Commit Graph
192 Commits
Author SHA1 Message Date
Netcatty-Botandsi1enteee 16376a2117 fix(terminal): use xterm synchronized-output state for clear handling (#3596)
Use xterm's native synchronized output state for viewport preservation, remove the redundant mode tracker, and exercise the production callback path in regression tests.

Fixes #3595

Co-authored-by: si1enteee <si1enteee@users.noreply.github.com>
2026-10-08 15:08:11 +08:00
1d10d1679e fix(terminal): opt-in broadcast without password protection during interactive prompts (#3491)
* fix(#3488): automated AI fix

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* fix: address Codex review on PR #3491

* test: verify broadcast password prompts on Windows desktop

* test: run desktop SSH check only in Windows job

* fix: preserve sensitive peer input across compose wake

* fix: retain compose history when sensitive peer receives nothing

* fix: protect compose history during sensitive fallback writes

---------

Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
2026-09-30 00:38:50 +08:00
陈大猫 8d2058267a fix(terminal): commit live IME transcription once (#3568)
* fix(terminal): commit live IME transcription once

xterm sends a rewritten dictation hypothesis as the interim, the full
final string, and the tail. Keep the pending textarea diff from firing
again when composition starts, and send only the part that is not
already on the PTY.

* fix(terminal): retype rewritten IME tails as separate backspaces

A shared suffix has to be deleted and typed again because the cursor
stays at the end of the line. One-character hypotheses are real
prefixes, and each backspace is its own input event so Ctrl-H hosts
still erase.

* fix(terminal): rewrite a revised IME hypothesis in place

A composition that replaces the snapshot, instead of extending it,
sends the cancelled textarea edit. Deletions are counted in graphemes
so a surrogate pair is one backspace.

* fix(terminal): send a new IME word that shares a finished prefix

A composition that starts at or after the snapshot is a new word.
Sharing the first character with text already on the PTY is not
enough to treat it as the same hypothesis.

* fix(terminal): remember a rewritten IME hypothesis in full

A 229 rewrite stored only the inserted tail, so a later
compositionend with no composition marker sent the revised
line again. Remember the whole revised textarea instead.

* fix(terminal): preserve IME rewrite state across delayed commits
2026-09-29 23:07:10 +08:00
陈大猫 bd8476fb34 fix(terminal): enable font ligatures from the terminal setting (#3557)
fontLigatures already defaulted to on, but the terminal never loaded
the ligatures addon, so Fira Code and Maple Mono kept drawing != as two
cells. Activate the addon after open and before WebGL, and rebuild the
WebGL atlas when the setting changes so glyphs are not rasterized
without the calt feature.

The addon bundles lru-cache's node:diagnostics_channel import. Alias
that builtin to a no-op channel in the Vite renderer build and in the
Electron live-test bundles.

Closes #3543
2026-09-29 17:34:00 +08:00
OOKAMISAMAandbincxz 6ece7e8222 feat(terminal): Add configurable terminal bar cursor width (#3495)
* feat(terminal): Add configurable terminal bar cursor width

Adds a configurable bar-cursor width for the terminal so thin vertical cursors are easier to find, particularly when vi/vim switches to insert mode.

* Add terminal bar cursor width setting

This adds the terminal bar cursor width control to the settings UI with an aria-label for accessibility and a matching test. It also adds localized labels/descriptions and search catalog entries so the setting is discoverable in both English and Chinese.

* Fix terminal cursor width override bug

Prevent width-only cursor updates from clearing DEC/vi cursor overrides by tracking the last applied shape/blink snapshot. The runtime now resets the snapshot on teardown, skips full cursor reapplication for width-only changes, and preserves remote bar/underline cursor styles. Added regression tests covering boot-time and later width updates.

* Fix terminal cursor preference lifecycle

Preserve DEC/vi cursor overrides during the first width-only terminal update after runtime creation. The runtime now records the user cursor preference snapshot when a term is published and applies width-only updates without clearing active cursor style or blink overrides. This prevents regressions when terminal settings change before the panel has fully booted.

---------

Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
2026-09-29 00:24:27 +08:00
GOUKI9999andClaude Opus 5.5 e3894eb850 terminal: allow Shift+Enter text fallback inside ConPTY Win32 input mode (#3464)
* terminal: allow Shift+Enter text inside ConPTY Win32 input mode

Win32 input mode (PR #3247) delivers Shift+Enter to ConPTY as a lossless
INPUT_RECORD, which CLIs reading console records natively (Codex, Grok) can
use. Node/Bun CLIs read input through libuv, which drops modifier state when
translating INPUT_RECORD to VT bytes, so Claude Code and CodeBuddy still see a
bare CR and submit instead of inserting a newline (anthropics/claude-code#92771,
reproducing on 2.1.278).

While the mode is active, Netcatty also skips its Shift+Enter text fallback and
Kitty encoding, so shiftEnterNewlineText had no effect on local Windows
sessions at all.

Add shiftEnterForceText (default false): when enabled, Shift+Enter sends the
configured text even inside Win32 input mode, and consumes the keydown so no
native record is emitted as well. Add `\e` (ESC) to the escape decoder so the
text can carry the legacy Alt+Enter encoding, which is the only sequence that
survives the ConPTY input path.

Defaults keep the #2974 behaviour untouched for Codex/Grok.

* docs: record libuv input findings and the Win32 input mode opt-out

* fix(terminal): don't swallow Shift+Enter when there is no text to send

The Shift+Enter send-text interception resolved its text after already
calling preventDefault/stopPropagation, and returned false outside the
`if (shiftEnterText)` guard. When resolveShiftEnterText() yields an empty
string (e.g. a persisted empty setting) the keydown was therefore consumed
with nothing written, and the handler never fell through to the Win32
bookkeeping or Kitty paths below, so no Enter reached the session either.
The result was indistinguishable from a dead feature.

