00:13 karl5fox: Thanks Kenny, let me check the version, its the one that is mirrored by MS so I'm sure its not the "latest"
03:59 Consolatis: <*> DemiMarie hopes that not too much will break without the protocol implemented (Qubes OS)
03:59 Consolatis: re shortcut protocol: No, it doesn't. Some user controlled (rather than application controlled) alternative mode to supress handling stuff like A-Tab / A-F4 when a specific window has keyboard focus would likely be beneficial for VM / remote clients though
04:04 Consolatis: I personally very much prefer to not allow random applications like browsers to disable compositor shortcuts just because they got focus, it should be the choice of the user to do so
05:39 dviola: damn it, 6.12-rc6 is also broken (sway inside QEMU has the mouse cursor inverted down)
05:40 dviola: time to bisect
07:18 dviola: doesn't look like a kernel regression
07:21 dviola: bluetail: I might have given you the wrong info in saying that kernel 6.11.6-arch1-1 broke things
07:21 dviola: looks like it's a problem with weston and hyprland too, will need to check mesa next
08:34 karenw: So updated ubuntu, and now it has "Default Queue" in the output of WAYLAND_DEBUG=client for objects not on a custom queue. Is there a way to name custom queues in a similar way?
08:36 kchibisov: yes, they should use the new with name api.
08:36 kchibisov: e.g. mesa uses it already.
08:40 karenw: Right. Looks like the mesa amd vulkan driver doesn't, so I may go and fix that then.
08:40 kchibisov: they do, if they are compiled against the new version.
08:41 karenw: Hmm, okay. I'll make sure I've grabbed the latest then.
08:41 kchibisov: the point is not the latest, but that mesa was built against the new libwayland with support for it.
08:42 kchibisov: so you could have latest mesa, but it could have built against previous libwayland.
08:43 karenw: Ooooh, that *would* make sense. I think the ppa I'm using is latest mesa but built against libwayland from ubuntu 24.04 not 24.10.
08:43 karenw: Thanks kchibisov
09:07 MrCooper: dviola: mostly not even trivial fixes, just churn for questionable if any benefit
09:07 MrCooper: Consolatis: FWIW, gnome-shell asks for user consent before inhibiting shortcuts
09:08 MrCooper: (and that dialogue explains how to break the inhibition)
10:34 karenw: Huh. wl_surface::frame doesn't work as I expected. It only seems to trigger on events like losing input focus. I expected to get 30/60/N callbacks a second.
10:35 kennylevinsen: karenw: The wl_surface_frame callback fires once when committed at the next time the compositor believes it's a good time to draw
10:35 kennylevinsen: for a visible surface in a common setup you'll find that it's generally tied to monitor refresh
10:35 karenw: kennylevinsen: Rubber Duck Debugging strikes again. I needed to call frame() *before* commit.
10:36 kennylevinsen: yup, it's double-buffered. :)
10:36 karenw: Like most things (although my mental model was not including server-triggered events as being part of that until now)
16:02 wlb: wayland-protocols Issue #228 opened by () Pedantic build tests cannot execute because of undeclared dependencies https://gitlab.freedesktop.org/wayland/wayland-protocols/-/issues/228
16:27 karenw: Hmm, I encountered a similar issue trying to compile code using cursor-shape, had to manually generate headers for tablet-unstable-v2 in order to get it to compile originally (it's now in stable/). Despite not using or supporting touch/tablets in any way.
16:30 karenw: I feel like some kind of machine-readable dependency tracking between protocols would be useful.
16:45 diego: MrCooper: I see
16:50 dviola: this doesn't look right, in my QEMU VM glxinfo reports virgl is in use, but also radeonsi, shouldn't radeonsi be limited to the host only?
16:53 dviola: if I'm not mistaken, virtio_gpu_dri.so should be loaded on the guest but it isn't
17:20 soreau: dviola: you said it used to work? do you know what version of mesa it was working with? if you have a good commit, you can perform a bisect
17:31 wlb: wayland-protocols Merge request !363 opened by () tests: Make build-only tests actually build-only https://gitlab.freedesktop.org/wayland/wayland-protocols/-/merge_requests/363
17:51 dviola: soreau: yes, it used to work
17:52 MrCooper: dviola: what does "virgl is in use, but also radeonsi" mean exactly?
