05:49 <throwthecheese> Is there a way to set the calibration matrix of the device via libinput?

05:50 <throwthecheese> For context: my stylus tip and eraser are not present in my device list at startup due to how the Goodix touchscreen kernel driver works

05:51 <soreau> https://wayland.freedesktop.org/libinput/doc/1.8.2/udev_config.html

05:52 <throwthecheese> This would not be that much of an issue on Wayland, but it is on Xorg as I can't change the orientation of the stylus root device

05:52 <soreau> then you're probably asking in the wrong channel

05:54 <throwthecheese> My hypothesis is that Wayland targets this root device but it's impossible to rotate it via xinput/libXi as doing so would result in a BadMatch or a BadValue error

05:56 <throwthecheese> So I'm trying to bypass it via using libinput but calling libinput_device_config_calibration_set_matrix() has no effect, even if the application is ran with root privileges

05:58 <throwthecheese> Udev also just don't cut it as as far as I know, there is no way in its API to edit properties

06:03 <throwthecheese> Full disclosure: I'm trying to get automatic screen rotation work seamlessly with my stylus on X but I can't afford to wait for its component devices to be added. Sure I can reload my device list periodically to get my stylus tip and eraser picked up but that has horrible latency, even if I use a 1ms interval or renice the screen rotator task

06:13 <soreau> that means you're definitely asking in the wrong channel

06:23 <whot> throwthecheese: yes, you can set it via the matrix as soreau linked to. at least mutter had a bug though that it would just overwrite the udev setting with the unity matrix, you'll need GNOME47 for that

06:24 <whot> throwthecheese: but if you're using it for dynamic screen rotation then that is the wrong approach, that matrix is for static brokenness, e.g. a touchscreen panel mounted upside down

06:26 <whot> throwthecheese: for the xorg bits best to file a bug against xf86-input-libinput to see what we can do for that use-case, I don't know off hand if that's possible (it *should* be with the wacom driver since that one just muliplexes everything)

11:17 <wlb> wayland-protocols Merge request !344 opened by () xdg-toplevel-drag: clarify offset is relative to the toplevel geometry https://gitlab.freedesktop.org/wayland/wayland-protocols/-/merge_requests/344

11:17 <kchibisov> soreau: this channel is also for general _libinput_ support (at least according to title).

12:07 <pq> yes, libinput is on-topic here

15:33 <vnd> after upgrading weston from 8 to 10 (together with entire BSP, yocto kirkstone), weston-screenshooter stopped working

15:33 <vnd> strace is showing that it's stuck in the endless loop of recvmsg/sendmsg

15:34 <vnd> XDG_RUNTIME_DIR and WAYLAND_DISPLAY are set correctly (weston-terminal runs fine). Any ideas?

15:47 <vyivel> vnd: run with WAYLAND_DEBUG=1 and check what's it doing

15:47 <vyivel> (run weston-screenshooter*)

16:02 <vnd> vyivel: wl_callback@8.done(76), wl_display@1.sync(new id wl_callback@8), wl_display@1.delete_id(8) are looping indefinitely

16:02 <vyivel> must be waiting for something

16:03 <vyivel> maybe check where weston-screenshooter is calling wl_display_roundtrip() in a loop

16:03 <vyivel> and keep debugging from there

16:06 <daniels> vnd: which platform?

16:09 <soreau> vyivel: does weston-screenshooter still operate on a priv'd proto? i.e. you can only invoke it from weston, not manually?

16:10 <vyivel> no idea

16:11 <soreau> at least that's how it was in it's inception

16:18 <daniels> yeah, that is true, though you can use --debug to have it exposed to arbitrary clients

16:22 <vnd> daniels: i.mx6 embedded board, yocto kirkstone, DRM backend

16:22 <vnd> weston is running with --debug, weston-screenshooter is invoked from SSH session with XDG_RUNTIME_DIR and WAYLAND_DISPLAY set correctly

16:23 <vnd> not sure what other options for invoking it are there in principle

16:23 <daniels> hmm, that's odd indeed - is it running on the drm display?

16:23 <daniels> btw we have much newer versions than 10

16:23 <vnd> yes

16:24 <vnd> we've just switched to yocto kirkstone, which is current LTS (one of the two)

16:25 <vnd> the only thing that differs is that meson is backported, so it's newer version than usually, but not sure if it can somehow result in this issue

16:25 <vnd> will try to narrow it down a bit in weston-screenshooter source code

17:05 <ManMower> does it work with the pixman renderer? which renderer are you using now?

18:38 <vnd> ManMower: there's only gl renderer out of the box

18:39 <daniels> vnd: —use-pixman