XShmAttach fails asynchronously with BadAccess when the X server cannot reach
the segment (for example a server in another container); Xlib's default error
handler exited the host before the existing XGetImage fallback could run.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A session that joins a host already streaming to another viewer can receive a
clean stream with no IDR in flight, so the loss-driven PLI never fired. Ask
every 30 access units until something is presented, and warn after 300.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Route host and viewer WebSocket signaling through the shared reqwest client so UU_USE_PROXY=1 applies to signaling as well (HTTP/HTTPS proxies only, no direct fallback, no redirects, strict upgrade validation). ICE/WebRTC media is not proxied.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Count only top-level hosts (helpers such as the Wayland capture process and
the login window are children) and explain that two instances with one
identity kick each other out of the room.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The host re-execs through /proc/self/exe, so its process name becomes "exe"
and pgrep -x uuremote-server missed duplicate hosts.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Classify QR request failures (timeout/connect/decode/body/request) with phase, elapsed time and HTTP/API status only; never print tickets, URLs or server content. Extract and test the polling decisions; keep the 60 s timeout and official refresh behaviour.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Wayland: type mobile text_input via clipboard paste on Portal and compositors without spare keycodes
* Wayland: address review comments for mobile text_input clipboard fallback
- Guard external clipboard commands in unit tests to eliminate desktop side effects
- Narrow paste fallback to Wayland and Portal backends only
- Document clipboard override and two-way sync behavior in README (both zh and en)
- Unify disable switch via UU_WAYLAND_TEXT=0
---------
Co-authored-by: shofukantkts <[email protected]>
Wayland shows a toplevel only after its first buffer, so a stream that never
yields a picture left no window at all. Present a black buffer on creation and
on resize/expose while no frame has arrived.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Tiling or maximizing the SDL viewer resized the surface, but nothing was
drawn until the next decoded frame, so a static remote desktop stayed blank.
Repaint the last frame on SizeChanged/Resized/Exposed/Maximized/Restored.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
After a network change the DNS/TCP/TLS/upgrade could stall forever after
room credentials were acquired; the reconnect loop never retried. The
handshake now fails after 20s and the existing 5s retry takes over.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A tool peer promoted by capture_desktop stayed marked as the desktop after its
encoder exited on its own (Wayland capture failing while locked/DPMS off), so a
later capture_desktop on the same peer only logged and never restarted the
video. Treat a finished stream task as ended: capture_desktop asks for the
desktop again and promote() restarts the video on the existing input lease.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The ICE transport state callback was created before the DTLS transport and
captured the default placeholder, whose state never leaves New. Every later
ICE change therefore computed the peer state against a "New" DTLS: after a
brief ICE Disconnected -> Connected the PeerConnection settled on New instead
of Connected, so the idle room renewal treated the still-working
port-mapping peer as idle and tore the room down.
Wire the callback to the real DTLS transport through a weak slot filled right
after it is created.
Fixes#109
Co-Authored-By: Claude Opus 5.5 <[email protected]>
SDL MouseWheel events had no arm in the event translator, so wheel input was dropped. Send one notch as 120 wire units (the Mac convention), use SDL's precise deltas for trackpads, honour the flipped direction for natural scrolling, and keep the existing focus and view-only gating.
The CLI decoder pads every frame into a fixed 1920x1080 canvas so rawvideo frames have a known byte length. Input mapping and the SDL viewport treated that padded canvas as the picture, so non-16:9 hosts (official Macs are usually 16:10) had the pillarbox bars inside the 0..1 range: the remote pointer lagged inward and the error grew towards the screen edges.
Parse the coded picture size from ffmpeg's one-time "Video:" stream line, carry it on Frame.content, and render/map only the content sub-rectangle. 16:9 streams are unchanged; UU_VIEWER_SIZE keeps working.
* Signaling: keep heartbeat detection alive when delivery stalls
The receive loop awaited event_sender.send() without a bound. When the main loop was parked in PeerConnection::close after a released peer, the 64-slot event queue filled and the receiver blocked inside the loop body, so the Engine.IO deadline in the same select! never ran. The host stayed online with a half-closed socket for hours.