Resolve the text first and only enter the interception block when it is
non-empty; empty text now falls through exactly as it did before the
`shiftEnterForceText` setting existed, so a normal Enter is still produced.
The opt-out semantics for the non-empty case are unchanged.

Also tighten the source-shape test so it verifies the real intent: the
interception (preventDefault through return false) must live inside the
empty-text guard, which fails against the previous code.

* fix(terminal): scope Shift+Enter force-text to Win32 mode, honor it per broadcast target

Address review on #3464:

- The force-text opt-out could override a negotiated Kitty encoding that
  preserves Shift+Enter in non-Win32 sessions. The keydown decision now
  lives in shouldClaimShiftEnterForText: under Win32 input mode it follows
  shiftEnterForceText, otherwise only the existing collapse fallback
  applies. Covered by a regression test for a non-Win32 session whose Kitty
  encoding already preserves the chord.
- Broadcast Win32 targets returned the native record without consulting
  their own shiftEnterForceText. They now send their configured text for
  Shift+Enter when opted in, whether the source forwarded a key chord or a
  native Win32 record, and drop the paired release so ConPTY never sees a
  key-up without its key-down. Targets without the opt-out keep the native
  press/release pair. Tests cover source and target settings that differ.

ShiftEnterEvent now accepts KittyKeyboardEvent's optional type/modifier
fields so broadcast events can be checked with the same helper.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-28 23:36:10 +08:00
Netcatty-Botandnetcatty-bot ff65339e04 fix(terminal): keep IME preedit commits alive when Shift ends composition (#3444)
Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
2026-09-21 14:56:48 +08:00
cb06857480 feat(terminal): confirm before pasting multiple lines (#3398) (#3400)
* fix(#3398): automated AI fix

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix: address Codex review on PR #3400

* fix(terminal): protect popup and sanitized multiline pastes

* fix(terminal): preserve input state across paste confirmation

* fix(terminal): retain serial character-mode paste echo

* test(terminal): cover paced input write options

* fix(terminal): retain telnet echo for paced pastes

* test(terminal): scope serial buffer assertion to serial sessions

* fix(terminal): align runtime write options with bridge contract

* fix(terminal): preserve bookkeeping for all paced pastes

* fix(terminal): record confirmed paste lines in order

* fix: record paced paste commands after transport writes

* fix: omit paste receipts for filtered input

* fix: cancel superseded paste batches across input paths

* fix: classify each acknowledged paste line before recording

* fix: stop paced paste when editing buffered serial input

* fix: preserve recording pause boundaries and serial edit cancellation

* fix: treat paced terminal pastes as user input

* fix: preserve paste drafts until write and snippet broadcast origin

* fix: retain filtered paste drafts and invalidate urgent input ownership

* fix: broadcast paced user paste only on guarded line receipts

---------

Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
2026-09-18 15:19:34 +08:00
0d21e37a19 fix(terminal): map macOS Command+. to interrupt running command like Ctrl+C (#3409)
* fix(#3408): automated AI fix

* fix: address Codex review on PR #3409

* fix: address Codex review on PR #3409

* fix: address Codex review on PR #3409

* fix: address Codex review on PR #3409

* fix: address Codex review on PR #3409

* fix: address Codex review on PR #3409

* fix: address Codex review on PR #3409

* fix: preserve physical releases for Command-period interrupts

* fix: allow Command-period with selected terminal text

* style: keep Command-period additions ASCII-only

* test: validate Command-period fixture without renderer errors

* fix: support logical period across macOS keyboard layouts

---------

Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
2026-09-18 14:11:05 +08:00
9082aa08a4 fix(serial): auto-login on serial consoles with saved host credentials (#3418)
* fix(#3417): automated AI fix

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: address Codex review on PR #3418

* fix: guard serial quick login and reconnect startup commands

* fix: preserve serial startup while tabs sleep

---------

Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
2026-09-18 14:10:58 +08:00
bincxz a0d032ec3c fix(terminal): guard password prompts and bypass directory completion for Command arrows 2026-09-12 17:16:38 +08:00
bincxz 28f7193c4a fix(terminal): keep Command arrow edits local during broadcast 2026-09-12 17:03:51 +08:00
bincxz 51d4d97419 feat(terminal): support macOS Command arrows for line navigation 2026-09-12 16:48:31 +08:00
keroroyukiandbincxz 15b16ab7b2 fix: serial backspace leaves hidden chars for multi-byte CJK input (#3325)
* fix: serial backspace leaves hidden chars for multi-byte CJK input

Serial terminals process backspace at the byte level: one 0x7F removes
one byte from the device input buffer. For multi-byte CJK characters
(UTF-8: 3 bytes, GB18030: 2 bytes), a single backspace leaves orphan
bytes that render as hidden or garbled characters.

Fixes three layers of the problem:

1. Character mode (createXTermRuntime.ts): when backspace is received
   and the command-buffer's last character is wider than one byte,
   send as many backspace bytes as the character occupies on the wire
   so the device deletes the whole character.

2. Local echo (serialLocalEcho.ts): formatSerialLocalEcho now accepts
   a backspaceCells parameter so CJK ideographs (2 terminal cells)
   erase both cells instead of just one.

3. Line mode (serialLineInput.ts, Terminal.tsx): backspace local echo
   now erases the correct number of cells based on the deleted
   character's display width.

Adds serialCharMetrics.ts with getCharDisplayWidth (East Asian Width)
and getCharByteLength (UTF-8 / GB18030) utilities.

39 new and updated tests all pass.

* fix: address codex review — surrogate pairs, effective encoding, cursor guard, broadcast

Address all 5 P2 Codex review comments on PR #3325:

1. GB18030 4-byte supplementary chars: CJK Extension B+ (cp > 0xFFFF)
   now returns 4 bytes instead of 2, matching iconv-lite's GB18030
   encoder.

