09:41 kennylevinsen: jadahl: OT-ish, but the xdp-gnome hangs still do happen on latest releases (1.16 and 44.1) - seems like xdg-gnome fails immediately, but dbus/systemd decides to wait the full service activation timeout
09:42 jadahl: kennylevinsen: needs 1.16.1, but think that'll happen soon. make sure not to have xdp-kde installed too, it seems to have the same problem
09:42 kennylevinsen: ah
09:46 kennylevinsen: I feel that there is still merit to the PR allowing Settings to be configured like the rest
09:48 kennylevinsen: but the service activation timeout issue is definitely the most annoying issue, so let's see what 1.16.1 can do :)
09:50 davidre: problem is that gtk requires the gtk settings to have sensible font rendering :/
09:50 kennylevinsen: does the gtk portal provide that too?
09:51 kennylevinsen: at least works fine without the gnome portal
09:52 kennylevinsen: might make sense for the gnome portal to have a Settings only mode like jadahl suggested in the discussion, but there's precedence and preference issues with just using all backends...
09:52 davidre: yes
09:53 davidre: with xdg-desk-portal-gtk uninstalled this is how bustle looks on my system
09:53 davidre:uploaded an image: (17KiB) < https://matrix.org/_matrix/media/v3/download/kde.org/bb7fbb305f41c24759d15b1cae0779857e87fe38/Screenshot_2023-05-15_11%3A52%3A41.png >
09:54 kennylevinsen: ah yeah I remember that issue, and crops up in support occasionally
09:56 davidre: So you probably want to run at least the gtk portal on every system if you expect to run gtk apps
09:57 kennylevinsen: yeah, I think that makes sense - the tricky part becomes DE-specific things. A good example was auto-darkmode, where you have Darkman, GNOME and possible KDE all providing the same setting controlled differently.
09:59 kennylevinsen: (and maybe Gtk's font setting fallback just need a small adjustment)
09:59 davidre: I vaguely remember that xdp would at least ask the one belonging to the session first
09:59 kennylevinsen: is there a dedicated place for XDP discussions?
10:00 davidre: I dont know of any except the github :D
10:00 jadahl: davidre: if it doesn't already, it probably should, at lesat
10:00 jadahl: kennylevinsen: there is a matrix room
10:01 kennylevinsen: ah, fair enough
10:02 davidre: jadahl: I remember from reading the code once, and at least on my system it first asks the kde portal for a Setting
10:02 kennylevinsen: my matrix home server is resting in bit heaven, and have been a bit detached since that
10:03 pq: In this Recommendation, when we refer to the luminance of a single color component, we mean NOT the luminance of a single color component. *sigh*
10:04 pq: FWIW, both HLG and PQ use the same definition at least.
10:06 pq: well, inside BT.2100
10:06 jadahl: davidre: that code was changed a bit recently, from "look at UseIn for priority" to "lets see what DE's is asking for"
10:06 davidre: as I understand kennylevinsen he complains that it was not changed for the settings portal :D
10:07 davidre: but I would still hope it retains priority
10:09 kennylevinsen: "complains" is a strong word, I just think that our new, better system is applicable to Settings too :)
10:09 kennylevinsen: context: https://github.com/flatpak/xdg-desktop-portal/pull/1018#issuecomment-1543541263
10:18 q234rty: fwiw https://github.com/flatpak/xdg-desktop-portal/commit/2a219279997c2124c8a639b0d009c9946ec97b40 does not seem to resolve the time out issue with x-d-p-gnome for me
10:19 q234rty: I think the reason is that that commit does not change find_all_portal_implementations so settings will still try to launch x-d-p-gnome and times out
10:24 jadahl: kennylevinsen: if xdp-gnome and xdp-kde can learn how to enter "settings only mode", and xdp dealing with priority correctly, I suspect it can work with darkman as long as it gets to have a say before anything else
10:58 kennylevinsen: jadahl: I guess the question becomes how to handle the priority
11:11 q234rty: jadahl: can we reopen https://github.com/flatpak/xdg-desktop-portal/issues/986 maybe? I'm pretty sure the issue is not fixed by the mentioned PRs
11:14 jadahl: kennylevinsen: prioritize whatever the DE has in its "want this portal" file, and use anything else after ?
11:18 kennylevinsen: yeah mean specifyign an ordered list for Settings? Or just a single backend as priority?
11:19 kennylevinsen: the former is basically my suggestion, except you can also decide not to include a backend
11:20 kennylevinsen: but if we allowed a value of "gnome;*", wouldn't that be the best of everything?
11:20 jadahl: kennylevinsen: I mean first "filling out" the settings with whatever comes from the preferred list, and then continuing with all other backends for the rest
11:20 jadahl: if 'UseIn' is respected, seems the end result is in the screenshot from before
11:24 kennylevinsen: yeah UseIn wasn't great
11:35 jadahl: nor the fact that there is no "good" way to start in settings-only mode
11:39 davidre: Now that there is a list of which impl to use for which settings maybe it could make sense for a way for xdp to forward that info
11:39 davidre: Tell each impl which interfaces it should support
12:41 q234rty: jadahl: see the new comment in the issue, turns out it does seem to be related to dbus-daemon vs dbus-broker...
12:43 jadahl: ah, great..
12:44 jadahl: either way, adding "settings only" mode and having xdp-{gnome,kde} handle them differently should address it. will add it to the gnome one
12:45 q234rty: side note: initially I was facing this issue in greetd but not in the session greetd launched
12:46 q234rty: turns out its because I ran enable and enable --user but not enable --global on dbus-broker
12:52 pq: ookay, then when an EOTF says that color channel is at C cd/m², it actually means that you take a white of (C, C, C) optical which results in C cd/m², and then zero the other channels. The light that is left is the chosen channel at "nominal C cd/m²" which is always less or much less than an actual C cd/m². That depends on both the RGB emitter white point and the... I forgot the name of the human spectral
12:52 pq: brightness sensitivity curve.
12:58 pq: I guess that's a trick that makes it possible to use the same EOTF for Y/I and per RGB channel, instead of defining a differently scaled EOTF for each.
13:03 pq: while reading about PQ and HLG, I've also seen comments that the peak cd/m² is only achievable at the device white point, and any other chromaticity is necessarily less. Which I knew, but nice to see confirmation.