06:06 zzag: coucouf1: check whether rebuilding kwin with the newer libwayland helps. I suspect that there are two wl_shm_interface symbols and the wrong one (outdated) is being used
06:48 dottedmag: Yeah, the error message is a dead giveaway for ABI problems
07:04 q234rty: I don't think just a rebuild will fix it, I think kwin would still use the outdated one generated by the wayland.xml bundled by qtwayland
07:04 zzag: it doesn't use those
07:18 coucouf1: yes, rebuilding fixes it
07:18 coucouf1: isn’t that an ABI break though ?
07:18 coucouf1: here’s the symbols diff caught by our tooling https://paste.debian.net/1324304/
08:55 zzag: coucouf1: I think there are two wl_shm_interface symbols: one in libkwin (generated by qtwaylandscanner), and one from libwayland-server
08:56 zzag: my guess is that wl_display_init_shm() picks up libkwin's wl_shm_interface
08:56 zzag: which has version 1
08:57 zzag: in plasma 6, kwin doesn't use wl_display_init_shm(), so I think it should not be affected by that issue
09:00 coucouf1: I can confirm plasma 6 is not affected
09:01 coucouf1: Can I add your above comments to our bug report for reference ?
09:01 zzag: sure
09:15 MrCooper: zzag: so is it a bug in libkwin for having an externally visible wl_shm_interface symbol?
09:19 zzag: MrCooper: it might be an ECM bug. `private-code` mode in `waylandscanenr` should fix that, right?
09:20 MrCooper: not sure, sounds plausible though
09:22 coucouf1: We don't track symbols on kwin, their libs are supposed to be private so I don't have that info at hand
09:30 daniels: yeah, it really shouldn't be leaking symbols into the global namespace
09:30 daniels: coucouf1: I've just pushed two commits to the shm-dso-pollution branch of daniels/wayland
09:30 daniels: coucouf1: can you please test that, with the first one, KWin only exposes wl_shm v1 but doesn't die, and that on the second one, KWin exposes wl_shm v2 and still works fine?
09:36 coucouf1: Not just now but yes, will test
09:38 coucouf1: When you say kwin exposes is it enough to look for that "implemented version" log I shared earlier or do you want me to actually look at kwin's symbols ?
09:42 MrCooper: look at the output of wayland_info running on it
09:43 MrCooper: *wayland-info
09:43 coucouf1: ack
10:30 daniels: yep
10:36 karenthedorf: zzag: Good news, I wrote that PRIVATE_CODE setting for ECM and it's been merged :)
11:44 davidre: I also added PRIVATE_CODE setting to qt 6.8 due to a similar problem client side
11:44 davidre: in the past
11:47 coucouf1: daniels: do you want me to tests the 2 commits independently or the 2nd one on top of the first ?
11:47 daniels: karenthedorf: thanks for the quick fix!
11:47 daniels: coucouf1: please the first one on its own, then the combination of first & second together
11:47 coucouf1: ack
12:21 coucouf1: daniels: first patch : same crash
12:21 coucouf1: 2nd patch : works and wayland-info gives:
12:21 coucouf1: wl_shm version: 1, name: 11
12:33 karenthedorf: daniels: Ha, I wrote it a week or two ago because I was using it in a non-KDE project and found it's absence annoying.
13:09 daniels: coucouf1: hmm, that was the opposite effect I was hoping for tbh
13:32 wlb: weston Merge request !1577 merged \o/ (backend-drm: get EOTF mask and colorimetry mask from EDID https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/1577)
13:32 wlb: weston/main: Daniel Stone * 5 commits https://gitlab.freedesktop.org/wayland/weston/compare/95c3f9687016391fc73edee7978719db3020348e...fa587e3fc018730d9a42ae33172ad12f334b6354
20:28 wlb: wayland Issue #483 opened by Benjamin Otte (company) Add a #define for the protocol name/global in the header https://gitlab.freedesktop.org/wayland/wayland/-/issues/483
21:37 wlb: wayland Issue #483 closed \o/ (Add a #define for the protocol name/global in the header https://gitlab.freedesktop.org/wayland/wayland/-/issues/483)