2. Effective session encoding: add currentEncodingRef to runtime
   context so toolbar encoding changes propagate to backspace byte
   length calculation (not just the saved host.charset).

3. Cursor-movement guard: track lastInputWasPrintable so backspace
   byte expansion only fires when the cursor is at the end of the
   typed buffer, not after arrow-key escape sequences.

4. Surrogate-pair-safe slicing: getLastChar/removeLastChar helpers
   handle UTF-16 surrogate pairs (emoji, CJK Extension B) so slice(0, -1)
   never leaves an orphan high surrogate.

5. Broadcast expansion: when backspace is expanded for the source
   serial session, the same expanded data is sent to broadcast peers
   instead of the original single-byte backspace.

Adds isPrintableInput, getLastChar, removeLastChar helpers.
56 tests pass.

* fix: address codex review — scope, bracketed paste, cursor confidence

Address 3 new Codex review comments on PR #3325:

1. P1: isExpanded/outData were declared inside the character-mode else
   block but referenced after it (ReferenceError in serial line mode).
   Moved declarations before the if/else so both branches share scope.

2. P2: Bracketed paste (\x1b[200~...\x1b[201~) starts with ESC so
   isPrintableInput returned false, losing tail confidence. Now detected
   and lastInputWasPrintable set to true.

3. P2: Printable input after a cursor movement reset lastInputWasPrintable
   to true even though the cursor was not at the buffer end. Now the flag
   is kept as-is when printable text is typed — only escape sequences
   reset it to false.

* fix: establish cursor-tail confidence for initial and post-submission typing

Address Codex review: lastInputWasPrintable started false, so the first
printable input after session start or line submission would keep it
false, defeating the backspace byte expansion guard.

- Initialize to true: cursor is at the end of an empty buffer at
  session start and after each line submission.
- Reset to true after Enter (handledSubmittedInput), Ctrl+C, and
  Ctrl+U — all three clear the command buffer, placing the cursor
  at the beginning of a fresh line.

This restores the original fix for the common case: type 你, press
Backspace, device receives 3 bytes and deletes the whole character.

* fix: address codex review — broadcast safety, paste cursor state, hibernation wake

Address 3 new Codex review comments on PR #3325:

1. Broadcast safety: stop forwarding expanded backspaces to broadcast
   peers. Each peer has its own protocol, encoding, and command buffer,
   so the source's repeated DELs can over-delete on non-serial peers
   (SSH character-level) or on serial peers with different encodings
   (GB18030 needs 2 DELs, not 3). The source session still receives
   the expanded data; broadcast peers get the original single DEL.

2. Bracketed paste cursor state: no longer unconditionally set
   lastInputWasPrintable=true on paste. After a cursor movement the
   paste inserts at the moved cursor, not the buffer tail, so the
   flag is now preserved as-is (false stays false).

3. Hibernation wake: initialize lastInputWasPrintable conservatively
   when commandBufferRef is nonempty. A hibernated session retains
   its buffer but the pre-hibernation cursor may have been moved,
   so we start with false instead of true.

* fix: grapheme-safe backspace + gate expansion behind byteOrientedBackspace

Address 2 new Codex review comments on PR #3325:

1. Remove the whole grapheme on backspace (P2):
   getLastChar/removeLastChar only handled UTF-16 surrogate pairs, so a
   decomposed grapheme like e + U+0301 returned only the combining mark
   and left the base. Rewrite both with Intl.Segmenter (grapheme
   granularity) so surrogate pairs, combining marks, variation
   selectors, and ZWJ sequences are removed as one unit. getCharByteLength
   and getCharDisplayWidth already compose correctly for graphemes.

2. Gate byte expansion on byte-oriented line editing (P2):
   Character-aware serial endpoints (e.g. a UTF-8 Linux console with
   readline) delete one whole character per DEL, so repeating the
   backspace 3x for a 3-byte char over-deletes. Add an explicit
   serialConfig.byteOrientedBackspace flag (default true, preserving the
   original fix) that users can disable for character-aware editors.
   Wired through the serial connect modal and host details panel, with
   i18n in en/zh-CN/zh-TW/ru/es, vault host validation, and session
   restore/snapshot preservation.

170 tests pass.

* fix: xterm grapheme width for local echo + always compute cell width

Address 2 new Codex review comments on PR #3325:

1. Use xterm's grapheme width for local-echo erasure (P2):
   getCharDisplayWidth returned 1 for emoji-presentation graphemes
   (e.g. ❤️, 1️⃣, ©️) whose base characters fall outside the standard
   emoji range. The 15-graphemes Unicode provider active in
   createXTermRuntime renders these as 2 cells. Now import
   stringCellWidth from terminalStringCellWidth.ts (which uses
   _core.unicodeService.getStringCellWidth when available) in both
   character mode (createXTermRuntime) and line mode (Terminal.tsx,
   serialLineInput). Add VS16 (U+FE0F) detection to both the xterm
   fallback and the domain getCharDisplayWidth so emoji-presentation
   graphemes are correctly 2 cells even without xterm.

2. Compute local-echo width when byte deletion is disabled (P2):
   When byteOrientedBackspace is false, the entire backspace block was
   skipped, leaving backspaceCells at 1 and erasing only half of a
   2-cell CJK/emoji glyph in local echo. Restructured the condition:
   display width is now always computed when backspacing a non-empty
   buffer, while wire-byte expansion is independently gated on
   byteOrientedBackspace.

182 tests pass.

* fix: read byteOrientedBackspace from session snapshot, not live host

Address Codex review (P2): a restored or hibernated session's
snapshotted serialConfig.byteOrientedBackspace can differ from the
current vault host (e.g. session saved false, host later changed to
true). The runtime read the live host instead of the snapshot, so
after recreation a character-aware session could start expanding DEL
and over-delete again.