- add a watchdog over the last inbound frame, independent of the receive loop
- bound event, pong and control sends with a timeout or try_send
- timeout PeerConnection::close so one dead peer cannot stall signaling
- regression test with a blocked event consumer
Co-Authored-By: Claude Code <[email protected]>
* Signaling: always report Disconnected when the receiver stops
The receive loop can break without enqueueing Disconnected: event delivery exceeded SEND_STALL_TIMEOUT, or the writer already stopped. receiver_done then lets the watchdog exit, while the main loop's rx_event stays open because the idle-renewal timer holds a sender. The host stays in a dead room until the 15-minute renewal, which is deferred while a peer is connected.
Hand a Disconnected off from a detached task after every loop exit. Duplicates are harmless: the main loop breaks on the first one and drops the old receiver.
The new test keeps the consumer blocked past SEND_STALL_TIMEOUT, then drains the queue and expects Disconnected.
Co-Authored-By: Claude Code <[email protected]>
---------
Co-authored-by: Claude Code <[email protected]>
The 100-500 kbit NVENC default squeezes a 1440p IDR into ~62 KB, so the
picture blurs and sharpens once per GOP. Share the VAAPI buffer parser so
UU_NVENC_BUFFER_MS (10-1000 ms of the bitrate cap) applies to the FFmpeg
pipeline and the in-process engine; default unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
When a port-mapping connection is already open, the Mac viewer opens the desktop by sending SimpleAction 4 (capture_desktop) on that connection's CONTROL channel instead of a new offer. The host now promotes that tool connection to the desktop: video, cursor, TEXT clipboard (with the capability echo replayed) and session counting follow a per-peer active flag. A later desktop demotes it without closing its port mappings. Stream tasks are fenced by an abort handle, and the X11 capture is rebuilt if the screen size changed.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Opt-in X11 monitor selection for #98: RandR monitors from xrandr --listmonitors through the existing screen-switch path; x11grab reads display+x,y at the monitor size, absolute input adds the monitor offset, in-process engine and xrandr resize are skipped while a region is captured.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The WARN filter let webrtc_ice/dtls lines push the useful ones out of the
last 80 lines (#99). Also mask IPv6 and MAC addresses, and ask for the
graphical session terminal: doctor from SSH/tty reports no DISPLAY.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
set-encoder / set-codec signal the running host (only when /proc/<pid>/exe is this exact binary); the live desktop stream re-plans its encoder at the next frame boundary, sends CaptureSettingSync when 4:4:4 falls back, and prints a note when no host was notified. Step 3 of #85.
Co-authored-by: ferstar <[email protected]>
v0.2.25 showed KDE's unknown icon on Plasma 6.8 (#87). Plasma resolves
SNI icons theme by theme, trying dash-trimmed names in each theme before
the IconThemePath app theme, which comes last, so a cache-only icon is
easy to lose. Write it to $XDG_DATA_HOME/icons/hicolor (found by every
panel's default lookup; only when changed) and keep that as the theme
path. The new name "uuremote-tray-symbolic" cannot trim into the colored
"uu-remote-linux" app icon and avoids entries cached under the old name.
Checked with KIconLoader on Plasma 6.7 (Plasma's IconThemePath setup and
the global loader both resolve it, with the colored icon installed too).
Co-authored-by: VallierDev <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
#16: with display, capture and encode all on an RTX 5070 the packed-RGB
input path is 19% slower in the probe and 25% end to end than CPU
swscale, while on an iGPU-composited desktop it was neutral to faster.
Keep it on by default only when no connected DRM connector belongs to an
nvidia-bound card; UU_NVENC_RGB=1 forces it, =0 still disables it.
