So a summary of what's been done in this fork/PR. Instead of moving the cursor, we're warp_closest ing. warp_absolute didn't work, so warp closest has kindof an auto constrain feature so it works beautifully. I'm also contraining to the right side of the window - 1, because the cursors are treating that pixel as the next pixel over, so it was actually thinking it was on the next monitor (assuming fullscreen) when the cursor got to the right or bottom of the window. TL;DR rounding issue.
This fix didn't work when you had a fullscreen video playing on a monitor beside, so I also fixed that by ignoring all focus changes or whatever when a constraint actually occured, we obviously just don't want to focus another monitor when constrained, so why not just *disable it when a constraint happened*
The PR is now ready, I love Hyprland, and I can't wait to contribute more :)
One failure case: When there is a fullscreen window directly to the right of a game, for example, in a multimonitor setup, then the cursor will bug out in between the monitors. sometimes it will get constrained to the border, but just outside the window (which is what this PR fixes, so it's unusual) and sometimes the cursor will just ignore the constraint entirely. However this is only in the one case.
The window properly constrains the mouse now
I do still notice a bug with moving the mouse in games, if you don't move the mouse fast enough, some games will not register the mouse movement. This doesn't happen in KDE so I know it's related somehow
This is a first version of my test to properly constrain cursors. This is mostly working in the buggy applications I had before, but I feel that the cursor needs to actually move around, instead of being locked to the center of the window.
This may cause problems when locking to the edge, but yeah.
meson will set add -DHYPRLAND_DEBUG to CXXFLAGS during compilation of
debug builds. this avoids NDEBUG issues with wlroots and ensures asserts
will also work on release builds.
prevents build-failure when xwayland is disabled to to
add_project_arguments('-DNO_XWAYLAND', language: 'cpp')
beeing called after the required wayland protocols have been built.
this makes for a far better experience in combination with wlroots,
since that whole makefile mess is not required.
Additionaly, handling of wayland protocol sources is also slightly
better, but could be improved with mesons inbuilt wayland module.
To build Hyprland using meson:
meson _build -Ddefault_library=static
ninja -C _build
ninja -C _build install
