[00:59] <utkarsh2102> vorlon: apologies for not checking it thoroughly and thanks for fixing it! \o/
[01:00] <utkarsh2102> aaah, wrong place. :)
=== tomreyn_ is now known as tomreyn
[08:06] <juliank> hi seb128, did you poke packagekit and glib a bit more?
[08:07] <seb128> juliank, no, I was about to resume on that now, just tried the hirsute glib and noted that it didn't fix the issue
[08:07] <juliank> ack
[08:07] <juliank> I did read glib code a bit and it seems it has 3 network montiros
[08:07] <juliank> * monitors
[08:07] <juliank> Like one network-manager one, one netlink one, and a dummy
[08:08] <juliank> I feel like maybe we need to order packagekit After=NetworkManager.service to make sure it starts after that and picks up the correct one
[08:10] <juliank> hmm it should already, though by network-online.target?
[08:10] <juliank> seb128: OK, I mean, we only have Wants=network-online.target, we probably need After=network-online.target too
[08:11] <juliank> because right now, there is no ordering between the two
[08:12] <juliank> update-notifier has After=network.target network-online.target systemd-networkd.service NetworkManager.service connman.service
[08:14] <juliank> I think there is a glib bug in that it should just discover all that automatically but ugh, this might do
[08:14] <seb128> juliank, if that works, I'm unclear why it's not able to change its status to online when the system gets online, also it was working until impish without that so I'm not sure what changed
[08:14] <juliank> seb128: subtle ordering changes maybe?
[08:16] <juliank> seb128: Though, I looked at the boot plot, and NetworkManager started before packagekit (but packagekit started before network-online)
[08:17] <juliank> anyway it's highly problematic if glib does not pick up changes if started too early
[08:18] <juliank> I'll go boot the other laptop and see what it does :D
[08:18] <seb128> I would be surprised if glib wasn't picking status changes
[08:19] <juliank> seb128: Presumably it initialized the wrong network monitor?
[08:19] <seb128> could be
[08:19] <juliank> let's gdb attach and see what it does :D
[08:21] <juliank> break pk_backend_is_online and step
[08:21] <juliank> or break g_network_monitor_get_network_available really
[08:23] <juliank> It seems to call g_network_monitor_nm_get_property
[08:37] <juliank> seb128: Further dug into it, and while proxy_signal_cb() is called on PropertiesChanged on network-manager changes, it never gets to where that calls sync_properties(), so it never updates its cached values
[08:38] <juliank> I can't figure out where it exits though
[08:38] <seb128> so sounds like a bug on the packagekit side, the cached values are there?
[08:40] <juliank> seb128: no, that's all in glib, all in gnetworkmonitornm.c to be precise
[08:41] <juliank> seb128: So it seems to listen for PropertiesChanged on the freedesktop interface on the network manager path, but it only gets the NetworkManager interface notifications
[08:41] <juliank> I only get DeviceAdded/DeviceRemoved/StateChanged on interface=org.freedesktop.NetworkManager
[08:41] <juliank> Don't see us being called for the interface=org.freedesktop.DBus.Properties PropertiesChanged signals
[08:42] <juliank> hence it exits early
[08:42] <juliank> seb128: It kind of makes sense, we only request the org.freedesktop.NetworkManager interface
[08:44] <juliank> seb128: So glib never sees any state changes
[08:48] <seb128> juliank, hum :/
[08:49] <seb128> juliank, maybe you could open an upstream glib bug see what they think?
[08:50] <juliank> seb128: I can also come up with a patch I think
[08:50] <seb128> juliank, 👍
[08:57] <juliank> seb128: but filed as https://gitlab.gnome.org/GNOME/glib/-/issues/2505 :)
[08:57] <ubottu> Issue 2505 in GNOME/glib "g_network_monitor_nm never updates on connection change (listens for signal on wrong dbus interface)" [Opened]
[08:58] <juliank> seb128: If you have an older machine where it seems to work, turn it offline, and see if it sees that it became offline?
[08:59] <juliank> seb128: It seems NetworkManager used to emit its own PropertiesChanged signal, but no longer does so.
