00:24 <ifreund> leon-p: I know what's happening there, I just haven't bothered to do anything about it yet

00:24 <ifreund> notice that this only happens with mpv while it is paused

00:25 <ifreund> what happens is that river handles focus change using the transaction system and waits for clients to commit a new frame with the activated state taken into account

00:25 <ifreund> which makes sense if the client has CSD, but not in the general case

00:33 <leon-p> i see, makes sense

00:34 <ifreund> to be clear, this is a bug with river and we should fix this, I just haven't found the time yet

02:11 <leon-p> ifreund: found a bug: I can reliably soft-lock river with a faulty layout

02:11 <leon-p> interacting with views still works, but I can no longer move then (neither with a layout nor via mouse) nor does tag switching work

02:12 <leon-p> it seems to work logically, as the views on screen loose focus, but it does not end up on screen

02:12 <leon-p> totally weird

02:19 <leon-p> ifreund: the problem occurs when a layout crashes while handling a layout demand.

02:19 <leon-p> I believe the output.active_layout value may be stuck

02:19 <leon-p> it fixes itself when a new layout is used

03:15 <ifreund> leon-p: interesting, thanks for investigating. I should have time to take a look tomorrow

08:25 <novakane> So I made a quick scrip to bind to a key and send a notif https://paste.sr.ht/~novakane/ea64f608d66dd669a1694eb8f3a55eec47ff817c

08:26 <novakane> works good with river-tag-overlay

08:26 <novakane> I'll see if I like it as a replacement of a bar

08:59 <nib9888> Has anyone here had any issues with intel iris? I keep getting permission denied errors from river, but strace doesn't actually show any system calls erroring with permission denied

09:01 <novakane> can you share the full errors?

09:01 <nib9888> Hold on, I can't copy from a tty, give me a sec to write them out

09:02 <nib9888> [backend/drm/atomic.c:35] connector eDP-1: Atomic commit failed (modeset): Permission denied

09:02 <nib9888> I run river with chmod +s because it won't work with either elogind or seatd

09:07 <novakane> hmm I don't know what's the modeset error, what's your problem with elogind and seatd it should works fine

09:08 <nib9888> Hold on again - Need to recompile wlroots (gentoo system)

09:11 <nib9888> It's just as simple as "libseat_open_seat: Backend 'seatd' failed to open seat, skipping"

09:11 <nib9888> With seatd running as root in another terminal

09:13 <nib9888> Just some extra information: Sway doesn't work with seatd either, but does work with chmod +s

09:16 <novakane> looks like modeset is the backend failing to set your monitor resolution/refresh rate

09:16 <novakane> for seatd I needed to add my user to _seatd group

09:16 <nib9888> I'll try that

09:17 <nib9888> I suspect it might be that, because I just found a permission denied error in strace for connecting to /run/seatd.sock

09:17 <novakane> also launch river with dbus-run-session river

09:19 <nib9888> dbus-run-session alone didn't help, but I'm working on getting write permissions to /run/seatd.sock

09:19 <nib9888> That worked

09:19 <nib9888> Thanks

09:20 <nib9888> Seems gentoo's seatd ebuild is a bit broken, as it didn't provide a seatd group, and had the socked owned by root:root

09:22 <novakane> ah yeah that's not ideal

09:22 <nib9888> I still can't help but wonder why river didn't work with +s, but sway did

09:22 <nib9888> Oh well, thanks for the help!

09:23 <novakane> no problem

12:13 <leon-p> novakane: thanks again for tha patch. I think you are overcomplicating it a bit. Instead of an anchor struct, try a global `enum zwlr_layer_surface_anchor anchors`, set to 0 initially.

12:13 <leon-p> and you should check the input. The only allowed values are 0 and 1.

12:15 <novakane> yeah I always overcomplicate things... I'll try that

12:15 <novakane> yep I want to add a check, but I was trying to make it works first tbh

12:17 <Guest32> i want to modify rivertile to use tag specific values, like location etc. But i dont get a simple way to get the focused tag, i want something like a oneliner without listening to events

12:20 <Guest32> i cleaned status.zig so i only get the focused tag but i think it whould reqire something like a second thread to run in.

12:21 <leon-p> Guest32: The layout_demand event sends the currently active tags

12:21 <leon-p> In Wayland it is generally a good idea to read the protocol before writing code

12:21 <leon-p> there is no need to use the river_status protocol

12:56 <Guest32> thanks a lot it worked perfectly, but why i cant acces it inside .set_fixed_value

12:56 <Guest32> in layout demand i cant use ev.tags but there it doesnt work

12:57 <novakane> leon-p: hell yeah it finally works!

12:57 <novakane> looks better for you? https://0x0.st/-2jr.patch

12:57 <novakane> need to add the check and margins back now

13:35 <leon-p> Guest32: because that event does not send the tags. You'll have to store the ones you get from the layout demand

13:41 <leon-p> novakane: way better. A few small nitpicks: sscanf can fail; the thing with anchor_vertical and anchor_horizontal is overcomplicated, you can just use `|=` on the global anchors variable; I would prefer the option to be called anchors; I don't like the anchors being set in the render function, set them when the surface is created; You do not need to give get_anchors() a pointer to the anchors variable,

13:41 <leon-p> since it's global; get_anchors should be called parse_anchors

13:47 <novakane> leon-p: alright I'll make the changes

13:47 <novakane> also I'm not sure how to check the input, without a lot of if else

13:52 <novakane> the surface is created in update_surface()?

14:09 <leon-p> novakane: yes, the surface is created in update_surface(). You can put the anchor setting right under zwlr_layer_surface_v1_set_size()

14:10 <leon-p> and for checking the input, let's just say 0 is no and any postive integer is yes, makes it a bit easier

14:12 <novakane> alright I changed everything else I just need to check the input and I should be done

14:13 <novakane> damn 995 lines, 5 more and you need to rewrite in zig :P

14:24 <novakane> damn I tried a negative value for the margins and my pc crash

14:24 <novakane> not ideal :P

14:37 <leon-p> :D

14:38 <leon-p> probably implicitly converting from int to uint somewhere. I probably should make the compiler flags more aggressive

14:39 <novakane> well anchors use uint and margins use int and that pissed me off to check the values, can't use the same thing :p

14:53 <novakane> leon-p: here where I am and I can't make some good check for now tbh https://0x0.st/-2jt.patch

14:55 <leon-p> novakane: it's fine, I can add the tests myself

14:56 <leon-p> I'll commit that, thanks for the patch

14:57 <leon-p> novakane: by the way, you probably want to teach your editor to respect the indentation. It used spaces almost everywhare despite the surrounding code being indented with tabs

15:01 <novakane> oops yeah sorry I never used tab

15:01 <leon-p> it's fine, I'll quickly ammend the commit

15:02 <novakane> thanks

15:02 <novakane> I really like having it at the bottom, look really good now

15:02 <novakane> :D

15:14 <leon-p> novakane: cleaned up your patch and pushed it to master, thanks for the contribution

15:18 <novakane> leon-p: nice! oof 997 lines we're good :P

16:17 <Guest32> if i get my tag value saving working should i implement a flag for it so it could be maintreamed or is there no interest

16:19 <leon-p> Guest32: I remember ifreund saying that rivertile upstream is not intended to have that feature.

16:19 <leon-p> but you can fork it

16:58 <ifreund> Yeah I don't want to maintain that complexity upstream, the official rivertile is feature complete as far as I'm concerned

17:02 <Guest32> k no problem