17:52 dviola: let me show you, one sec
17:52 MrCooper: the GL renderer string says both virgl and radeonsi?
17:52 dviola: MrCooper: indeed
17:52 MrCooper: then I suspect virgl simply also reports the host driver
17:53 MrCooper: it's not possible for two drivers to be active at the same time for the same GL context
17:54 dviola: http://0x0.st/XD4P.txt
17:55 dviola: that's on the guest
17:57 dviola: I'm pretty sure something broke, just can't tell exactly what yet
18:00 dviola: lsof shows a bunch of libdrm_*.so open but no virtio_gpu_dri.so
18:00 MrCooper: it's clearly a feature, not a bug :) virgl gets the host driver name from the host and includes it in its renderer string
18:01 dviola: MrCooper: ah, then I just never seen it before
18:01 MrCooper: *the host renderer string
18:01 dviola: or maybe I did and don't remember
18:01 MrCooper: maybe it was added recently
18:02 dviola: ah ok, thanks
18:03 MrCooper: well it was added in Mesa over 3 years ago, maybe your hypervisor didn't report the information before though
18:04 dviola: makes sense
18:09 dviola: this is really odd now :D -- I started recording the screen with wf-recorder and as soon as the recording starts, the mouse pointer becomes usable again, although I still see the inverted mouse below, so there are two cursors on the screen
18:10 dviola: if I stop recording, the only cursor left on screen is the odd one
18:11 soreau: dviola: wf-recorder makes sw cursor happen
18:11 soreau: (FWIW)
18:11 soreau: and wf-recorder is per-output
18:11 dviola: I see
18:13 dviola: ahh, WLR_NO_HARDWARE_CURSORS=1 made it go away again
18:14 dviola: had this issue in the past with sway on QEMU, but it went away for a while... now it's back
18:16 soreau: dviola: sounds like a possible wlroots issue in that case
18:16 soreau: but you mentioned it happens on weston too?
18:16 dviola: iirc, around kernel 6.8 it went away
18:18 dviola: just tried weston now, and it's fine...
18:21 dviola: yeah maybe I did something wrong in my previous test
18:22 dviola: river also has the issue
18:23 soreau: probably wlroots related
18:23 soreau: maybe try 0.17 vs 0.18
18:23 soreau: wayfire master is on 0.17, track-wlroots branch on 0.18 too
18:23 dviola: yep, will do, thanks
18:23 soreau: 👍
18:25 soreau: the unfortunate thing is, wlroots is kinda difficult to bisect with constant unstable API churn :P
18:25 dviola: I'm also running sway on the host and the cursor on the host also looks weird, maybe that got me confused when I tested weston on the VM
18:27 dviola: soreau: ah heh
18:27 soreau: dviola: if you find 0.17 works, then try to gather info and complain loudly in #wlroots ;)
18:28 soreau: and maybe file an issue
18:33 dviola: ha :P they probably get enough complaining already, but I will let them know ;)
18:34 dviola: I bet they get less now, with hyprland not using it anymore... sorry
18:37 dviola: I know vaxry was complaining very loudly about wlroots for a while
19:16 karl5fox: hey guys, so I've gotten weston to output the three monitors over the RDP shell, they show as XWAYLAND0 - XWAYLAND02 at the shell, but in the gui terminal in GNOME I see WAYLAND0 connected and rdp-0, rdp-4, rdp-3 disconnected. Can I put those RDP-0 etc in weston.ini as outputs?
19:34 dviola: tbh I think wayland is too good to be true sometimes, it mostly just works for me nowadays... I went back to X a couple of days ago and couldn't believe I never saw the tearing before
19:35 DemiMarie: How does one see tearing?
19:35 dviola: one thing I often hear people complain about wayland is how they need portals and pipewire for screen sharing, are portals always going to be mandatory for screen sharin or will that be replaced in the future?
