https://bugs.kde.org/show_bug.cgi?id=525847

--- Comment #6 from David Edmundson <[email protected]> ---
I don't think GTK is doing anything wrong from what's described and I agree
there's an issue somewhere in the stack.

I'm questioning the ibus->zwp_input_method_context_v1 bridge is doing the right
thing and whether that code should be the one sending the pre-edit styling
again rather than kwin.

```
6. kwin receives `QtWaylandServer::zwp_text_input_v3::handle_commit()`
7. kwin calls `KWin::TextInputV3InterfacePrivate::zwp_text_input_v3_commit()`
-> `QtWaylandServer::zwp_text_input_v3::send_done()` ->
`zwp_text_input_v3_send_done()`
```

I don't think this is right. (even if you replace that with
zwp_text_input_v3_commit_string)
I can't see why we would call send_done in response to something from the
text_input client only we only do it because of something from the input_method
client.

What I suspect what is presumably happening is:

6.0 kwin receives `QtWaylandServer::zwp_text_input_v3::handle_commit()`
6.1 kwin sends inputContext->commit_state
6.2 *IBUS does stuff and repsonds to this commit  and sends a preedit*
6.3 kwin recieves inputContext->preedit string
7. kwin calls `zwp_text_input_v3_preedit_string()` ->
`zwp_text_input_v3_send_done()`

Based on the input-method docs, should ibus not be the one resending the
preedit styling before the new preedit string.

FYI, if you can export this 'KWIN_IM_WAYLAND_DEBUG' kwin will dump traffic
between ibus and kwin.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to