Co-authored-by: VallierDev <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
* Tray: monochrome icon that follows light/dark panels
The tray used the colored app icon (or the theme's "computer" when it was
not installed), which is hard to see on dark panels. Ship a 16px symbolic
icon embedded in the binary, write it to the user cache and pass that
directory as the indicator's icon theme path: GTK and the GNOME
AppIndicator extension recolor "-symbolic" names, KDE recolors the
ColorScheme-Text class. Fixes#87.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* AGENTS: config.json is parsed once, re-read only on LoginExpired
Reported in #85: the 60 s sleep is the room renewal poll, not a config reload.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: VallierDev <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
* Restart X11 capture at the new size when the desktop is resized mid-stream
A RandR resize during a session made x11grab fail with BadMatch on the old
size. read_encoder treated that as an encoder failure and walked the whole
fallback chain (NVENC -> CPU, HEVC -> H.264) with the stale size, then the
stream stopped ("Encoder output ended") and the viewer froze. New connections
already re-probed the geometry (#18); live ones did not.
- When an X11 stream fails and the root size differs, skip the encoder
fallback and restart capture at the new size with the same settings.
- Absolute pointer mapping follows the refreshed X11 size.
- New connections compare the root size with their capture instead of the
engine's change flag, which a live stream may already have consumed.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Change the X11 desktop resolution when the viewer picks one
The official Windows host changes its display resolution for the viewer's
分辨率 choice; this host only scaled the encoded stream, so a smaller choice
made text smaller and blurrier instead of larger.
On X11 the capture/resolution RPC now switches the primary output's mode with
xrandr (an existing mode of that size, else a CVT reduced-blanking mode at
60 Hz, or 30 Hz above an HDMI 2.0 pixel clock), re-applying it when mutter
resets the mode after the list changes. resolution_type 3 (original) and the
end of the last desktop stream restore the mode in use before the first
change. The stream restarts capture at the new native size and reports it in
the screen sources and CaptureSettingSync. UU_DESKTOP_RESIZE=0 keeps the old
encoder scaling; Wayland is unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: VallierDev <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
webrtc-srtp 0.12.0 stored every sequence number and used a 100-packet window to
guess the rollover counter. A NACK retransmission of a pre-wrap packet sent after
the wrap (65000 after 85) rewound that state, so the next new packets were
protected with ROC+1 and the viewer failed authentication on all of them: frozen
video after the Windows viewer returned from the background.
Vendor webrtc-srtp 0.12.0 with the RFC 3711 Appendix A half-space estimate and a
high-water index that retransmissions cannot move back. Regression tests fail on
upstream (65000 != 85) and pass with the fix for AES-CM/HMAC and AES-GCM.
Co-authored-by: VallierDev <[email protected]>
Co-authored-by: AYssu <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
README/README.en/AGENTS: AMD iGPUs (HawkPoint, Cezanne) are accepted on the inproc-av VAAPI engine with the official
Mac viewer; the FFmpeg process pipeline on AMD is not separately verified. inproc-av now covers VAAPI in process;
build headers listed. Thanks @ferstar.
With --features inproc-av on X11 and UU_VIDEO_ENCODER=vaapi, VAAPI runs in process: a DRI3 DMA-BUF of the root
(XShm upload without DRI3) goes through VPP to NV12 and the hardware encoder. SPS/PPS(/VPS) and slice headers are
packed by the host because radeonsi (Mesa 26) does not write them. libva/libva-drm are dlopen'ed; release builds do
not enable inproc-av. The NVENC path is unchanged.
Accepted by the contributor on AMD HawkPoint1 (Mesa 26.2.3), 1920x1200@60, with the official Mac viewer
(HEVC, presets, fps, scaling). Intel iHD and 4K are not verified. Thanks @ferstar.
H.264 High 4:4:4 / HEVC FRExt with chromaFormatIDC 3 and AQ off when 真彩 is requested; GPUs without YUV444 encode fail the capability check and fall back. full_chroma_available() no longer excludes the in-process path. UU_PROBE_CHROMA=444 for the probe.
Co-authored-by: ferstar <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Replaces the FFmpeg-based in-process encoder with a native NVENC engine (vendored nv-codec-headers n12.1.14.0, libcuda/libnvidia-encode dlopen'ed at runtime, primary context reuse). Fallback: native NVENC -> FFmpeg NVENC process -> libx264. inproc-av no longer links FFmpeg; VAAPI/CPU always use the process pipeline. Default builds are unchanged.
Co-authored-by: ferstar <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>