19:35 DemiMarie: Depends on your compositor
19:35 dviola: sharing*
19:35 soreau: dviola: I made wayfire tear and it didn't show up in the video ;) https://streamable.com/h5njy4
19:36 DemiMarie: Portals are the portable solution, but some compositors offer non-portable extension APIs
19:37 kennylevinsen: wayland protocols are also a perfectly portable solution, it’s just a different solution
19:38 kennylevinsen: dviola: portals are currently the design flatpak and it’s proponents favor so it’s unlikely to go away anytime soon
19:39 kennylevinsen: pipewire isn’t going away either - and people were used to pulseaudio, so a more capable modern replacement isn’t an issue
19:41 dviola: DemiMarie: I see a noticeable difference when scrolling with firefox on wayland and X, when I was using X I was like "why do I even need a compositor?" because it was good enough, but I see the difference now
19:41 dviola: soreau: hah
19:41 dviola: kennylevinsen: ah yes, pipewire is really good
19:42 DemiMarie: dviola: is it a case of one being able to see the different images or is it more something that feels wrong subconsciously?
19:43 psykose: you see a torn line on the screen
19:43 psykose: where e.g. above it is the new frame and below it is the old frame
19:43 dviola: DemiMarie: good question, it's mostly about text not flowing as smoothly, what psykose said (I think)
19:49 kennylevinsen: Tearing is very obvious, but you also often ran separate compositors to fix it
19:51 psykose: separate compositors that crashed 4 times a day
19:51 psykose: i wonder if that ever got fixed, sure didn't in all the time i used X
19:54 dviola: for me crashes related to X were almost always due to some issues with xf86-video-intel, went away after I switched to -modesetting
19:58 kennylevinsen: The path to get pixels to your screen is a lot simpler now, that’s for sure
19:58 dviola: good
20:02 dviola: kennylevinsen: I suspect it's going to take a while to fully get rid of xwayland, if ever though
20:10 dviola: brodie just made a video that his windows doesn't resize as smoothly under xwayland :P
20:11 dviola: https://www.youtube.com/watch?v=8G9zX8CkGTY
20:55 kennylevinsen: I ran Xwayland free for a year, I barely noticed anything missing
20:58 jadahl: kennylevinsen: gitk
20:59 jadahl: it's the main thing that keeps my Xwayland running
20:59 dviola: I should try that sometime but for me it's Steam
20:59 jadahl: I should try to find time to game again :|
21:01 dviola: heh
21:02 dviola: yeah I don't do a lot of gaming anymore, maybe a few hours a month, sometimes not even that :P
21:23 kennylevinsen:cheats in that regard with consoles
21:26 linkmauve:still needs Xwayland exactly because of consoles… Dolphin reverted its Wayland backend years ago, and no one proposed it again.
21:29 jadahl: kennylevinsen: me too, I have a NES I played on once last month until a toddler also wanted to hold the controller
21:56 kennylevinsen: jadahl: as in that the NES is now the property of your toddler or reduced to something no longer resembling a NES?
21:58 wlb: weston Merge request !1640 opened by () Draft: backend-drm: Implement DRM_FORMAT_MOD_BROADCOM_SAND128 support https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/1640
22:05 jadahl: kennylevinsen: the NES is waiting for the toddler to gain a few more years until it becomes theirs :P
22:07 kennylevinsen: Better start educating them early :)
22:08 kennylevinsen: Ugh those modifiers with size-dependent parameters encoded into them again…
22:11 soreau: jadahl: I had this plan for an old P4 machine but now probably would have to get a tcouhscreen for it, since there's a certain addiction to touchscreens that has developed
22:12 soreau: if you can't touch it - it's broke
22:13 soreau: (the DS4->rpi->rc car project is a bit more successful, but they still want me to drive it)
22:30 jadahl: soreau: did the plan work out?
22:35 soreau: jadahl: yet to be seen, we have to decide where it will live first
22:45 soreau: jadahl: it might help in your situation to acquire the 'stomp/jump pad' or whatever it's called, so they can just dance around to make input happen (only supported by certains games, like track meet IIRC)