Fix IME issues by sawka · Pull Request #2938 · wavetermdev/waveterm

sawka · 2026-02-25T23:26:06Z

No description provided.

…l still work

coderabbitai · 2026-02-25T23:26:31Z

Walkthrough

This change modifies the IME (Input Method Editor) event handling in the terminal wrapper component. It replaces a blanket suppression approach for composition events with a time-based deduplication window of 20 milliseconds. The modification introduces conditional logic to suppress data during composition only when it occurs outside this window, extends the composition state reset to clear additional tracking fields, wraps the keydown handler with composition bypass logic, and adds logging for suppressed IME events. The public API signatures remain unchanged.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	No pull request description was provided by the author, making it impossible to assess whether the description relates to the changeset.	Add a pull request description explaining the IME issues being fixed, the approach taken, and any relevant context for reviewers.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main change: fixing IME (Input Method Editor) issues, which is the core focus of all modifications in the changeset.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings (stacked PR)
📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment
Commit unit tests in branch sawka/ime

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

kilo-code-bot · 2026-02-25T23:27:22Z

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

frontend/app/view/term/termwrap.ts - IME composition handling refactored and simplified

Review Notes

This PR refactors the IME (Input Method Editor) composition handling logic:

Changes:

Extracted IMEDedupWindowMs constant (20ms) for better maintainability
Simplified handleTermData() IME logic from ~30 lines to ~10 lines
Updated resetCompositionState() to reset all composition-related fields
Added isComposing check in attachCustomKeyEventHandler to prevent key events during composition

Logic:
The new approach suppresses data events when isComposing=true unless they occur within 20ms of a compositionend event. This handles the edge case where CapsLock input method switching causes spurious isComposing=true events immediately after composition ends.

Verification:

The logic is sound and correctly handles the composition lifecycle
State management is properly reset in resetCompositionState()
The refactoring maintains the same behavior while improving code clarity

No bugs or issues detected.

sawka · 2026-02-25T23:29:41Z

addresses #2915 and #2523

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@frontend/app/view/term/termwrap.ts`:
- Around line 366-373: The dlog call inside the IME suppression branch is
logging raw IME payloads via data; change it to avoid exposing user text by
logging only metadata (e.g., data length, timeSinceCompositionEnd,
this.isComposing state, and IMEDedupWindowMs) or a masked representation instead
of the verbatim data. Update the logging call in the block where isComposing is
checked (referencing isComposing, lastCompositionEnd, IMEDedupWindowMs, and the
dlog invocation) to include those safe fields and remove the raw data value.

ℹ️ Review info

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between ba7aea7 and a419cc8.

📒 Files selected for processing (1)

frontend/app/view/term/termwrap.ts

coderabbitai · 2026-02-25T23:30:16Z

+        // IME fix: suppress isComposing=true events unless they immediately follow
+        // a compositionend (within 20ms). This handles CapsLock input method switching
+        // where the composition buffer gets flushed as a spurious isComposing=true event
        if (this.isComposing) {
-            dlog("Blocked data during composition:", data);
-            return;
-        }
-
-        // IME Deduplication (for Capslock input method switching)
-        // When switching input methods with Capslock during composition, some systems send the
-        // composed text twice. We allow the first send and block subsequent duplicates.
-        const IMEDedupWindowMs = 50;
-        const now = Date.now();
-        const timeSinceCompositionEnd = now - this.lastCompositionEnd;
-        if (timeSinceCompositionEnd < IMEDedupWindowMs && data === this.lastComposedText && this.lastComposedText) {
-            if (!this.firstDataAfterCompositionSent) {
-                // First send after composition - allow it but mark as sent
-                this.firstDataAfterCompositionSent = true;
-                dlog("First data after composition, allowing:", data);
-            } else {
-                // Second send of the same data - this is a duplicate from Capslock switching, block it
-                dlog("Blocked duplicate IME data:", data);
-                this.lastComposedText = ""; // Clear to allow same text to be typed again later
-                this.firstDataAfterCompositionSent = false;
+            const timeSinceCompositionEnd = Date.now() - this.lastCompositionEnd;
+            if (timeSinceCompositionEnd > IMEDedupWindowMs) {
+                dlog("Suppressed IME data (composing, not near compositionend):", data);
                return;


⚠️ Potential issue | 🟠 Major

Avoid logging raw IME input payloads.

