 5 follow) * https://deb.li/3QlPi [15:54] gnome-software ubuntu/master 99c5f85 Milan Crha lib/gs-external-appstream-utils.c * gs-external-appstream: Remove old files from legacy path * https://deb.li/14x0 [15:55] gnome-software ubuntu/master 40e5e8e Philip Withnall plugins/fwupd/gs-fwupd-app.c * Merge branch 'fwupd-release-urgency' into 'main' * https://deb.li/3VRm2 [15:55] gnome-software ubuntu/master 11d51c8 Philip Withnall lib/gs-utils.c * Merge branch 'wip/notify-pending-updates' into 'main' * https://deb.li/DLgs [15:55] gnome-software ubuntu/master 3b27009 Philip Withnall gs-install-appstream/gs-install-appstream.c lib/gs-external-appstream-utils.c lib/gs-external-appstream-utils.h * https://deb.li/3RoJU [15:55] Merge branch '1909-external-appstream-installer-leaves-old-files-behind-in-legacy-locations' into 'main' [15:55] gnome-software ubuntu/master e1e3263 Milan Crha gs-install-appstream/gs-install-appstream.c * https://deb.li/Tmln [15:55] gs-install-appstream: Ensure the installed file has correct permissions and owner [15:55] gnome-software signed tags 506f67e Jeremy Bicha ubuntu/43.1-2ubuntu1 * gnome-software Debian release 43.1-2ubuntu1 * https://deb.li/iipa7 [15:56] lissyx: I don't think I'd worry about firefox so far [15:56] $ grep CLASS direct snap [15:56] direct:WM_CLASS(STRING) = "Navigator", "foobar" [15:57] snap:WM_CLASS(STRING) = "Navigator", "foobar" [15:57] running xprop on the windows when launched directly vs by snap [15:57] with --class foobar [15:57] same for both [15:58] OK so snapd produced a firefox_firefox.desktop without a WMClass that overrode everything, nice [15:59] I guess *something* is matching on the X-SnapInstanceName=firefox in there [16:01] the firefox -> snap transition installs a hidden firefox.desktop that works === Trevinho is now known as Trevinho[irc] === Trevinho[m] is now known as Trevinho [16:05] Oh and I found out that Chrome simply has broken startup notification, hooray [16:08] Apparently it supproted startup notification with system title bars but also those seem gone [16:16] ok, still having issues with lxd :( [16:37] juliank: sorry, I've been reading a bit but what's the whole problem? Matching a snap launched app with its desktop name? [16:37] gnome-initial-setup ubuntu/master 9c3fff6 Jeremy Bicha debian/changelog * releasing package gnome-initial-setup version 43.1-2ubuntu1 * https://deb.li/3zc0g [16:38] Trevinho: yes, the snap process is always matched to the snap desktop, so custom overrides with --class don't work [16:38] So you can't create a separate firefox-foo.desktop where you do StartupWMClass=firefox-foo and pass --class=firefox-foo to firefox [16:38] juliank: indeed because that's happening at mutter level, we figure out the desktop name from its group data [16:39] ah ok now I found it [16:40] that's bad [16:40] juliank: that's expected, you should not be able to redefine an app that is in a container https://www.pastery.net/pcbnnk/ [16:40] that's the code leading to it [16:41] the fact is that a desktop-id is not just a matching element, it also involves security (as per the portals), so we need to be sure that it always matches [16:41] that is just wrong and means you can't snap chrome properly because all PWAs are broken [16:41] looks like i dont have ipv4 in the container? [16:42] Because each PWA obviously needs to have its own desktop file with its own icon and its own window class so it appears as a separate app [16:42] And I'm kind of evaluating how the hell I could switch to firefox [16:43] because I run a bunch of PWAs in chrome, like Mattermost [16:43] I can't have them all in a random browser window [16:43] I think it could be possible but requires some