Add serialByteOrientedBackspace to the runtime context (mirroring
serialLineMode and serialLocalEcho) and pass it from the session
snapshot in useTerminalEffects. Default true preserves the existing
fix for byte-oriented devices.

182 tests pass.

* fix: carry serialByteOrientedBackspace and currentEncodingRef into hibernation wake context

Address Codex review (P2): xTermRuntimeContextRef.current in Terminal.tsx
omitted serialByteOrientedBackspace and currentEncodingRef, so when
hibernateHiddenTabs woke a serial tab the runtime fell back to default
values (byte-oriented=true, host charset). A session snapshot with
byteOrientedBackspace=false would re-enable byte expansion after waking,
and a toolbar-selected encoding would be lost.

Added both fields to the wake context alongside serialLocalEcho and
serialLineMode, mirroring the useTerminalEffects mount path.

182 tests pass.

* fix: tail confidence after line submission, GB18030 4-byte BMP, free-form encodings

Address 3 Codex review comments (all P2):

1. Preserve tail confidence after line submission:
   After Enter submission sets lastInputWasPrintable = true, the same
   \r/\n went through the generic control-char branch and immediately
   reset it to false. This disabled byte expansion and correct local-echo
   width for every subsequent command. Now skip the reset when
   handledSubmittedInput is true.

2. Measure GB18030 bytes with the backend encoder:
   getCharByteLength assumed every BMP non-ASCII character is 2 bytes.
   But BMP characters outside the GBK character set (e.g. U+0100 'Ā')
   encode to 4 bytes in GB18030. Added isGbkCharacter() to check the
   GBK character set ranges; non-GBK BMP characters now return 4.

3. Use the backend charset for free-form serial encodings:
   currentEncodingRef is now always present (hibernation wake fix), so
   the ?? fallback to host.charset never triggered. For a Shift_JIS host
   with no toolbar override, currentEncodingRef.current is 'utf-8' (the
   default), causing UTF-8 byte lengths (3 for あ) instead of Shift_JIS
   (2). Now prefer host.charset when it's non-UTF-8 and the toolbar
   encoding is the default. Also added Shift_JIS support to
   resolveWireEncoding and getCharByteLength (1 byte ASCII, 2 bytes
   non-ASCII).

182 tests pass.

* fix: honor remembered UTF-8 when counting serial bytes

Address Codex review (P2): the previous fix treated every
toolbarEnc === 'utf-8' as an unselected default and overrode it with
the host charset. But a GB18030 host with a remembered UTF-8 preference
has its backend session synced to UTF-8 (via shouldSyncTerminalEncodingOnAttach),
so using the host's GB18030 charset for byte counting is wrong.

Now use resolveTerminalEncodingFromCharset to check if the host charset
is recognized by the toolbar system (UTF-8 or GB18030). For recognized
charsets, the toolbar encoding is authoritative (it may reflect a
remembered override). For unrecognized charsets (e.g. Shift_JIS), the
toolbar encoding falls back to the default (utf-8), so the host charset
is the correct wire encoding.

193 tests pass (including terminalEncodingPreference.test.ts).

* fix: honor toolbar changes for free-form charsets + GB18030 4-byte CJK Ext A

Address 2 Codex review comments (both P2):

1. Honor toolbar changes for free-form serial charsets:
   When a Shift_JIS session starts and the user later selects UTF-8 in
   the toolbar, setSessionEncoding switches the backend to UTF-8, but
   the runtime continued using host.charset because it is outside the
   two-option resolver. Now pass userPickedEncodingRef to the runtime
   context (mirroring existing use in shouldSyncTerminalEncodingOnAttach).
   The effective encoding logic:
   - Non-utf-8 toolbar → trust toolbar (explicit selection)
   - utf-8 toolbar + userPickedEncoding → trust toolbar (explicit utf-8)
   - utf-8 toolbar + recognized charset → trust toolbar (remembered override)
   - utf-8 toolbar + unrecognized charset → use host.charset (backend default)

2. GB18030 4-byte for CJK Extension A:
   isGbkCharacter classified U+3400–U+4DBF (CJK Extension A) as 2-byte
   GBK characters, but iconv-lite encodes this block as 4 bytes in
   GB18030. Removed the range; these characters now correctly return 4.

193 tests pass.

* fix: simplify effective encoding logic with backendSynced flag

Follow-up to previous fix: the 4-level decision tree had an edge case
where a Shift_JIS host with remembered UTF-8 would use host.charset
instead of the synced toolbar encoding. Simplified to a 3-level logic
using backendSynced = userPicked || hasRemembered:

- backendSynced OR recognized charset → trust toolbar encoding
- NOT synced AND unrecognized charset → use host.charset (backend default)

Added hasRememberedEncodingRef to runtime context (alongside existing
userPickedEncodingRef) so the logic mirrors shouldSyncTerminalEncodingOnAttach.

193 tests pass.

* fix: sum grapheme code points, remove Latin-1 from GBK, generic multi-byte fallback

Address 3 Codex review comments (all P2):

1. Sum every code point in non-UTF-8 graphemes:
   GB18030 and Shift_JIS returned after the first non-ASCII code point
   instead of counting the complete grapheme that getLastChar returns.
   ❤️ (U+2764 U+FE0F) is 8 GB18030 bytes, but the old code returned 4.
   Now accumulate byte lengths for all code points, like the UTF-8 path.

2. Stop classifying Latin-1 as two-byte GBK:
   U+00A1–U+00FE was in isGbkCharacter, but iconv-lite encodes these
   as 4 bytes in GB18030. Removed the range; these characters now
   correctly return 4 bytes.