Line 372 logs user-typed text verbatim. Even under debug logging, this can expose sensitive data/PII. Log metadata only (length/timing/state).

🔒 Suggested safe logging change

- dlog("Suppressed IME data (composing, not near compositionend):", data); + dlog("Suppressed IME data (composing, not near compositionend)", { + length: data.length, + timeSinceCompositionEnd, + });

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed. In `@frontend/app/view/term/termwrap.ts` around lines 366 - 373, The dlog call inside the IME suppression branch is logging raw IME payloads via data; change it to avoid exposing user text by logging only metadata (e.g., data length, timeSinceCompositionEnd, this.isComposing state, and IMEDedupWindowMs) or a masked representation instead of the verbatim data. Update the logging call in the block where isComposing is checked (referencing isComposing, lastCompositionEnd, IMEDedupWindowMs, and the dlog invocation) to include those safe fields and remove the raw data value.

…v#3164) xterm.js v6.x defers compositionend's final text via setTimeout(0). On fast IME input — an issue that can affect any CJK IME — a next-keydown can reach the PTY before the deferred IME text, producing reordered output. Two-pronged fix: 1. term-model.ts: drop keydown events while event.isComposing or keyCode === 229, so xterm doesn't forward the composition trigger key to the PTY during composition. 2. termwrap.ts: on compositionend, immediately send e.data via sendDataHandler and queue it in pendingImeDedup. handleTermData drops the matching string when xterm's deferred setTimeout(0) onData fires it again. As a best-effort additional guard, reset xterm's internal _isSendingComposition flag so the deferred callback becomes a no-op when the field name isn't minified. The dedup queue (string[]) supports overlapping composition cycles for very fast input. History: This is the same class of bug fixed in wavetermdev#2938, but the patches added there were intentionally removed in wavetermdev#3095 ("upgrade xterm.js to v6.0.0") under the assumption that xterm v6 would handle composition correctly. v6.0.0's CompositionHelper still defers final-text onData via setTimeout(0), so the race resurfaced and was reported as wavetermdev#3164 against v0.14.4. This PR re-fixes it in a way compatible with the v6 codebase, using a dedup queue rather than the v5-era patches. Closes wavetermdev#3164

sawka added 3 commits February 25, 2026 14:15

stop wave keybindings when composing...

997662d

fix IME issues

afac1a0

dont block modifier key keypresses when composing... so Cmd-1 etc wil…

a419cc8

…l still work

sawka mentioned this pull request Feb 25, 2026

Korean IME composition not committed — characters only appear after pressing Space #2915

Closed

coderabbitai Bot reviewed Feb 25, 2026

View reviewed changes

sawka merged commit 296760d into main Feb 26, 2026
8 checks passed

sawka deleted the sawka/ime branch February 26, 2026 00:07

sawka mentioned this pull request Feb 26, 2026

change IME composing handling order to support implicitly confirming IME #2767

Closed

This was referenced May 1, 2026

fix: serialize IME composition with subsequent keystrokes hyeonjinyoon/waveterm#1

Closed

fix: serialize IME composition with subsequent keystrokes (#3164) #3264

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix IME issues#2938

Fix IME issues#2938
sawka merged 3 commits intomainfrom
sawka/ime

sawka commented Feb 25, 2026

Uh oh!

coderabbitai Bot commented Feb 25, 2026 •

edited

Loading

❌ Failed checks (1 warning)

Uh oh!

kilo-code-bot Bot commented Feb 25, 2026

Uh oh!

sawka commented Feb 25, 2026

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Feb 25, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

sawka commented Feb 25, 2026

Uh oh!

coderabbitai Bot commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Estimated code review effort

❌ Failed checks (1 warning)

Uh oh!

kilo-code-bot Bot commented Feb 25, 2026

Code Review Summary

Review Notes

Uh oh!

sawka commented Feb 25, 2026

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

coderabbitai Bot commented Feb 25, 2026 •

edited

Loading