09:15 <novakane[m]> I really see no reasons why there is this problem, it only happens in river but I look into other compositor code and saw nothing different

09:16 <novakane[m]> if someone got an idea to help me find where it come from

09:18 <ifreund> I haven't had time to look closely yet, hopefully soon

09:20 <novakane[m]> do you have an idea what I could look for in priority?

09:23 <novakane[m]> gdb show that it happen in `xkb.Keymap.newFromNames(context, &rules, .no_flags)` but I don't see why we would have a null pointer here, and `&rules` ptr are optionals so

09:28 <novakane[m]> I guess it could be something from zig--xkbcommon?

09:32 <ifreund> it's possible the extern function definition in zig-xkbcommon has a typo

09:32 <ifreund> or the semantics of the xkbcommon function have changed and passing null is no longer ok

09:54 <novakane[m]> well I continue my investigation

09:56 <novakane[m]> I don't know what I could pass in the case we can't pass null anymore

10:05 <novakane[m]> doesn't look like they changed this though

15:57 <novakane[m]> Alright I'm gonna replace arch with void maybe it gonna resolve all my problem in river :p

19:28 <ifreund> wlroots release \o/

22:24 <leon-p> so what are the current blockers for the layout protocol PR? I am aware of the transaction issue (and the idea to have per-output atomic transaction instead of an atomic global one). I also really need to get around to removing the tag from the advertise view event. The issue snakedye found regarding the order of the layout demand and tag status events may also be nice to fix. Oh, and default options and

22:24 <leon-p> direction options.

22:25 <ifreund> leon-p: I'm not aware of an existing transaction issue after my commits

22:25 <leon-p> Actually the issue of the event ordering could be worked around by the layout generator, after receiving the layout demand event, first wait for the tag status event before comitting a layout

22:25 <leon-p> ifreund: that's good to hear.

22:26 <ifreund> the only blocker I'm aware of is the UX surrounding per-output options, which should be fixed by my wip modifications to river-options

22:26 <ifreund> leon-p: yeah I've been daily driving for like a month now or something :)

22:26 <leon-p> are per-output atomic commits still planned?

22:26 <ifreund> the only pain point is the riverctl options interface

22:26 <ifreund> leon-p: no

22:26 <ifreund> I decided atomicity across outputs is not that hard and worth it

22:27 <leon-p> I see

22:27 <ifreund> right now moving views between outputs isn't frame perfect though

22:28 <ifreund> just pushed the wlroots 0.13.0 update by the way, going to rebase the layout branch shortly

22:29 <leon-p> sure, I'll wait for that before removing the tags from advertise_view

22:29 <ifreund> no conflicts, let me just test to make sure nothing broke and I'll force push

22:30 <ifreund> yeah, all seems fine

22:30 <leon-p> any major changes in 0.13.0 that affect river?

22:31 <ifreund> several bug fixes!

22:32 <ifreund> also I implemented the API we need to handle xdg/layer surface popups and subsurfaces thereof properly without duplicating lots of code

22:34 <ifreund> I should probably just build wlroots from source for the CI tbh...

22:37 <leon-p> by the way, is zig-waylands scanner intended to eventually also scan the license and the description? I was playing around with turning it into a man page generator (Xlib functions are documented that way as well, which was a great help) but found those to be missing.

22:38 <ifreund> leon-p: I'd love for it to scan all the documentation text from the xml and convert that into comments/doc comments in the generated zig code

22:38 <ifreund> It just hasn't made it to the top of my todo list yet

22:42 <leon-p> wait, I think I actually had a valid reason to have the tags in the advertise_view events. I'll leave it in for now.

22:48 <leon-p> oh, are custom modes now working?

22:50 <ifreund> leon-p: yeah probably, though I didn't test

22:51 <ifreund> I'm off to bed now though o7

22:52 <leon-p> night!