3. Count bytes for every backend-supported serial charset:
   Unknown encodings fell back to UTF-8 byte length, but the backend
   uses iconv-lite for Big5, EUC-JP, Johab, etc. (2 bytes for CJK).
   Changed fallback to 1 byte ASCII + 2 bytes non-ASCII, matching
   iconv-lite's multi-byte encoding for common serial charsets.

Comment 4 (remembered override for free-form charsets) was already
fixed in commit fca4e7c0: backendSynced = userPicked || hasRemembered.

193 tests pass.

* fix: Shift_JIS half-width Katakana as 1 byte (iconv-lite 0xA1-0xDF)

Address Codex review (P2): half-width Katakana characters like ア
(U+FF71) are encoded as single bytes (0xB1) in Shift_JIS by iconv-lite,
but the previous code counted every non-ASCII character as 2 bytes.
Backspace sent 2 DELs, deleting the character and one preceding byte.

Now handle the three 1-byte non-ASCII cases in Shift_JIS:
- Half-width Katakana (U+FF61-U+FF9F, mapped to 0xA1-0xDF) → 1 byte
- ¥ (U+00A5, mapped to 0x5C) → 1 byte
- All other non-ASCII → 2 bytes

Self-review: traced all 4 encoding paths (UTF-8, GB18030, Shift_JIS,
unknown fallback) × all edge cases (single/multi-code-point graphemes,
ASCII/BMP/supplementary, GBK/non-GBK ranges). All paths verified correct.

193 tests pass.

* fix: deduplicate terminal width import after main merge

* test: retain input safety checks with serial erase metadata

* fix: restore serial editing after paste and document byte mode

* fix: keep pending serial input on its original encoding

* fix: restore serial backspace after urgent interruption

* fix: retain editable serial snippet input for backspace

* fix(serial): synchronize pending snippet tail state

* fix(serial): unify empty-line input state across text sources

* fix(serial): track editable startup snippets

* fix(serial): share pending-tail tracking for multiline input

---------

Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
2026-09-12 13:10:52 +08:00
陈大猫andnetcatty-bot 35ed0b6dbd fix(terminal): preserve row boundaries in alternate history preview (#3297)
* fix(terminal): preserve row boundaries in alternate history preview

* fix(terminal): retain columns across history row moves

* fix(terminal): keep cursor-only moves empty and overwrite whole glyphs

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix(terminal): consume unprintable wide glyphs in history capture

* fix(terminal): preserve deferred wrap between display chunks

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix(terminal): retain blank rows when reflowing history

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix: address Codex review on PR #3297

* fix(terminal): refresh cached history after resize trims text

* fix(terminal): synchronize history with actual resized buffer

* fix(terminal): preserve final wide glyph during nonwrapping overflow

* fix(terminal): restore saved cursor modes in history

---------

Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
2026-09-07 11:11:10 +08:00
陈大猫 5d61462b20 fix(terminal): keep full-width history preview rows visible (#3304)
* fix(terminal): align history preview with rendered cell dimensions

* fix(terminal): fit wide history rows without changing copied text

* fix(terminal): select every history preview row
2026-09-07 10:30:38 +08:00
陈大猫 4ba196510d fix(terminal): avoid duplicate broadcast keypress text (#3292) 2026-09-07 10:30:07 +08:00
bincxz 7451fed016 fix(terminal): preserve modified keys in Windows ConPTY 2026-09-02 17:58:52 +08:00
bincxz f2ec2a4842 fix(terminal): respect Shift+Enter negotiation 2026-09-02 16:01:10 +08:00
bincxz be63f6de88 fix: address Codex review on PR #3204 2026-08-29 14:39:21 +08:00
bincxz cf84e61451 fix(terminal): keep autocomplete popup on the wrapped command start
Fixes #3061
2026-08-29 14:21:46 +08:00
bincxzandClaude a3380d26e8 fix: base raw-paste splitting on the sanitized payload
Co-Authored-By: Claude <noreply@anthropic.com>
2026-08-27 15:58:21 +08:00
bincxz 5dbede1368 Merge commit '08044ee98' into pr-3156-resolve
# Conflicts:
#	components/terminal/runtime/createXTermRuntime.ts
#	components/terminal/runtime/shiftEnterText.test.ts
#	components/terminal/runtime/terminalImeTextInput.test.ts
2026-08-27 15:43:24 +08:00
bincxzandClaude d809b6c3fa fix(terminal): send IME commits and short pastes per character
Strict bastion prompts (QAX/奇安信) treat one SSH channel write as a single keystroke and silently drop every multi-character chunk, so an IME commit of two or more hanzi and a Chinese paste both reach the bastion's search prompt as nothing at all. Single characters work, which is why typing them one at a time is the only path that works today.

Split plain multi-character text at the send boundary instead of adding a setting: the write side of handleTerminalInputData now iterates one chunk per Unicode code point, and all bookkeeping (command buffer, script recorder, autocomplete, sudo autofill, broadcast, scroll) still sees the original payload exactly once. IME commits split whenever the committed text has more than one code point; raw (non-bracketed) onData splits up to 32 characters and longer pastes keep the single write. Anything containing an escape sequence - bracketed paste wrappers, CSI-u composition, forwarded Kitty key sequences, cursor reports - is never split, and broadcast peers still receive the original string once.

Each chunk carries the same writeToSession options as the original single write, so the sensitive marker and automation flags are kept per character. Bridge writes stay one channel write per IPC call, which reproduces what Xshell sends on the same bastion (#3077).

