Pointer constraints protocol

[ids1024]

The protocol implementation should generally work, though there's some more work still needed to get the implementation in Anvil fully functional. And testing, docs, etc.

[Drakulix]

Looks good for a first draft!

[ids1024]

I've implemented pointer confinement in Anvil; confining to just the window boundary, ignoring the region.

Handling this correctly seems complicated. The logic in wlroots is slightly complicated (https://gitlab.freedesktop.org/wlroots/wlroots/-/blob/master/util/region.c#L190-256), but more significantly seems to depend on the pixman_region32_t data structure. I don't see any crates providing something similar. (I guess other compositors are all in C and using pixman, so it's not an issue for them.)

I think this is good enough to merge now, at least given the part that's incomplete is the Anvil implementation.

[codecov-commenter]

Codecov Report

Patch coverage: 14.77% and project coverage change: -0.13% ⚠️

Comparison is base (d8600f2) 22.88% compared to head (d978a2b) 22.76%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #872      +/-   ##
==========================================
- Coverage   22.88%   22.76%   -0.13%     
==========================================
  Files         143      144       +1     
  Lines       23050    23365     +315     
==========================================
+ Hits         5275     5319      +44     
- Misses      17775    18046     +271     

Flag	Coverage Δ
wlcs-buffer	19.79% <4.40%> (-0.21%)	⬇️
wlcs-core	19.45% <4.40%> (-0.21%)	⬇️
wlcs-output	8.05% <4.40%> (-0.05%)	⬇️
wlcs-pointer-input	21.56% <14.77%> (-0.11%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
anvil/src/input_handler.rs	7.33% <ø> (ø)
src/input/pointer/mod.rs	33.09% <0.00%> (-0.25%)	⬇️
anvil/src/state.rs	37.62% <8.33%> (-0.60%)	⬇️
src/wayland/pointer_constraints.rs	14.09% <14.09%> (ø)
src/wayland/seat/pointer.rs	17.40% <80.00%> (+0.77%)	⬆️

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Drakulix]

Handling this correctly seems complicated. The logic in wlroots is slightly complicated (https://gitlab.freedesktop.org/wlroots/wlroots/-/blob/master/util/region.c#L190-256), but more significantly seems to depend on the pixman_region32_t data structure. I don't see any crates providing something similar. (I guess other compositors are all in C and using pixman, so it's not an issue for them.)

I am slightly confused, because I am pretty sure we have to deal with this for input_regions of surfaces already. So I am not sure, why the same code wouldn't work here?
It probably needs logic to properly clamp to the region, but a cheap implementation could just discard cursor movements, that would result in a place outside the region. And we already have a "is point in wl_region" test for certain.

[Drakulix]

Found it: https://github.com/Smithay/smithay/blob/master/src/wayland/compositor/mod.rs#L275-L303

[ids1024]

Yeah, I was thinking of the clamping part. Just ignoring movement if the new position would be outside the region might work alright. Technically less pixel-perfect. Also with a really weird region that could jump over a tiny area that isn't in the region to another that is, but that's even less likely to be an issue (I'm not sure exactly how other implementations handle these cases).

pixman_region32_t should be a more efficient data structure in general, but hopefully applications aren't adding and subtracting 500 rectangles to a wl_region...

[Drakulix]

pixman_region32_t should be a more efficient data structure in general, but hopefully applications aren't adding and subtracting 500 rectangles to a wl_region...

Oh absolutely and yes the issues above can in theory happen. But I think implementing it this way in anvil is better, than not implementing it at all. Especially because compositors are likely going to copy&paste anvil's implementation. 😅

[ids1024]

This isn't doing any fancy pointer warping, but setting a region seems to work now, with Smithay/client-toolkit#407.

Just testing if the pointer ends up outside the region and dropping the event, instead of clamping, seems to be working alright.

[Drakulix]

LGTM