Chuchu v0.5.1, Android, using voice dictation (Gboard voice typing). Happens on any session.
When I start voice dictation while Chuchu is in the foreground, nothing gets inserted into the terminal. My dictation normally falls back to writing the system clipboard when there's no usable text field, but that fallback doesn't fire while Chuchu is foreground either, so the text is just lost. Outside Chuchu the same dictation writes to the clipboard fine.
Steps
- Open a session in Chuchu.
- Start voice dictation and speak.
Expected: the dictated text is inserted into the terminal, or failing that the dictation falls back to the clipboard like it does in other apps.
Actual: nothing is inserted and the clipboard isn't written. The text is lost.
Where it might be
The terminal uses a custom BaseInputConnection (TerminalInputView / TerminalInputConnection) that advertises a text field (TYPE_CLASS_TEXT, NO_SUGGESTIONS, with suppression/mirror logic to avoid double-sends). Because it looks like a real text field, the dictation IME targets it and never uses its own no-field clipboard fallback. But voice input comes in as streaming composing text (setComposingText), and the terminal-side commit/suppression logic doesn't seem to commit that composing text, so it gets dropped. End result: dictation neither inserts nor falls back.
Would it be feasible to either commit composing text from voice IMEs, or let the field decline input so the IME uses its own fallback? Happy to test builds if that helps.
Chuchu v0.5.1, Android, using voice dictation (Gboard voice typing). Happens on any session.
When I start voice dictation while Chuchu is in the foreground, nothing gets inserted into the terminal. My dictation normally falls back to writing the system clipboard when there's no usable text field, but that fallback doesn't fire while Chuchu is foreground either, so the text is just lost. Outside Chuchu the same dictation writes to the clipboard fine.
Steps
Expected: the dictated text is inserted into the terminal, or failing that the dictation falls back to the clipboard like it does in other apps.
Actual: nothing is inserted and the clipboard isn't written. The text is lost.
Where it might be
The terminal uses a custom BaseInputConnection (TerminalInputView / TerminalInputConnection) that advertises a text field (TYPE_CLASS_TEXT, NO_SUGGESTIONS, with suppression/mirror logic to avoid double-sends). Because it looks like a real text field, the dictation IME targets it and never uses its own no-field clipboard fallback. But voice input comes in as streaming composing text (setComposingText), and the terminal-side commit/suppression logic doesn't seem to commit that composing text, so it gets dropped. End result: dictation neither inserts nor falls back.
Would it be feasible to either commit composing text from voice IMEs, or let the field decline input so the IME uses its own fallback? Happy to test builds if that helps.