08:38 pq: sewn, I think usually the bigger problem to support another OS is to get the automated CI testing going for it, rather than getting someone to write the OS integration code.
08:39 pq: for libwayland
11:40 ifreund: hmm, is it expected that xdg_toplevel.resize_edge.none causes the compositor to ignore the resize request or to pick an edge itself?
11:41 ifreund: the protocol doesn't say anything about the meaning but doesn't make it a protocol error
11:43 vyivel: the latter i think
11:44 ifreund: from reading code, mutter ignores the resize request if the edge is none while sway interprets none as bottom_right
11:44 vyivel: if it was ignoring the client wouldn't have a reason to the request in the first place
11:44 vyivel: hm
11:44 vyivel: my compositor iirc falls back to choosing based on the pointer location relative to the window center
11:45 vyivel: i.e. same logic as user-initiated resize
11:45 ifreund: yeah, there's no reason for the none enum variant to exist in the first place if the intent is that the request is ignored
11:45 ifreund: but it could have slipped into the protocol without much scrutiny...
11:45 mclasen: you typically need a value to indicate
11:46 mclasen: 'not dragging on any specific edge, but still resizing'
11:46 mclasen: I guess you only need that in case you have constraints associated with the other values
11:47 ifreund: mclasen: gnome ignoring resize requests with "none" as the edge makes me think this isn't a situation that matters much in practice...
11:49 mclasen: Yeah, probably not. But I think mutters behavior is wrong
11:49 ifreund: I think I agree, it makes sense for the copositor to pick something reasonable here
11:50 ifreund: maybe I would have left the none enum variant out of the protocol in the first case personally but there's no changing that now
11:51 mclasen: the case to use none for
11:51 mclasen: is the "Resize" menu item in the window menu, if done client-side
11:52 mclasen: we later added a compositor-side window menu, so now none is a bit useless
11:56 ifreund: makes sense
11:56 vyivel: oh nvm wlroots straight up disconnects the client on none
11:56 vyivel: which sounds wrong
11:57 ifreund: heh, yeah that's not what the protocol says
11:58 ifreund: I guess it hasn't really been a problem though...
11:59 vyivel: fwiw weston doesn't resize the window with none too
12:00 ifreund: heh, I'm the one who broke this in wlroots: https://gitlab.freedesktop.org/wlroots/wlroots/-/commit/5ae17de23f5fd9bb252a698f3771c840280e2c05
12:02 vyivel: smithay's anvil ignores the request as well, and kwin seems to handle edge-less resize request as a move request instead
12:03 ifreund: handling it as a move seems like a bug to me
12:03 vyivel: maybe they have some common move/resize path with edges=0 meaning move
12:04 vyivel: anyway, the protocol should say at least something about all this
12:04 ifreund: I agree
12:35 any1: Maybe there ought to be a library that can set up texture importing and generate shaders for conversion from any DRM_FORMAT_*.
12:50 pq: any1, there have been some wishes in that direction but with color management. I don't know of anyone working on it.
12:56 pq: so I spent two hours on one Wayland email... HTH
12:56 any1: Yeah, I was thinking that it ought to support colour management as well
12:59 Company: "generate shaders" is always tricky to integrate with because shaders don't have API/ABI
13:01 Company: like, I have a certain setup for handling samplers and clipping an texture coordinates that needs to work with whatever that tool generates
13:01 Company: and with the drm formats, you can map the multiplane formats as multiple textures to workaround missing yuv support, and once you do that, you suddenly need to integrate with that, too
13:07 pq: plus all the shader language variants/versions
13:07 pq: maybe it becomes easier to define a compositing API instead of deliver texturing setup and shader snippets
15:06 wlb: wayland-protocols Merge request !310 opened by Isaac Freund (ifreund) xdg-shell: document resize with "none" edges https://gitlab.freedesktop.org/wayland/wayland-protocols/-/merge_requests/310
15:32 Company: pq: "shader snippets" as part of a compositing spec would be very welcome I think (no idea how big they would be of course, but GLSL is often short enough)
16:18 wlb: weston Issue #903 opened by Leandro Ribeiro (leandrohrb) Dma-buf feedback unnecessarily making clients use linear modifier https://gitlab.freedesktop.org/wayland/weston/-/issues/903