Laneylaney@nightingale> snap run gedit                                                                                                                                                                                                 ~00:36
Laney/bin/bash: warning: setlocale: LC_ALL: cannot change locale (en_GB.UTF-8)00:36
Laney(org.gnome.gedit:22673): GLib-GIO-WARNING **: 00:35:05.530: Error creating IO channel for /proc/self/mountinfo: Permission denied (g-file-error-quark, 2)00:36
Laney(org.gnome.gedit:22673): Gtk-WARNING **: 00:35:17.517: Calling org.gnome.SessionManager.Inhibit failed: GDBus.Error:org.freedesktop.DBus.Error.AccessDenied: An AppArmor policy prevents this sender from sending this message to this recipient; type="method_call", sender=":1.256" (uid=1000 pid=22673 comm="/snap/gedit/368/usr/bin/gedit " label="snap.gedit.gedit (enforce)")00:36
Laneyinterface="org.gnome.SessionManager" member="Inhibit" error ...00:36
Laney... name="(unset)" requested_reply="0" destination=":1.20" (uid=1000 pid=1564 comm="/usr/lib/gnome-session/gnome-session-binary --syst" label="unconfined")00:36
LaneyI feel like that method call should have been allowed00:36
Laneythat's what lets gedit block shutdown until you decide to save the document or not00:36
Laneykenvandine: jamesh: any opinion?00:38
kenvandineLaney: i'd think that would be allowed by connecting the desktop interface00:43
kenvandineLaney: i don't see that output from gedit in my X session though, so maybe wayland related as well?00:44
Laneydesktop                   gedit:desktop          :desktop                         -00:44
Laneydoes that mean it's there?00:44
kenvandinenot sure, but I think we should allow access via that interface00:45
Laneyit only denies once you've typed something btw00:45
kenvandineah... i am seeing that00:45
kenvandinejdstrand: ^^ thoughts on the allowing inhibit session via the desktop interface?00:49
Laneythat's gtk_application_inhibit() https://developer.gnome.org/gtk3/stable/GtkApplication.html#gtk-application-inhibit00:51
jameshLaney: there is an inhibit interface in xdg-desktop-portal, so I suspect the main API is more powerful than is ideal01:06
Laneyah I see01:10
Laneyjamesh: https://gitlab.gnome.org/GNOME/gtk/blob/gtk-3-24/gtk/gtkapplication-dbus.c#L43601:11
Laneyso I would guess that we can see gnome-session on the bus, which means we're trying to use it01:11
Laneydoes that sound plausible?01:11
jameshLaney: try running the app with GTK_USE_PORTAL=101:15
kenvandinejamesh: doesn't help01:15
kenvandinejamesh: probably the gtk version in the platform is too old?01:16
Laneyah if it's bionic's version that would make sense (271 isn't there)01:19
jameshperhaps? gnome-3-28-1804 looks like it contains 3.2201:19
kenvandinethat's what i'm thinking01:19
kenvandinei can verify by building gedit with gnome-3-34-180401:19
jameshwhich doesn't have the gtk_should_use_portal() short circuit01:20
Laneyneed 3.24.0 for that01:21
jameshI do wonder if the gtk_application_get_proxy_if_service_present() call to get the inhibit portal proxy is going to work or not though01:21
jameshi.e. will the calls it uses to determine if the name is available be blocked, since they'd allow enumerating any bus name01:23
LaneyD-Bus using clients are going to have a bad time if this stuff (GetNameOwner / NameOwnerChanged) doesn't work01:27
Laneybtw (slightly different topic)01:28
jameshthey're allowed by desktop-legacy01:28
jameshthe main problem is that the AppArmor dbus integration is too low level for the type of filtering we usually want to do01:29
jameshbut then we need desktop-legacy anyway for a11y and input methods01:30
jamesh(those APIs leak like a sieve as far as confinement goes)01:30
Laneywe found that if a gsettings key is at its default value, and the value is overridden in the host by a gsettings override, you don't see that overridden value inside the snap01:32
LaneyI guess we're only getting things which are set in the user's dconf database?01:32
jameshwe've got settings mostly worked out for the case where portals are enabled01:33
jameshthe desktop wide settings required by the Wayland GTK backend are accessed via a D-Bus portal API that provides read-only access to theme name, font name, etc01:34
jameshapplication settings are stored in a keyfile local to the app01:34
jameshno need to talk to dconf at all01:34
Laneyit's about reading where the host's default differs from the snap's01:35
Laneybut indeed it was for the cursor theme in this case, which will be covered by that portla01:35
jameshright.  we get the host default theme names because the xdg-desktop-portal service is running outside of confinement01:35
jameshfor application settings, the confined app never sees host settings: default or otherwise01:36
LaneyI guess I can see that this makes sense01:37