more discussion in how implement that for the portals [16:44] to be fair the PWA processes should be running in their own cgroup really :D [16:46] I don't know why I did not find this when grepping mutter code for snap earlier [16:46] yeah, but I mean we can't provide the security model without having a way to define allowed desktop files we can match on... Probably a good way would be using desktop-id's as KDE does so basically group-app.id.desktop, so ideally you could have org.firefox-sub.app.desktop and then we allow that, by using different security rules though [16:46] maybe I just grepped for scope [16:47] it's all upstream so debian's codesearch should have been working [16:47] I had a local checkout I iterally ran grep on [16:47] but must have searched wrong words [16:50] anyhow was getting confused for a time [16:51] oSoMoN, https://forum.snapcraft.io/t/docker-snap-interferes-with-lxd-snap/13789/3 [16:51] oSoMoN, stopping docker.service seems to work better ... [16:51] but that's going to be messy, tc runs on docker ... [16:51] So I need to file bugs (1) against chrome because they advertise startupnotify, but don't notify; (2) against mutter to figure out how we can have "sub applications" [discuss on snapcraft.io?], (3) against something else I forgot [16:52] And complain again about firefox not supporting PWAs [16:52] :D [16:52] well I did that yesterday [16:53] not that chrome is perfect, sometimes it sends me notifications for the PWA and they open in the browser instead ... [16:53] in any case, I think it's the right approach in the snap context, I think it could be changed to use snap routine portal-info instead, and then rely on that to support other desktop files if any, but that's snapd work anyways [16:56] same issue applies to flatpak of course [16:57] anyway [16:58] I prefer the flatpak, it installs as Firefox instead of Firefox Web Bro... [16:58] the snap so hipster bro [16:59] ogra, I dont want to be painful, but I'm hitting quite a lot of issues with snapcraft and lxd ... [17:02] lissyx, well, i never do ... but then i also only use "snapcraft --destructive.mode" from a shell inside the container ... and i rarely build things at the size of firefox either [17:03] once I stopped docker.service, the build moves further but I hit some issue with snap being unable to reach api.snapcraft.io [17:03] ah, i have only used docker twice in my life i think (... found it unusable and went back to doing my stuff in lxc) [17:04] so not a prob i have ever hit [17:05] https://status.snapcraft.io shows there was some outage with api.s.io today ... pehaps bad timing ? [17:06] 3 PM CET [17:06] I was having issues like 5 minutes ago [17:06] either there is a new one not yet reflected here, or something else is going on [17:07] yeah [17:08] probably more docker mess with your network settings ? not sure what it does to that [17:08] lxd should have its own bridge [17:08] I dont know why it would interfere with docker's one [17:08] me neither ... [17:09] reported my chrome issue https://bugs.chromium.org/p/chromium/issues/detail?id=1393062 [17:09] I thought a couple months ago they moved to issuetracker.google.com but it's still there [17:09] and I do need docker, so uninstalling is not really an option [17:09] you could use podman? [17:09] :D [17:10] and the way it fails is even different depending on core20 or core22 [17:10] I like had 30 veth yesterday [17:10] I don't know why I think for me it's multipass going mad [17:11] everything is messy :D [17:12] well, yeah, snapcraft has the very bad habid to re-invet itself completely for new base snaps so it executes completely different code paths for 20 and 22 [17:12] *habit [17:14] ogra, https://paste.mozilla.org/ngrmLYQe/raw [17:14] but I'm running out of time for today sadly [17:15] it was able to perform some apt steps earlier so network should be okay [17:16] inside that container ? [17:16] hm ipv4 seems broken inside the container [17:16] and api.snapcraft.io returns only v4 [17:16] yeah 😞 [17:18] hmm, it seems to unset all proxy vaiables in the first lines ... do you have any proxy ? [17:18] (in use that is) [17:20] no [17:21] I give up. [17:22] there's no masquerading rules in iptables nat table for the lxd network [17:22] and packets are coming out of lxdbr0 [17:23] if you shell into the container, can you "snap install core20" ? [17:25] no [17:25] it's unable to contact snap store [17:25] since there's nno ipv4 connnectivity inside the container [17:25] yeah, weird [17:26] not sure how you even got to such a state [17:26] I already did a bunch of "snap run snapcraft clean" [17:27] since I can ping the container from the host [17:27] so weird [17:29] right "iptables -L -v -n" sshows FORWARD in DROP policy and counter of packets dropped inncrements when I ping ... [17:30] that's it [17:30] iptables --policy FORWARD ACCEPT [17:31] not the best solution, but will do for the moment. [17:32] there are a few debugging hints at https://discuss.linuxcontainers.org/t/containers-not-getting-ipv4/6668 [17:35] annd now a new error ... [17:36] https://paste.mozilla.org/s2AxSMCb/raw [17:39] likely the github issue :( [18:47] ogra, ok that last one was a temporary hiccup [18:47] build from scratch finished in 33m [18:58] 👍 [19:14] gnome-shell ubuntu/kinetic 221a933 Jeremy Bicha * pushed 593 commits (first 5 follow) * https://deb.li/3qEpU [19:14] gnome-shell ubuntu/kinetic 6b30471 Jeremy Bicha debian/changelog * releasing package gnome-shell version 43~beta-1ubuntu2 * https://deb.li/30dZG [19:14] gnome-shell ubuntu/kinetic 28c2dc4 Jeremy Bicha (66 files in 19 dirs) * Merge tag 'debian/43_rc-1' into ubuntu/master * https://deb.li/3iQAn [19:15] gnome-shell ubuntu/kinetic 11bb28a Jeremy Bicha debian/ gnome-shell-common.postinst gnome-shell-common.preinst gnome-shell-common.prerm * Drop obsolete maintainer scripts handling Ubuntu theme migration * https://deb.li/i2zH7 [19:15] gnome-shell ubuntu/kinetic d6ed115 Jeremy Bicha debian/changelog * releasing package gnome-shell version 43~rc-1ubuntu1 * https://deb.li/U61l [19:15] gnome-shell ubuntu/kinetic 293a552 Jeremy Bicha debian/ control control.in * Suggest instead of depend on gnome-backgrounds * https://deb.li/m1ex [19:15] gnome-shell signed tags 8f14038 Jeremy Bicha ubuntu/43.1-0ubuntu1 * gnome-shell Debian release 43.1-0ubuntu1 * https://deb.li/3tExI [20:21] gtk3 ubuntu/master 4da562c Jeremy Bicha * pushed 12 commits (first 5 follow) * https://deb.li/aTth [20:21] gtk3 ubuntu/master b41b8e9 Simon McVittie debian/ control control.in * (Build-)Depend on libfontconfig-dev, not transitional libfontconfig1-dev * https://deb.li/3MKBc [20:21] gtk3 ubuntu/master d03dca1 Simon McVittie debian/patches/ (94 files) * d/patches: Update to upstream gtk-3-24 branch commit 3.24.34-204-g2fcc114870 * https://deb.li/Ihw7 [20:21] gtk3 ubuntu/master 98b1965 Simon McVittie debian/ control control.in tests/control * Build-/test-depend on dbus-daemon for dbus-run-session, instead of dbus * https://deb.li/3eGYd [20:21] gtk3 ubuntu/master 705fb86 Simon McVittie debian/changelog * Update changelog * https://deb.li/3kdrF [20:21] gtk3 ubuntu/master c8db199 Simon McVittie debian/ (7 files in 2 dirs) * Update patch series to 3.24.34-209-gf8d26e2acc * https://deb.li/iWNT1 [22:14] gtk3 signed tags a237edc Jeremy Bicha ubuntu/3.24.34-5ubuntu1 * gtk+3.0 Debian release 3.24.34-5ubuntu1 * https://deb.li/5JJd