Add support for OSC 22 control sequence to change mouse cursor shape #6292
Conversation
reykjalin
changed the title
| MuxNotification::Alert { | ||
| alert: Alert::MouseCursorShapeChanged { .. }, | ||
| .. | ||
| } => { | ||
| // fall through | ||
| } |
There was a problem hiding this comment.
I copied what's done for Alert::PaletteChanged here, not sure if that's correct.
wezterm-gui/src/termwindow/mod.rs
Outdated
| alert: Alert::MouseCursorShapeChanged, | ||
| pane_id, | ||
| } => { | ||
| // FIXME: Is there a cache that needs to be invalidated like for PaletteChanged? |
There was a problem hiding this comment.
There isn't an explicit cache, but you will need to cause the logic in the normal mouse event processing flow to trigger and assess what the cursor should be here, based on its position.
I don't have a great suggestion for this off the top of my head. Most likely will need to add a helper function to figure out if the mouse cursor is over the terminal area and if so, get the shape from the pane identified by pane_id and apply it to the window.
You and me both 😅 I think I've addressed most of the feedback, but I'm still seeing the issue where it's like the mouse shape is cached. I'll dig into that some more next week :) |
|
I think I figured out the reason the mouse cursor shape gets "stuck" on whatever was last set through the OSC 22 sequence. It's because the TUI doesn't set the OSC22 code back to the default, or doesn't clear it I guess? So once the TUI is closed whichever pointer was active is still stored in I'm not sure if that's "correct" behavior or not. I also don't know what the right time to clear the OSC 22 setting would be, if that even is the correct approach here. The xterm docs don't seem to specify that, at least not in relation to what the control sequence does. I'll keep looking into this over the next couple of days. |
| } else if let Some(shape) = pane.get_mouse_cursor_shape() { | ||
| if pane.is_mouse_grabbed() { | ||
| parse_mouse_cursor_shape(shape.as_str()) | ||
| } else { | ||
| // If the mouse is no longer grabbed by a TUI, reset the cursor shape. | ||
| pane.clear_mouse_cursor_shape(); | ||
| MouseCursor::Arrow | ||
| } | ||
| } else if pane.is_mouse_grabbed() { |
There was a problem hiding this comment.
As it turns out, this fixed the issue of the mouse cursor shape being wrong. I have absolutely no idea if this is the right way to approach it though.
But basically; if the mouse is grabbed by whichever application is running, we defer to whatever mouse shape is set by any OSC 22 escape sequences.
If we find that a shape has been set by an OSC 22 escape sequence but the mouse isn't grabbed we clear it and just use the default pointer shape instead.
Some thoughts:
- I'm realizing that using
mouse_cursor_shapefor the names of this might be confusing? Should we specify that this is a shape set through an OSC 22 escape sequence somehow?- It's more confusing in the GUI layer, less so in the terminal state itself.
- But it may be useful to use something like
osc_22_mouse_cursor_shapeinstead of justmouse_cursor_shape?
- I'm not sure this is the best place to clear the mouse cursor shape. Maybe there's an event sent when the mouse is grabbed and released that would be more appropriate?
- As stated earlier; I have no clue if this is the right way to manage OSC 22 control sequences, but at least it behaves the way I expect.
| Before this change | After this change |
|---|---|
Screen.Recording.2024-10-22.at.2.36.17.PM.mov |
Screen.Recording.2024-10-22.at.2.31.45.PM.mov |
Happy to get some guidance on any improvements here!
Description
pointer,text,row-resize,ns-resize,col-resize, andew-resize, based on the currently availableMouseCursoroptions in the codebase. Any other value defaults toMouseCursor::Arrow.Videos
Running comlink (terminal IRC client) before and after:
Screen.Recording.2024-10-17.at.11.39.06.PM.mov
Screen.Recording.2024-10-22.at.2.31.45.PM.mov
Testing
This requires a terminal program that makes use of OSC 22 codes to change the mouse cursor. The only one I know of and actively use is comlink, but I'm sure there are other TUIs out there that do use the codes, I'm just not aware of them.
Run a terminal program that uses OSC 22 codes and check if they work.
I'll try to make a terminal program to demonstrate at least some of them as I work on this PR. It should make testing easier.