08:44pq: emersion, the long threads in Mastodon between me and Troy for the past week or so are about tone mapping. I guess the main take-away is that a tone mapping algorithm is still an open research question.
09:27emersion: aha, that's what I feared
09:33pq: Tony Vladusich raised a good point: https://mathstodon.xyz/@TonyVladusich/113935668726882902
09:34pq: my interpretation is that when down-mapping dynamic range, at some point when that range compression increases the picture becomes unnatural if you try to keep all details.
09:37pq: I think that's the problem of "HDR photos" mapped to SDR present in some camera apps.
10:55JEEB: pq: yea, tone mapping is the thing that is very much not specified and everyone has their own special sauce
11:26wlb: weston Merge request !1682 opened by Robert Mader (rmader) Draft: libweston: Implement color-representation protocol https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/1682
12:51soreau: for something a bit less wild-west: github.com/soreau/wl-backdrop (wl-shm time/weather widget applet thingie)
16:05zzag: emersion: is it okay with wlroots to reuse buffer storage after destroying a wl_buffer? https://gitlab.freedesktop.org/wayland/wayland/-/merge_requests/447
16:05zzag: speaking for kwin, it won't be
16:06zzag: because kwin can upload data from a shm buffer to a texture later
16:06zzag: so we expect the client not to touch the backing storage
16:06zzag: I believe wlroots and kwin handle buffers in a similar fashion
16:08zzag: I've been thinking about suggesting that the client is not allowed to reuse buffer storage if it destroys a buffer before receiving a release event
16:10daniels: that's the only possible answer
16:10vyivel: yeah i think that's how it always meant to be
16:10daniels: it's not just delayed SHM uploads, but there could also be delayed GPU work (even if it's flushed, it's not guaranteed to have finished), overlays, etc etc
16:16zzag: okay, cool!
16:20zzag: "I've been thinking about suggesting that the client is not allowed to reuse buffer storage if it destroys a buffer before receiving a release event" ah, wl_surface.attach documentation already covers it \o/
18:26emersion: zzxyb[m]: client can release, but not reuse
18:26emersion: er
18:27emersion: zzag: ^
18:27emersion: release the backing storage, i mean
18:27emersion: e.g. on shutdown
18:27emersion: or when re-allocating a swapchain
18:28emersion: since the compositor still has a ref to the backing storage via mmap, the OS will not re-use that memory again for a new alloc until the compositor unmaps it
19:03DemiMarie: emersion: can you look at my MR for libwayland to fix the infinite loop?
19:50linyaa: hi, i'm searching for the latest in-progress protocol work for VK_EXT_present_timing.
19:50linyaa: can someone point me to the latest MRs or protocol?
19:50linyaa: sorry, i'm not skilled at navigating wayland development.
20:47ezzieyguywuf: is there a way to request from the server what size it thinks my window should be? or do I have to specify a prefered default width/height?
21:07vyivel: ezzieyguywuf: the server will tell you either way in xdg_toplevel.configure
21:07vyivel: unless the size is 0,0 in which case pick something sensible on your own
21:34Markiplier: hello?
21:53Markiplier: are you all just daemons?
22:38DemiMarie: Markiplier: why do you ask?
22:38DemiMarie: did you hear something negative about Wayland?
22:38Markiplier: erm no
22:38Markiplier: there's little activity here
22:56bl4ckb0ne: its late
23:07ezzieyguywuf: vyivel: it's telling me 0 width and 0 height lol
23:07ezzieyguywuf: unless I'm doing something w rong
23:08ezzieyguywuf: e.g. this is the log I'm currently dumping: Got CONFIGURE for xdg_toplevel, xdg_shell_client.Toplevel.Event__struct_3355{ .width = 0, .height = 0, .states = common.Array{ .size = 0, .alloc = 0, .data = anyopaque@9aa5f5c } }
23:13soreau: ezzieyguywuf: vyivel mentioned that you can choose your own size if the configure event size is 0x0
23:14ezzieyguywuf: ahh I missed that
23:15ezzieyguywuf: so 0,0 is a valid response from the server - is there a way for me to coerce the server into giving me a value? e.g. on sway (which I'm using right now), I don't understand what it means for a window to be 0,0, since there's no concept of "minimize"
23:18ezzieyguywuf: ah, ok so if I go ahead and set my own width/height, I get _another_ event (after 0,0), where sway tries to set the window to its desired dimensions
23:19ezzieyguywuf: I find that odd though I wish it would just tell me from the rip what size it wants, I hate to guess at an "appropriate" size, especially for non-tiling compositors