jameshand since the settings end up in a keyfile in $XDG_CONFIG_HOME (which will be in $SNAP_USER_DATA or $SNAP_USER_COMMON), they get managed with the rest of the confined app's data01:39
jameshso e.g. you won't end up with stale settings in your dconf database after uninstalling a snap01:39
didrocksgood morning07:07
jibelsalut didrocks07:29
didrockssalut jibel, ça va ?07:29
jibeldidrocks, bien et toi?07:29
jibeltrying to reproduce bug 184885607:30
ubot5bug 1848856 in grub2 (Ubuntu Focal) "Upgrade from 19.04 to 19.10 with zfs on root fails with grub syntax error" [Medium,Triaged] https://launchpad.net/bugs/184885607:30
jibelI don't know where this cr charachter comes from07:30
didrocksjibel: ça va. Didn't way say that it could come in a mirror disk situation?07:31
jibeldidrocks, yeah I created the same setup mirrored disk + cache and log devices on another partition07:34
didrocksinteresting, we got 2 people at least getting it IIRC, so unsure07:34
didrocksI think we just adapt the mock to return it07:34
didrocksand ensure it's fixed07:34
jibeldidrocks, I cannot find 10_linux_zfs in grub's vcs. where is it?07:36
didrocksthe guy fixed it himself that way, so, we shouldn't be far :)07:36
didrocksjibel: .in ;)07:36
jibeldidrocks, right but I cannot find it07:36
didrockshum, oh, you need to checkout the patched branch07:37
didrocksgit dpm checkout-patch07:37
didrocks(or patch*ed*)07:37
didrockswith ed, indeed07:37
didrocks(btw, sorry for the email github action failing spam, but as it's working again… ;))07:38
didrocksphew, passed!07:46
=== pstolowski|afk is now known as pstolowski
marcustomlinsonmorning didrocks and jibel08:24
didrockshey marcustomlinson08:26
jibelmorning marcustomlinson08:29
=== TheFuzzStone[m] is now known as asd[m]
Trevinhomarcustomlinson: hey, still trying to get that bug happen, what kind of vm was reproducible?12:59
Trevinholow-powered one maybe?12:59
Trevinhoso that it might look like a race12:59
marcustomlinsonTrevinho: the VM's I was testing on had 8GB RAM, 8 cores13:00
Trevinhomarcustomlinson: so not really bad13:00
Trevinhokvm or what?13:00
marcustomlinsonno, but I don't know at this point, perhaps a slower machine will trigger it more13:00
marcustomlinsonTrevinho: kvm yes13:01
Trevinhomarcustomlinson: uff, trying focal daily kvm, again I'm still getting the normal text pasted... -_-13:30
Trevinhook, got it now13:30
Trevinhomarcustomlinson: mh, so it looks that mutter uses the png format as default for saving data when there are various possible types...16:12
marcustomlinsonTrevinho: nice catch16:13
marcustomlinsonthat would make sense16:13
Trevinhoalthough, not sure why we're then pasting it in the way to ignore the actual type, this might be because when doing multiple requests we ignore the FORMAT request from the client16:13
Trevinhoas the client asks us the type to paste, then we should give it the one required16:14
marcustomlinsonTrevinho: I'm not sure LibreOffice would strictly specify type at the time of pasting. I mean, either text or an image is acceptable in a spreadsheet16:22
marcustomlinsonand the same would go for pasting into an email16:24
marcustomlinsonMy guess is that it's just the copy end that is broken16:24
marcustomlinsonTrevinho: I might have misunderstood what you meant. I think we are in agreement anyway.16:32
Trevinhomarcustomlinson: well the copy end by default marks such copy as image....16:33
TrevinhoHowever, libreoffice should then require the actual type (like when you do "ctrl+shift+v" and you can choose the type to paste)16:33
Trevinhoso by default a cell can be also an image16:33
Trevinhowhat could be wrong is both libreoffice choosing the type or mutter not doing the required conversion sometimes16:33
seb128goooood morning systems16:37
seb128hey marcustomlinson, Trevinho, how is it going?16:37
Trevinhoseb128: hey, good... Got an house :)16:37
seb128oh, nice, congrats! :16:37
marcustomlinsonhey seb128, good thanks, how's things there?16:38
didrockshey seb12816:39
seb128lut didrocks16:39
seb128didrocks, en forme?16:43
didrocksça va, et toi ?16:43
seb128ça va bien :)16:48
seb128hum, does anyone here remember how to change the nick completion mode from hexchat? (I assume it's working the same as under old xchat)16:51
Trevinhono idea16:52
=== pstolowski is now known as pstolowski|afk
marcustomlinsonTrevinho: I'm beginning to believe this may actually be LibreOffice's fault. While there's been 2 reports of "other apps doing the same thing", I fear there may have been a misunderstanding. Some people were saying sometimes the paste doesn't work or pastes something old17:05
Trevinhomarcustomlinson: i'm thinking there could be some faulty thing there as well, but still in debugging17:06
marcustomlinsonI haven't reproduce this on anything other than LibreOffice yet17:07
Trevinhomarcustomlinson: have you tried somwhere else that create a multi-type selection?17:08