Co-Authored-By: Claude <noreply@anthropic.com>
2026-08-27 14:35:27 +08:00
陈大猫 65d334ff8d Merge pull request #3165 from binaricat/fix/alternate-screen-history-preview
fix(terminal): scrollable history preview inside the alternate screen
2026-08-27 14:14:36 +08:00
bincxz 101b686dbd fix(terminal): release the deferred press after an unrelated keyup
When the punctuation keyup was dropped and a previously held key was released
next, the recovery flushed the deferred key (recording a Kitty press) but then
released only the unrelated key, leaving the deferred press forwarded until
focus loss.

Replace the sentinel-rewrite predicate with resolveDeferredKeyupRelease, which
enumerates all three cases: sentinel releases take over the deferred key's
release identity, the deferred key's own release pairs the press directly, and
an unrelated release keeps its identity while the deferred press gets a
synthesized release.
2026-08-27 13:51:58 +08:00
bincxz 4f3ca3ec9d fix(terminal): drop the stale IME deferral on modified keydowns
A modified keydown (Ctrl+C, Alt+...) is a command, not the continuation of the
lost punctuation keystroke. Leaving the deferral armed across it meant the next
ordinary keydown flushed the pending character in front of the user's input,
right after the interrupt or shortcut ran.

Discard the stale deferral when the next keydown carries Ctrl/Alt/Meta; an
active composition still resolves it via insertText.
2026-08-27 13:20:27 +08:00
bincxz 00d1c819cf fix(terminal): pair the stale-flush press with a synthesized release
The keydown recovery for a swallowed punctuation keyup flushes the deferred
key, which records a Kitty press locally and for broadcast peers. Since that
branch only runs when the release will never arrive, the press stayed
forwarded and the TUI saw the key held until focus loss or session reset.

Extract the forwarded-press release pairing into releaseForwardedKittyPress and
call it from the recovery with a release synthesized from the deferred physical
key, so the press/release pair completes in the same keystroke.
2026-08-27 13:15:30 +08:00
bincxz 5ad4eaa309 fix(terminal): only rewrite IME sentinel keyups to the deferred key
A real keyup for another held key could land while the punctuation deferral
was still armed (the deferred key's own release was swallowed). That release
was rewritten to the deferred physical key, so in Kitty report-event modes the
held key's forwarded press never got its release and the TUI saw it stuck down
until focus loss or session reset.

Restrict the rewrite to IME sentinel releases (Process/229/Unidentified/
Dead/Compose, mirroring the non-text DOM keys the Kitty encoder already
refuses to send as text). Any real release still ends the stale deferral, so
the #3103 recovery is unchanged.
2026-08-27 13:11:13 +08:00
bincxzandClaude 981807e6ca fix(terminal): keep the history preview usable inside the alternate screen
Plain wheel in screen/vim/codex still forwards arrow keys (xterm's
alternate-scroll), and shift+wheel previewed xterm's normal buffer, which
holds no scrollback once an app owns the alternate buffer - so the preview
clamped to a single screen and looked dead (#2516).

When the alternate buffer is active and the normal buffer has no rows above
the viewport, the preview now renders the session's captured output instead:
a bounded plain-text transcript kept next to the connection log capture,
after the programmatic command rewriter (masked commands stay masked) and
before the replay sanitizer, which drops alternate-screen output on purpose.
Rows are wrapped to the terminal columns with wide-character awareness and
reuse the existing wrap-flag selection joining.

Mouse-reporting apps are untouched: their wheel events still go to the app.

Co-Authored-By: Claude <noreply@anthropic.com>
2026-08-27 12:13:35 +08:00
bincxz 2f020e6513 fix(terminal): recover stuck IME punctuation deferral blocking input
The CJK IME punctuation deferral (#2833) armed on a punctuation keydown and
only resolved on an insertText remap, an exact-key keyup, or a session reset.
Windows IMEs report Process/keyCode 229 as the keyup of a key they consumed,
or drop the keyup, so the deferred key stayed armed: the deferral then blocked
every keypress, and a second "/" search in vi (or any later keystroke routed
through keypress) never reached the PTY (#3103).

The deferral now ends on any real key release (composing releases excluded, so
an active composition still resolves via insertText) and a stale deferral is
flushed when the next unmodified, non-composing keydown for a different key
arrives. The keypress block is scoped to the deferred key instead of every
keypress, so a stale deferral can no longer swallow unrelated input. A mangled
release pairs its kitty release from the deferred physical key instead of the
Process/229 event.
2026-08-27 12:10:06 +08:00
keroroyuki 08044ee987 fix: address codex review — preserve ZWNJ/ZWJ and sanitize before Kitty encoding
Two changes based on Codex bot review feedback on PR #3156:

1. Preserve ZWNJ (U+200C) and ZWJ (U+200D) in terminal input.
   These carry meaning in Persian orthography and emoji sequences;
   PTYs can process the raw bytes, so stripping them could corrupt
   filenames, args, and passwords that legitimately contain them.

2. Sanitize composition text before Kitty encoding.
   When Kitty report-all + associated-text mode is negotiated,
   encodeKittyCompositionText converts U+200B into an escape sequence
   (e.g. ESC[0;;8203u) BEFORE reaching the handleTerminalInputData
   guard. Sanitize the raw text in both composition paths
   (commitImeTextInput remap path and onData handler) so zero-width
   artifacts are stripped before encoding, not after.

Update pattern-matching tests to reflect the new variable names
(sanitizedText / sanitizedData) in the Kitty composition paths.
2026-08-26 18:29:25 +08:00
keroroyuki b27aef1cd8 fix: strip zero-width characters from terminal input to prevent hidden command chars (#3138)
CJK IMEs (notably Microsoft Pinyin / Sogou on Windows) occasionally emit
zero-width Unicode characters (U+200B ZWSP, U+FEFF BOM, etc.) when switching
composition modes. xterm.js sends these to the PTY via onData, and with the
15-graphemes Unicode version they render at width 0 — so the command line
looks normal but contains hidden characters that cause the executed command
to fail.

