The flush-char handling backspaced and re-typed the already-on-screen
word before/around the forwarded space. In the grabbed-device injection
path this raced against the separately-forwarded space, eating spaces
and merging finished words (e.g. "mất sự" -> "mấtsự",
"đầu ngã xuống" -> "đầungãxuống").
The composed word is already correct on screen, so a non-macro flush
now finalizes state without backspace+retype:
- engine: process_key returns None on flush (macros still Replace)
- daemon replay_and_inject: just types the flush char
- daemon did_flush branch: clears state without retyping
Add regression tests for flush behavior and multi-word spacing.
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: vndangkhoa <vonguyendangkhoa@gmail.com>
Co-authored-by: vndangkhoa <60398697+vndangkhoa@users.noreply.github.com>
The flush-char handling backspaced and re-typed the already-on-screen
word before/around the forwarded space. In the grabbed-device injection
path this raced against the separately-forwarded space, eating spaces
and merging finished words (e.g. "mất sự" -> "mấtsự",
"đầu ngã xuống" -> "đầungãxuống").
The composed word is already correct on screen, so a non-macro flush
now finalizes state without backspace+retype:
- engine: process_key returns None on flush (macros still Replace)
- daemon replay_and_inject: just types the flush char
- daemon did_flush branch: clears state without retyping
Add regression tests for flush behavior and multi-word spacing.
Co-Authored-By: vndangkhoa <vonguyendangkhoa@gmail.com>
When Vietnamese mode is on, the engine transformed every word including
English (test->tét, cargo->cảgo, status->státu). This wires up the
previously-dead english.rs dictionary and spelling.rs validator so that on
word commit, words that are clearly English or not phonologically valid
Vietnamese are reverted to the raw keystrokes typed. Genuine Vietnamese
(tiếng, việt, quả) is kept. Gated by the existing [auto_restore] enabled
config (default on).
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: vndangkhoa <vonguyendangkhoa@gmail.com>
When Vietnamese mode is on, the engine transformed every word including
English (test->tét, cargo->cảgo, status->státu). This wires up the
previously-dead english.rs dictionary and spelling.rs validator so that on
word commit, words that are clearly English or not phonologically valid
Vietnamese are reverted to the raw keystrokes typed. Genuine Vietnamese
(tiếng, việt, quả) is kept. Gated by the existing [auto_restore] enabled
config (default on).
Co-Authored-By: vndangkhoa <vonguyendangkhoa@gmail.com>
- xclip -o read triggered unwanted paste in some apps
- Clipboard save/restore was over-engineering the Ctrl+C/V fix
- Simple clipboard write+paste is sufficient
- Engine no longer includes flush char in Replace insert
- Daemon forwards raw flush key after Replace injection
- Test helper simulates this by adding Insert event after Replace
- Engine no longer includes flush char in Replace insert text
- Daemon forwards raw flush key (space/enter/etc) after injection
- Clipboard paste only contains the word, not trailing whitespace
- Fixes 'thịtrâm' → 'thị trâm' (space reliably arrives via raw forward)
- xdotool type depends on keyboard layout for Unicode — fails on US layout
- Backspaces were sent AFTER text (wrong order — erased what was just typed)
- Reverted to clipboard paste which is layout-independent
- Copies xdotool from system or /tmp/xdotool-extract
- Fallback message if not found
- xdotool type is preferred for Unicode injection (no clipboard hacks)
- xdotool types text directly into X11 focus window — no clipboard hack
- More reliable than clipboard paste (no trimming, no timing issues)
- Fallback to clipboard if xdotool not available
- Only on X11 (Wayland uses clipboard fallback)
- Splitting spaces into separate uinput events caused them to arrive
after the user's next keystrokes, resulting in 'thịtrâm' (no space).
- Now paste entire text including spaces via clipboard in one operation.
- Trailing spaces may be trimmed by some apps — but this is rarer than
the timing-induced space loss from split injection.
