Update smithay-client-toolkit to v0.18 #5276
Conversation
Also updates:
- wayland-{client,protocols} to v0.31
- wayland-egl to v0.32
|
This appears to break the primary selection, and I'm not sure why. However, I've noticed that the update also has a new |
Also converts other DataDevice-handling code to use Read/WritePipe instead of the FileDescriptor API.
rust-analyzer can't find the standard library source code otherwise.
|
This is now complete, and the primary selection works once again. 😊 Pipe handling is a little cleaner as well, and there aren't quite as many |
|
Fixed an existing crash that occurred when switching TTYs. Looks like someone else has found and tried to fix this (#5173), but this implementation is more complete (handles loss of each different capability type, not just keyboard) and more correct (releases the associated resources). Should resolve #5154 and #5079. |
|
Fixed a rare, latent crash in code which assumed that capabilities couldn't be lost, uncovered by the previous fix. |
| match capability { | ||
| Capability::Keyboard => { | ||
| log::trace!("Lost keyboard capability"); | ||
| self.keyboard.take().unwrap().release(); |
There was a problem hiding this comment.
Please avoid the unwrap for this; I'm assuming that this will work:
| self.keyboard.take().unwrap().release(); | |
| self.keyboard.take().map(|p| p.release()); |
|
I'll fixup the unwrap and merge this, otherwise it may be a while before I get back to it! |
|
I rebased and pushed to main, thanks! |
|
Some of the issues I linked in the initial comment can be closed as well now, I don't think they get auto-closed when the commits are merged in manually and the enclosing PR closed. |
|
Same with the issues/PR linked in #5276 (comment) |
|
I closed those as duplicates, thanks for pointing me to them! |
I've taken a look and am unsure: it's possible that hyprland and river support will have been improved with the dependency bump, but I also don't use either of these compositors— and the errors reported don't seem familiar to me. I guess we'll see if people report seeing further issues. |
|
I do use Hyprland. The reported issues are familiar and were resolved by this PR :-) |
SCTK v0.18 includes some changes to how cursors are set, which fixes not being able to change the cursor when
XCURSOR_THEMEis unset (and emitting an error message about this once every frame). Probably fixes #5071 and #4604, and, I think, #5209. Needs testing on GNOME to see if #3106 or #2687 are fixed by this (GNOME unfortunately does not yet implement the cursor-shape-v1 protocol), but that's obviously not a blocker.