Add sanitizeTerminalInput() and call it at the top of
handleTerminalInputData so every input path (keyboard, IME composition,
paste, Kitty keyboard, Shift+Enter) strips zero-width formatting chars
before they reach the PTY. If the input was zero-width-only, return early
to skip all downstream processing (command buffer, broadcast, sudo autofill).
2026-08-26 17:46:32 +08:00
陈大猫andCursor c4edefdbe4 feat(terminal): pass Esc+. through for readline yank-last-arg (#3047)
* feat(terminal): pass Esc+. through for readline yank-last-arg

Autocomplete was swallowing Escape, and macOS Option+. typed ≥ instead of Meta-., so the standard shell shortcut never reached bash/zsh.

Co-authored-by: Cursor <cursoragent@cursor.com>

* fix(terminal): allow Shift when Esc+_ yanks the last argument

Underscore is Shift+Minus, so rejecting all shiftKey follow-ups dropped M-_ after autocomplete dismissed.

Co-authored-by: Cursor <cursoragent@cursor.com>

* fix(terminal): keep Esc+_ armed across the Shift keydown

Typing underscore emits Shift before Minus. Clearing the Meta prefix on that modifier event dropped M-_ after autocomplete dismiss.

Co-authored-by: Cursor <cursoragent@cursor.com>

* fix(terminal): do not treat Esc+. as typed input

Yank-last-arg rewrites the shell line. Walking the ESC+. bytes appended a literal period to the command buffer and left autocomplete's typed buffer marked reliable.

Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-18 13:59:56 +08:00
陈大猫 a2532dde35 Merge pull request #3037 from binaricat/feat/3019-osc-desktop-notifications
feat(terminal): show desktop notifications for OSC 9/777/99
2026-08-18 10:26:50 +08:00
bincxz f44e200874 fix(terminal): scope Select All copy-on-select pulse to one pane
A split workspace must not treat a sibling SearchAddon match as a
user selection when Select All runs in another terminal.
2026-08-18 10:00:53 +08:00
bincxz d7f12834df fix(terminal): keep Select All and touchcancel on copy-on-select
User-invoked Select All / Select Word now pulse the gesture so they
still copy. touchcancel releases the same way as touchend so an
interrupted touch cannot leave SearchAddon able to overwrite the
clipboard.
2026-08-18 09:56:10 +08:00
bincxz 2a52837657 feat(terminal): show desktop notifications for OSC 9/777/99
Closes #3019

Codex and other agents emit OSC 9 when a turn completes or approval
is needed, but Netcatty ignored those sequences. Register OSC 9, 777
notify, and kitty OSC 99 handlers, map them to Electron notifications,
and reuse tray focus so a click jumps back to the emitting session.
Ignore ConEmu OSC 9;n progress sequences so they do not become toasts.
2026-08-18 09:28:11 +08:00
bincxz ff1f9efe5f fix(terminal): copy preview text after xterm loses focus
#3026

Dragging on the overlay can blur xterm, so native copy never hit the
container listener and the raw wrapped rows went to the clipboard.
Listen on document while the preview is open. Also keep a selected
hard newline at a row boundary instead of dropping it.
2026-08-17 20:45:52 +08:00
bincxz 19e89934c1 fix(terminal): keep preview copy and paste-selection consistent
#3026

Select All built an element-boundary range, so wrap joining never ran
and copied extra newlines. Context-menu paste selection also left the
overlay up, unlike the keyboard path. Select the text node, accept
element-boundary offsets, and dismiss the preview before pasting.
2026-08-17 20:38:58 +08:00
bincxz 137deac897 fix(terminal): dismiss history preview on click and join wraps
#3026

The overlay had to keep pointer events so history could be copied, but
then every later click stayed on the preview. A click without a drag
now closes it and restores terminal focus. Copy also joins overlay
rows that xterm marked as soft-wrapped, so a long command does not
pick up extra newlines.
2026-08-17 20:34:22 +08:00
bincxz f4ff04897e fix(terminal): treat history preview as a selection for Ctrl+C
#3026

Urgent interrupt still asked only xterm for a selection. With history
preview text selected, Ctrl+C therefore sent SIGINT into vim before the
copy path ran. Use the same copyable-selection flag for interrupt,
shortcut passthrough, and Kitty broadcast.
2026-08-17 20:28:11 +08:00
bincxz b6b433dc88 fix(terminal): copy history preview while vim owns the screen
#3026

Shift+scroll already overlays the normal-buffer history on the
alternate screen, but the overlay ignored pointer events and any
mousedown or copy chord dismissed it. Selection now stays on the
preview so the usual copy shortcuts, context menu, and copy-on-select
can take that text without leaving vi.
2026-08-17 19:26:59 +08:00
bincxz afed0f06b5 fix(terminal): drop leftover search highlights after Esc
Related to #2980

SearchAddon can leak incremental-match decorations, and WebGL bakes
those yellow backgrounds into the glyph atlas. After Esc the handles
are gone but stained cells remain. Track search decorations on
registerDecoration, paint fill on the HTML overlay only, and resweep
after the search-bar close refit.
2026-08-14 17:45:42 +08:00
bincxz 697758e440 fix(terminal): skip sudo autofill on Shift+Enter broadcast targets
Key-chord fan-out now delivers remapped LF through handleTerminalInputData
on each peer. handlingKittyBroadcast already blocks re-broadcast, but that
made canBroadcastInput false and armed password assistance on the target.
Keep the re-fan-out guard and only arm autofill for locally originated
submits.
2026-08-14 16:07:58 +08:00
bincxz 0eff48410b fix(terminal): keep sudo autofill off during Shift+Enter broadcast
skipBroadcast only stops the raw-string fan-out so peers can re-resolve
the chord from their own buffer. Sudo autofill still has to treat that
key as a broadcast, or a main-buffer sudo/su submit arms password
assistance on the source session.
2026-08-14 15:56:06 +08:00
netcatty-bot 193e88689a fix: address Codex review on PR #2975 2026-08-14 06:04:55 +00:00
netcatty-bot a254acbc75 fix: address Codex review on PR #2975 2026-08-14 05:27:01 +00:00
netcatty-bot 5b7af4414b fix(#2974): automated Cursor CLI fix 2026-08-14 05:01:34 +00:00
陈大猫andnetcatty-bot e8e696971f fix(terminal): avoid keyword highlight flicker (#2951)
* fix(terminal): avoid keyword highlight flicker