[09:00] <juliank> DEPRECATED. Use the standard "PropertiesChanged" signal from "org.freedesktop.DBus.Properties" instead which exists since version NetworkManager 1.2.0.
[09:00] <juliank> I guess it is gone now
[09:01] <seb128> juliank, https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/853
[09:01] <ubottu> Merge 853 in NetworkManager/NetworkManager "[th/dbus-drop-legacy-properties-changed] core: drop deprecated PropertiesChanged D-Bus signal (API break)" [Merged]
[09:02] <seb128> juliank, well done on figuring out the issue!
[09:15] <juliank> seb128: this patch seems to work https://paste.ubuntu.com/p/GVfshpPZFj/
[09:15] <seb128> juliank, the patch makes sense to me
[09:16] <seb128> shrug GNOME for not noticing that issue earlier :/
[09:18] <juliank> seb128: I think it doesn't actually work, it just falls back to netlink monitor, while it seems to work overall, I don't get any of the NM backend callbacks called
[09:18] <seb128> oh :(
[09:35] <juliank> seb128: Seems I need two proxies, oh well
[09:36] <juliank> because the Connectivity property can be retrieved from the NetworkManager iface
[10:00] <juliank> seb128: I think this does it
[10:00] <juliank> https://paste.ubuntu.com/p/FTbPfbsNmn/
[10:01] <seb128> juliank, we should do an upstream MP so we get a proper review before landing the patch
[10:02] <juliank> seb128: on it
[10:02] <seb128> juliank, thanks
[10:14] <juliank> sil2100: So I did add https://bugs.launchpad.net/ubuntu/+source/packagekit/+bug/1946196 to 21.10 milestone, people can't install software graphically without it :(
[10:14] <ubottu> Launchpad bug 1946196 in packagekit (Ubuntu) "Additional Drivers fails with pk-client-error-quark: Cannot download packages whilst offline (257)" [High, In Progress]
[10:17] <juliank> sil2100: there is question whether not being able to install software is a release blocker or notz
[10:17] <juliank> seb128: https://gitlab.gnome.org/GNOME/glib/-/merge_requests/2291
[10:17] <ubottu> Merge 2291 in GNOME/glib "gio: Stop using deprecated NM PropertiesChanged signal" [Opened]
[10:18] <seb128> juliank, thanks
[10:19] <seb128> on release blocker or not, today snap-store fails to install debsn  software-properties fails to install drivers, totem fails to install codecs
[10:20] <sil2100> juliank: eeek
[10:21] <sil2100> seb128: eeek
[10:21] <sil2100> I feel like I need more coffee for all of this
=== ginggs_ is now known as ginggs
[10:41] <sil2100> seb128: what is the bug for the snap-store failing to install debs? Since we had LP: #1944768 but from what I see people report it as fixed?
[10:41] <ubottu> Launchpad bug 1944768 in snap-store-desktop "snap-store hangs building catalog on first run (impish)" [Critical, Confirmed] https://launchpad.net/bugs/1944768
[10:41] <sil2100> juliank: so that affects all software installation via our GUI?
[10:42] <juliank> sil2100: Yeah, maybe not upgrades, as they still use aptdaemon
[10:42] <juliank> I don't know if aptdaemon does online checks :D
[10:43] <juliank> sil2100: Oh, I think aptdaemon has the same bug basically
[10:43] <juliank> sil2100: It just doesn't start during boot and exits, so it will usually start with the correct state
[10:44] <sil2100> juliank: once you get some more eyeballs on your fix, could you crank up a fix upload for impish?
[10:44] <juliank> sil2100: that's the goal, yes
[10:46] <seb128> sil2100, unsure if it's reported, I just noticed yesterday while testing that it was also impacting snap-store, basically any user of packagekit
[10:48] <sil2100> The day is getting better every minute!
[10:50] <juliank> So good news on aptdaemon, it does not actually use its monitoring code
[10:50] <juliank> it was only there for the packagekit compat layer that was dropped a while ago
=== genii-core is now known as genii
=== sem2peie- is now known as sem2peie
[19:03] <vorlon> sil2100: hi, are you able to chair this TB meeting?
[19:09] <sil2100> Hey! Ouchy, can we move me to the next one, if possible?