You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When typing Vietnamese text with diacritics in Orca's integrated terminal, characters are lost or corrupted as soon as I add the tone mark at the end.
A typical example using the word “chính”:
Using a Vietnamese Telex input method, I type: chinhs (which should produce the Vietnamese word “chính”)
However, in Orca’s terminal the final result is only ch — the ính part disappears after the IME commits the composition.
This happens for essentially any Vietnamese word typed with Telex: if I type the full Telex sequence (letters + tone key) so that it should become a Vietnamese word with diacritics, parts of the word are removed or changed once the IME finalizes the composed text.
This makes it effectively impossible to type normal Vietnamese text in the Orca terminal.
Steps to Reproduce
Open Orca and create/open a local terminal.
Enable a Vietnamese Telex input method on the OS (for example: Unikey/EVKey on Windows, or the built‑in Vietnamese IME on macOS/Linux) and set it to Telex mode.
In the Orca terminal, type the Telex sequence for the word “chính”: chinhs.
Observe what is rendered in the terminal.
Actual Behavior
The terminal only shows ch, or the rest of the word is otherwise lost/incorrect after the IME commits the composed text.
Similar corruption/loss happens for other Vietnamese words typed with Telex.
Expected Behavior
The terminal should correctly display the full Vietnamese word with diacritics, for example chính, and in general preserve all characters when the IME commits the composition.
Additional Info
This only happens inside Orca’s integrated terminal. Typing Vietnamese with the same IME (Telex) works normally in other apps and terminals outside Orca.
The problem appears to be related to how Orca/xterm handles IME composition and updates the buffer when the composed text is committed (especially for multi‑code‑point sequences used in Vietnamese).
This was filed against v1.4.104. The root cause — xterm's Kitty keyboard encoder canceling the printable keydown before Chromium committed the IME-transformed Vietnamese glyph, dropping everything past the un-transformed prefix (chinhs → ch) — has since been addressed. PRs #6682 (Support Vietnamese IME in terminal forwarder), #6699 (Fix macOS IME input mode detection) and #7102 added a macOS native-text forwarder: a Vietnamese input source enables forwardShortTextReplacements (terminal-ime-input-source.ts), which makes the Telex letter/digit keydown a native-text candidate (terminal-ime-native-text-candidates.ts); the forwarder then claims that keydown to bypass xterm and forwards the committed glyph from the input event straight to the PTY (terminal-ime-native-text-forwarder.ts).
A repro test drives the real modules with the exact issue word and shows every glyph of chính is forwarded intact (never ch):
pnpm exec vitest run --config config/vitest.config.ts src/renderer/src/components/terminal-pane/repro-6698-vietnamese-telex-forwarding.test.ts
4 passed. Full end-to-end confirmation still needs a live macOS app with a real Vietnamese IME, but the code path that caused the loss is remediated. If a residual report appears, likely gaps are a third-party IME whose source id lacks telex/unikey/vietnam/vni, or a tone commit delivered as insertReplacementText (forwarder only handles insertText) — both would be new, narrower issues.
Automated bug-reproduction pass. Artifact on branch nwparker/bug-repro-pass2; writeup docs/bug-reproductions/6698.md.
Activity
nwparker commentedon Jul 18, 2026
Status: NOT REPRODUCED (fixed after report)
This was filed against v1.4.104. The root cause — xterm's Kitty keyboard encoder canceling the printable keydown before Chromium committed the IME-transformed Vietnamese glyph, dropping everything past the un-transformed prefix (
chinhs→ch) — has since been addressed. PRs #6682 (Support Vietnamese IME in terminal forwarder), #6699 (Fix macOS IME input mode detection) and #7102 added a macOS native-text forwarder: a Vietnamese input source enablesforwardShortTextReplacements(terminal-ime-input-source.ts), which makes the Telex letter/digit keydown a native-text candidate (terminal-ime-native-text-candidates.ts); the forwarder then claims that keydown to bypass xterm and forwards the committed glyph from theinputevent straight to the PTY (terminal-ime-native-text-forwarder.ts).A repro test drives the real modules with the exact issue word and shows every glyph of
chínhis forwarded intact (neverch):4 passed. Full end-to-end confirmation still needs a live macOS app with a real Vietnamese IME, but the code path that caused the loss is remediated. If a residual report appears, likely gaps are a third-party IME whose source id lacks
telex/unikey/vietnam/vni, or a tone commit delivered asinsertReplacementText(forwarder only handlesinsertText) — both would be new, narrower issues.Automated bug-reproduction pass. Artifact on branch
nwparker/bug-repro-pass2; writeupdocs/bug-reproductions/6698.md.AmethystLiang commentedon Aug 1, 2026
Closing as fixed.
Merged PRs that close this:
Support Vietnamese IME and synthetic Unicode text forwarding in terminal)Fix macOS IME input mode detection)Fix ime input switch)Maintainer confirmation: filed against v1.4.104; IME forwarder path addresses the Telex character-drop root cause.