* fix: address Codex review on PR #2951

* fix: address Codex review on PR #2951

* perf(terminal): defer highlight catch-up under load

* perf(terminal): bound deferred highlight work

* fix(terminal): bound highlight catch-up backlog

* fix(terminal): mirror history wipes in highlighter

* fix(ci): load throughput baseline from PR base

* fix(terminal): preserve highlight rebuild ordering

* fix(terminal): drain highlights before exit capture

* fix(terminal): preserve line state across clear

* perf(terminal): slice history serialization

* fix(terminal): preserve parser state on recolor

* fix(terminal): preserve protocol state on recolor

* fix(terminal): retain live state during recolor

* fix(terminal): close recolor state gaps

* fix(terminal): retain saved colors across recolor

* fix(terminal): serialize recolor side effects

* fix(terminal): color keywords in place without history rebuild

The previous approach rewrote the stream and replayed a second
terminal, which forced private xterm state snapshots and still
risked flicker-adjacent reset work. Writes now stay pristine and
only dirty cells get a new foreground. Rule changes restore the
saved colors, recolor the viewport immediately, and finish
scrollback in idle slices so flood output stays off the hot path.

Related to #2784
Related to #2879

* fix(terminal): keep in-place keyword coloring correct under load

Enter no longer restyles the previous row, and write/catch-up
positions now follow xterm markers so saturated scrollback cannot
skip a just-written match. Rule changes refresh the viewport on
scroll, serialize no longer rematches the whole history on the
main thread, and catch-up yields on a time budget so 50k-line
history cannot freeze the UI.

Related to #2784
Related to #2879

* perf(terminal): keep flood and scroll off the highlight hot path

Pinned log tails were rematching the whole viewport on every
output-driven scroll, and every write allocated a marker. Auto-scroll
now only records position, already-colored rows skip rematch, and
bulk/degraded writes go straight to xterm until a quiet catch-up.

Related to #2784
Related to #2879

* perf(terminal): merge flood writes on a 16ms burst window

Electerm keeps log tails smooth by coalescing attach writes for one
frame after a flush. Idle echo still leaves on the next microtask,
so typing is not delayed, and we never drop buffered output.

Related to #2784
Related to #2879

* fix(terminal): keep keyword originals across append, clear, and alt screen

Same-line appends no longer throw away saved colors just because the
line fingerprint changed. Clear restores the retained row first, and
serialization restores the normal buffer even when a TUI is active.
Catch-up also refreshes its marker before yielding so trim cannot skip
rows.

Related to #2784
Related to #2879

* fix(ci): stop resetting highlight catch-up on every flood write

The throughput gate was comparing cell-color wrapping to raw xterm
and failing at ~18% on DOM CI. Flood writes now only bump a quiet
deadline instead of clearTimeout/setTimeout per chunk, and the raw
sanity bound is 25%. The product check remains 10% versus main.

Related to #2784
Related to #2879

* fix(terminal): rematch recycled rows and cap wrapped catch-up

Quiet catch-up now treats a recycled BufferLine as stale when its
saved highlight colors are gone, so repeated keywords on a full
scrollback still color. Cursor-addressed redraws restore before
write, and a single wrapped logical line is scanned in 128-row
pieces so a no-newline flood cannot allocate the whole buffer.

Related to #2784
Related to #2879

* fix(terminal): accept plugin #RRGGBBAA highlight colors

Plugin decorations allow 8-digit hex. Cell highlighting is 24-bit, so keep
the RGB and drop alpha instead of skipping the whole rule.

* fix(terminal): color cells for partial grapheme matches

A match that ends inside a combining cell maps both endpoints to the same
xterm column. Color that cell instead of building an empty range.

* fix(terminal): rematch flood rows after empty-line stamps

A hot-path \r\n write stamped the following empty cursor row as current.
The next flood chunk reused that BufferLine with ERROR text, and quiet
catch-up skipped it. Fingerprint the stamped content so filled or
recycled identical rows are recolored.

* perf(terminal): avoid flood-path registerMarker after trim

Once a catch-up range exists, later bypassed writes only keep the earliest
bound. A disposed marker collapses to 0 without allocating a new one on
every saturated-scrollback chunk.

* fix(terminal): recolor the row retained by clear

xterm clear keeps the cursor row after we restore originals. Recolor that
surviving line so enabled rules still apply until the next write.

* perf(terminal): stamp no-match rows without cell snapshots

Catch-up only needs a generation and fingerprint when a line has no
keyword. Allocate typed-array originals when a cell is actually colored.

* perf(terminal): treat pinned-bottom recycle as output scroll

Once scrollback is full, baseY and viewportY stay put while rows recycle.
That was treated as a user scroll, so every flood write rematched the
visible area. Keep pinned-to-bottom updates off the rematch path.

* perf(terminal): keep quiet catch-up to a single viewport paint

Catch-up already refreshes visible slices. A trailing recolorVisible()
caused a second settle paint and tripped the atomic-repaint gate.

---------

Co-authored-by: netcatty-bot <308658023+netcatty-bot@users.noreply.github.com>
2026-08-13 21:26:16 +08:00