- Server: clamp web session adaptive quality to H264-only levels (>=Good) in EvaluateQuality and ApplyQualityLevel; Ultra/High (DIFF/RGB565) caused the browser to freeze ~1 min into a session
- Server: move session-type detection to the top of ScreenSpyDlg::OnInitDialog and skip SetWindowPlacement/EnterFullScreen for hidden web sessions, eliminating the MFC dialog flash on web-triggered opens
- Linux client: default QualityLevel from QUALITY_ADAPTIVE to QUALITY_GOOD to match Windows/macOS so the server's adaptive controller doesn't auto-upgrade to non-H264 algorithms
- Web: clear the floating quick-action toolbar on fullscreen exit so its row of buttons (RDP reset / Mouse / Close) doesn't stay pinned to the top of the page
- Web: route F11 to the remote in control mode instead of toggling local fullscreen
- Web: route Esc to the remote in control mode via the Keyboard Lock API instead of exiting native fullscreen
Server sends COMMAND_SCREEN_PREVIEW_REQ when user double-clicks an
active (non-Locked/Inactive) host that advertises CLIENT_CAP_SCREEN_PREVIEW.
Client BitBlts primary screen, encodes to JPEG via GDI+ and replies. The
existing STATIC tooltip is replaced with a self-drawn CPreviewTipWnd
showing the thumbnail above the host info text, with wide-character
rendering so the popup also works on non-Chinese servers.
- Quality tiers reuse QualityProfile pattern: PreviewProfile + 6 levels
driven by GetTargetQualityLevel (FRP-aware), with 4K/ultrawide auto
upscale on Ultra/High tiers up to min(screenWidth/4, 1280).
- Client limits to 1 in-flight capture via atomic counter to defend
against flood/DoS; Send2Server is already mutex-serialized.
- Server validates responses by reqId only (single in-flight tip);
4s arrival timeout marks "preview unavailable" without blocking the
text fallback path.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ON_NOTIFY(HDN_ITEMCLICK, 0, ...) matches the inner header control's ID,
which is 0 for both m_CList_Online and m_CList_Message. So clicks on
either list's header reach OnHdnItemclickList, which always sorts the
host list by the clicked column index.
The cross-talk has existed since the initial migration commit (5a325a2).
It went unnoticed because pre-0aa7588 both lists' headers triggered the
handler in A mode and the columns happened to align (host list cols 0..2
== IP/Addr/Location, log list also has 3 cols), so log-header clicks
appeared to "sort plausibly". After 0aa7588 only the log list's A-mode
header reached the handler, surfacing the strange "click log header
re-sorts hosts" behavior.
Guard the handler by checking pNMHDR->hwndFrom against the online list's
header HWND. Log header clicks now have no effect on the host list.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The i18n commit (0aa7588) enabled LVM_SETUNICODEFORMAT(TRUE) on the online
list. That flag also flips the embedded header control to Unicode mode, so
header notifications switch from HDN_ITEMCLICKA (= HDN_ITEMCLICK in MBCS
build) to HDN_ITEMCLICKW. The existing ON_NOTIFY mapping only handles the
A version, so clicking the column header silently does nothing.
Add a parallel ON_NOTIFY for HDN_ITEMCLICKW dispatching to the same handler.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
NIF_GUID binds the tray icon to the EXE's full path. Once a GUID is
registered for one path, Shell_NotifyIcon(NIM_ADD) silently fails for any
other path using the same GUID. This caused Debug-vs-Release builds and
service+agent dual-process scenarios to fight over the same GUID slot.
Drop NIF_GUID and the static NOTIFY_ICON_GUID; revert to the traditional
(hWnd, uID) identification. The icon is freshly registered per process
launch, no path binding, no cross-instance interference.
The NIF_GUID was leftover from the AUMID/Toast experiment that was later
reverted; only the tray-icon side of that change wasn't cleaned up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Client first packet on every sub-connection signs (clientID || timestamp ||
nonce) and waits for server ack. Server verifies signature and pins clientID
on the sub-connection ctx, eliminating IP-reverse-lookup unreliability for
NAT/localhost scenarios. Sub-conn coverage: Win 12 sites, Linux/macOS 3-4
each. Main connection keeps existing TOKEN_LOGIN flow unchanged.
Includes:
- Protocol structs sized to 512/256 bytes with reserved space for future
extensions (locale, OS info, session token, etc.)
- 5-min timestamp tolerance (Kerberos-grade replay window)
- 10-sec client wait for cross-pacific / weak-network tolerance
- Fix RemoveFromHostList side-effect ordering: MarkDeviceOffline and
m_ActiveWndW.erase now only fire when ctx is actually removed from
m_HostList, preventing sub-conn disconnects from misreporting main as
offline (regression introduced by auth-set clientID on sub ctx)
- Fix latent bug: IOCPClient::m_conn was never assigned in ctor, leaving
GetConnectionAddress() always NULL and FileManager V2 transfer's
srcClientID always 0
Breaking change: new client cannot use sub-features against old server.
New server tolerates legacy clients (no auth). Future tightening can reject
unauthenticated sub-connections via IsAuthenticated() flag.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixed two bugs when closing ScreenSpyDlg with file transfer dialogs:
1. Access violation (0xC0000005): CDlgFileSend self-destructs via
PostNcDestroy (delete this) when closed, leaving dangling pointers
in m_FileRecvDlgs map.
2. Double-free: Original code called DestroyWindow() then delete,
but DestroyWindow already triggers delete this via PostNcDestroy.
Solution:
- Store {HWND, pointer} pairs instead of raw pointers
- Check HWND validity with IsWindow() before accessing pointer
- Use SendMessage(WM_CLOSE) to let dialog self-destruct safely
- Always erase map entries to prevent accumulation of invalid data
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