marcustomlinsonTrevinho: I've tried gnumeric, calligrasheets and abiword so far17:08
Trevinhomarcustomlinson: you can see the possible of conversions for the selection you copied via `xclip -o -target TARGETS -selection clipboard`17:08
TrevinhoLO has also image/png for those17:09
Trevinhoand bmp17:09
Trevinhomarcustomlinson: also the way I have to reproduce it is hitting ctrl+c/v quickly many times.. Not really a way I can reproduce with an exact schema, so not the easiest to debug17:11
Trevinhomarcustomlinson: so no image/*, no way it could be misunderstood17:11
marcustomlinsonI see17:12
Trevinhonot sure if gimp has something doing similar things17:12
marcustomlinsonTrevinho: abiword: https://pastebin.ubuntu.com/p/F4wpCXYQyF/17:12
marcustomlinsonthat was from copying the text "hello"17:13
marcustomlinsonso let me see if I can reproduce then17:13
Trevinhoall this depends if the target supports svg I assume17:17
Trevinhoas last image type is picked iirc17:17
marcustomlinsoncan't reproduce it from the copy in abiword17:17
marcustomlinsontried by hitting ctrl+a/c/v over and over17:18
Trevinhook, I got some useful logging maybe17:18
Trevinhomarcustomlinson: abiworld and paste to lo?17:19
marcustomlinsonwill keep trying that17:20
marcustomlinsonoh wait17:20
Trevinhomarcustomlinson: anyways, forgot to ask, there was not a bug reported upstream yet, right?17:20
marcustomlinsonI get "contents could not be pasted"17:21
marcustomlinsonwhen copying text from abiword and pasting into LO17:21
marcustomlinsonTrevinho: https://gitlab.gnome.org/GNOME/mutter/issues/91917:21
gitbotGNOME issue 919 in mutter "Copied text from rich text editors seems to be stored in the clipboard as a bitmap in X11" [Opened]17:21
Trevinhomarcustomlinson: ok, thanks17:22
marcustomlinsonTrevinho: don't know why I didn't think about this earlier: I ran the exact same version of the LO snap on Eoan (mutter 3.34) and Disco (mutter 3.32), reproducible on Eaon, not reproducible on Disco17:29
Trevinhomarcustomlinson: sure, sure this indeed involves the new mutter17:30
marcustomlinsonjust triple checking17:30
Trevinhoit's even true that mutter could be still right, and LO not doing all the things properly17:30
Trevinhoas now mutter puts itself in the middle17:30
marcustomlinsonI see17:31
Trevinhomarcustomlinson: it's true that when this happens, is because LO asks mutter the TARGETS types we supports17:33
Trevinhobut we return only the bmp...17:33
vickyTrevinho: ping17:42
Trevinhovicky: pong17:42
vickyTrevinho: I'm trying to get some help/pointers on a gdm issue. Do you have some time to discuss. Thank you.17:44
Trevinhovicky: tell me17:44
vickyTrevinho: The details are in https://gitlab.gnome.org/GNOME/gdm/issues/50717:45
gitbotGNOME issue 507 in gdm "gdm3 login screen is blank and does not show greeter" [2. Needs Information, Opened]17:45
vickyTrevinho: From my analysis it looks to be a gdm issue. Or the kernel update breaks the userspace application.17:47
Trevinhovicky: mh, it's a bit generic, could be many things18:17
Trevinhovicky: does it work in x11, at first?18:17
vickyTrevinho: Do you mean uncommenting 'waylandenable=false'?18:23
=== hyperair is now known as Guest84234
vickyTrevinho: Uncommented 'WaylandEnable=False' in /etc/gdm3/daemon.conf and tried. The issue is still seen. If i use lightdm then the greeter appears18:31
Trevinhovicky: do you get any jouranlctl issue related to gnome-shell/gdm?18:32
Trevinhovicky: also in that file enable debugging18:32
Trevinhoand then put all this in the bug you opened18:32
vickyTrevinho: yes enabled debug in that file and took journalctl logs18:33
vickyTrevinho: https://gitlab.gnome.org/GNOME/gdm/issues/507#note_64134418:33
gitbotGNOME issue 507 in gdm "gdm3 login screen is blank and does not show greeter" [2. Needs Information, Opened]18:33
Laneywhy does firefox believe that I've launched an older version after upgrading to focal O_O21:28
* Laney runs firefox --allow-downgrade21:31
sarnoldLaney: I'm going to guess it's because eoan's -security firefox was built 2019-11-05 and focal's -release firefox was built 2019-11-0222:08
Laneyinteresting theory!22:09
Laneylooks like it could be correct22:13
Laneyand the build ID for me is 2019103100013322:14
sarnoldI thought we had a local patch to disable that check or similar after this happened in *released* firefox versions22:16
sarnoldLaney: https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/183009622:17
ubot5Ubuntu bug 1830096 in firefox (Ubuntu) "Firefox 67 in Ubuntu 18.10 thinks it's an older version" [High,Fix released]22:17
Laneysarnold: aha22:20
Laneymight need to poke Olivier again22:21
Laneyi'm glad that has managed to pass me by until now :>22:21
sarnoldsame :)22:21
=== alan_g_ is now known as alan_g

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