11:10 <jadahl> gschwind: so figured out the problem, and it was a problem in freerdp: you sent a bogus 'serial' to the wl_pointer.set_cursor request (it must match wl_pointer.enter). what happens is you wl_surface.attach/commit which updates the cursor sprite, but the following set_cursor is ignored by compositors so it has no effect. the reason it might have worked on some compositors is they bothered with dealing with

11:11 <jadahl> moving the cursor given the offset, but at the same time, if (I think) they implemented it correctly, the cursor would "shift" each time it changes without leaving the wlfreerdp surface.

11:12 <jadahl> gschwind: https://paste.centos.org/view/7fe6c62a should fix it, let me know if you want a PR

12:47 <zamundaaa[m]> emersion: apart from maybe updating the Mesa impl to the new naming, is there anything left to do with the tearing protocol to merge it?

13:09 <wlb> wayland Issue #337 opened by Simón (simonbcn) Implement minimize to tray https://gitlab.freedesktop.org/wayland/wayland/-/issues/337

13:10 <emersion> zamundaaa[m]: i think it's clear to land

13:10 <emersion> just waiting for a few days to see if anyone has objections

13:11 <wlb> wayland Issue #337 closed \o/ (Implement minimize to tray https://gitlab.freedesktop.org/wayland/wayland/-/issues/337)

15:15 <bl4ckb0ne> what the protocol?

15:20 <yshui`> can a misbehaving compositor crash user applications?

15:21 <jadahl> yshui`: sure it can, clients aren't uncrashable, it just needs to misbehave in a way where the client is unlucky enough to crash

15:23 <yshui`> I am thinking of scenarios. a client should be able to catch protocol errors i think? but the compositor is considered trusted, right? so clients are probably not doing much validation.

15:31 <jadahl> yshui`: right, and tbf, compositors should be fixed if they misbehave, and clients shouldn't have to have work arounds

15:31 <yshui`> yeah that makes sense.

15:52 <i509VCB> yes the client can catch protocol errors. Some client libraries also handle the scenario of invalid messages (I know wayland-rs does this), but the dispatch implementations can panic still

15:55 <gschwind> jadahl, Thanks, sure you can submit a PR, link the pr to the issue : https://github.com/FreeRDP/FreeRDP/issues/6722

15:57 <gschwind> jadahl, I did noticed this issue you seen, but in my trial fixing the serial did not solved the issue, but maybe I did not tested both change together

15:57 <gschwind> or I did not fixed it propoerly :)

16:04 <emersion> PSA, i made this, not sure if anyone's interested https://gitlab.freedesktop.org/emersion/drmdata

16:04 <emersion> pq maybe

16:05 <gschwind> jadahl, Can I share with you the code that trigger the bug of alpha privatly ?

16:06 <gschwind> jadahl, the code is fairly simple but I do not want to publish as it is

16:07 <gschwind> jadahl, you will probaly need to compile it by hand with c++14

16:10 <jadahl> gschwind: https://github.com/FreeRDP/FreeRDP/pull/8411

16:11 <jadahl> gschwind: fixed the alpha thing thanks to your fix-003 branch, that mutter missing repaint the opaque buffer into a transparent one suitable for kms

16:12 <gschwind> jadahl, in my own client, I have the bug, and as far as I know my code is correct, i.e. no attache offset and the proper serial :)

16:13 <jadahl> gschwind: if you can, find the branch from your mutter bug, it's hopefully enough

16:14 <gschwind> jadahl, I use gentoo with : x11-wm/mutter-42.5

16:15 <gschwind> jadahl, I will try another version

16:16 <gschwind> jadahl, did you tryed with scale 200% ^^ ? just in case

16:17 <jadahl> gschwind: yea, scale problem was lodpi cursor on hidpi screen

16:19 <gschwind> jadahl, sadly it seems that I cannot instal unstable mutter eassyly :/

16:19 <jadahl> gschwind: would a .patch suitable for 42.5 help?

16:20 <gschwind> jadahl, what do you mean ? with gentoo it's should by not that difficult to try a patch over 42.5

16:21 <gschwind> jadahl, if you want the I test a .patch over 42.5 I think I can

16:21 <gschwind> s/the/that/

16:22 <jadahl> gschwind: https://paste.centos.org/view/cc3046c8 should apply on top of 42.5

16:23 <gschwind> jadahl, ok for now I do not have the time to build and test, I will just try to apply the patch and test it latter

16:25 <gschwind> jadahl, the patch seems to apply thus I will try it later :)

16:25 <gschwind> jadahl, I keep you in touch

21:43 <gschwind> jadahl, I did test your patch and it's actually scale the cursor proprelly with my test client, but if I use WL_SHM_FORMAT_XRGB8888 format, mutter still use the alpha channel.

21:43 <gschwind> if my testing code is correctly writen

21:44 <gschwind> without the patch it correctly ignore alpha

22:44 <emersion> swick: color-repr is already on wl_surface atm