- \n char had no keycode mapping — now sends KEY_ENTER via uinput
- Fixed in both ASCII path and trailing-ASCII-after-unicode path
- Enter now works on single press (was being silently consumed)
- skip_count=3 applied only to auto-repeat events, never to press/release
- Prevents rrrrrrrrr while letting spacebar and real typing through
- Reverted broken drain approach that corrupted source file
- After each Unicode injection, skip next 10 events (auto-repeat backlog)
- Prevents 'rrrrrrrrrrrrrr' and '555555' from auto-repeat during injection delay
- Also fixes Ctrl+V/Ctrl+C clipboard conflict during injection window
CHANGELOG: document v0.1.1 Telex fixes, injection improvements, AppImage flags
- Tone keys (f,s,r,x,j) now only apply when composition has a vowel
- Without a vowel, they fall through to normal character append
- Fixes 'r' disappearing in words like 'trời', 'trâm', 'trảm'
- Added test_telex_r_as_normal_char covering 4 scenarios
- Also: 15ms delay between clipboard paste and trailing uinput chars
- --quit: stops daemon, uinputd, xrecord, and tray
- --restart: stops all then re-launches
- GUI launch without tray shows zenity dialog with quit instructions
Only pure control keys (f,s,r,x,j,w) should be consumed silently.
Base letters used in double-letter marks (aa→â, ee→ê, etc.)
are normal typing keys that must be forwarded when no mark triggers.
Removed uppercase variants too — consolidated with to_ascii_lowercase().
- Fix Xutf8LookupString signature (missing XIC param caused all keys to map to \0)
- Port bamboo-core Vietnamese engine to Rust (bamboo.rs, input_method.rs)
- Flexible backtracking for mark/tone keys (scan up to 5 chars back)
- Correct tone placement for io, uâ, yê clusters
- Evdev capture preferred over X11 XRecord (more reliable)
- Uinput injection with correct Linux keycodes
- Vietnamese Unicode via clipboard paste + trailing ASCII via uinput
- Persistent X11 connection for Ctrl+V (no per-call dlopen overhead)
- Consume stale VNI/Telex control keys when no match found
- Fix execute_commands backspace count for evdev grabbing path
- Add vietc-uinputd privileged injection daemon
- AppImage: bundle uinputd, preserve LD_LIBRARY_PATH, fix xrecord build flags
- Remove old generated test files, add 63 focused engine tests
- Add aggressive drain loop in wait_for_event() when SKIP_RECORD_EVENTS
is true: poll 5ms + drain, repeat until quiet (up to 50ms). This closes
the timing gap where injected events arrived after drain_pipe returned
but before the flag was cleared in the next iteration.
- Remove verbose debug eprintln!/log_info from daemon (process_key,
replay, inject, toggle, window change, etc.)
- Add vietc-xrecord.c (C helper with XRecordEnableContext blocking mode)
- Update build-appimage.sh to compile and bundle C helper
XGrabKeyboard on the same display as XRecord breaks event delivery.
XRecord captures events globally without any grab needed.
Also: use XPending() before select() to check Xlib internal buffer,
and add XFlush before XRecordProcessReplies after select().
- XRecordRange is 32 bytes: core_requests(0), core_replies(2),
ext_requests(4), ext_replies(10), delivered_events(16),
device_events(18), errors(20), client_started(24), client_died(28)
- XRecordClientSpec is XID = unsigned long (8 bytes), not int (4 bytes)
- Use XRecordAllClients=3 instead of XRecordCurrentClients=1
XGrabKeyboard returns success but never delivers events to a windowless
client on modern X11. XRecord is the standard way to capture keyboard
events globally — used by xdotool, xbindkeys, and input methods.
Architecture:
- XRecord: captures all keyboard events (no grab needed)
- XGrabKeyboard: still used to block original events from reaching apps
- XTest: injects modified events
Dynamically loads libXtst.so.6 for XRecord functions.
Events flow: XRecord callback → thread-safe queue → daemon event loop.
CRITICAL BUG: XEvent was { _type, _pad[24], data } (data at offset 28)
but in X11 it's a union where ALL variants start at offset 0. This meant
event.data.key.state/keycode read from completely wrong offsets — every
keystroke produced garbage characters.
Fixed by replacing XEvent with a raw [u8; 192] byte buffer and using
event_type() and key() accessor methods that cast from offset 0.
Also fixed same bug in x11_inject.rs.
XPending returned 0 even with active keyboard grab. Using select() on
the X11 connection file descriptor with 100ms timeout to reliably detect
when the X server sends events. Also adds XSelectInput on root window
and XConnectionNumber for the fd.
Without XSelectInput, XGrabKeyboard grabs the keyboard but the X server
never sends KeyPress/KeyRelease events to our connection. Also flushes
after grab to ensure it takes effect.
- Use XPending() to check for events before XNextEvent (non-blocking)
- Add is_grabbed() and has_pending_events() public methods
- Auto re-grab keyboard when grab is silently lost (tray, WM focus)
- Fixes AppImage daemon receiving zero keystroke events
