Seen with Chrome on Windows. Was triggered with Youtube running and scrolling that browser window. The bug is in websock.js where it tries to avoid overfilling the underlying socket, but fails to restart again.
ThinLinc 4.7.0: Can't reproduce the problem nightly build: Can't reproduce the problem If there ever was a problem in between these versions, it's not there now. ¯\_(ツ)_/¯
The vendordrop for this bug breaks resize on desktop Chrome. If you resize to make the window smaller, you will end up with scroll bars.
All fixed now.
Tester should verify that mouse and resize works well now.
Resize in EDGE on Windows 10 from big to small will result in a disconnect: VNCSConnST: tl-shadow-notify: CloseAuthenticatedConnection Connections: closed: 127.0.0.1::36190 (Pixel buffer request 0x0 at 1030,880 exceeds framebuffer 782x535) EncodeManager: Framebuffer updates: 228 EncodeManager: Tight: EncodeManager: Solid: 461 rects, 3.29934 Mpixels EncodeManager: 7.20312 KiB (1:1789.98 ratio) EncodeManager: Bitmap RLE: 23 rects, 14.277 kpixels EncodeManager: 715 B (1:80.2573 ratio) EncodeManager: Indexed RLE: 747 rects, 683.402 kpixels EncodeManager: 136.849 KiB (1:19.5712 ratio) EncodeManager: Tight (JPEG): EncodeManager: Full Colour: 447 rects, 4.79015 Mpixels EncodeManager: 1.38256 MiB (1:13.2205 ratio) EncodeManager: Total: 1.678 krects, 8.78717 Mpixels EncodeManager: 1.52391 MiB (1:22.0088 ratio) ComparingUpdateTracker: 0 pixels in / 0 pixels out ComparingUpdateTracker: (1:-nan ratio)
(In reply to comment #10) > Resize in EDGE on Windows 10 from big to small will result in a disconnect: > > VNCSConnST: tl-shadow-notify: CloseAuthenticatedConnection > Connections: closed: 127.0.0.1::36190 (Pixel buffer request 0x0 at 1030,880 > exceeds framebuffer 782x535) > EncodeManager: Framebuffer updates: 228 > EncodeManager: Tight: > EncodeManager: Solid: 461 rects, 3.29934 Mpixels > EncodeManager: 7.20312 KiB (1:1789.98 ratio) > EncodeManager: Bitmap RLE: 23 rects, 14.277 kpixels > EncodeManager: 715 B (1:80.2573 ratio) > EncodeManager: Indexed RLE: 747 rects, 683.402 kpixels > EncodeManager: 136.849 KiB (1:19.5712 ratio) > EncodeManager: Tight (JPEG): > EncodeManager: Full Colour: 447 rects, 4.79015 Mpixels > EncodeManager: 1.38256 MiB (1:13.2205 ratio) > EncodeManager: Total: 1.678 krects, 8.78717 Mpixels > EncodeManager: 1.52391 MiB (1:22.0088 ratio) > ComparingUpdateTracker: 0 pixels in / 0 pixels out > ComparingUpdateTracker: (1:-nan ratio) Also reproduced using IE on Windows 10. Firefox and Chrome works as expected.
*** Bug 6171 has been marked as a duplicate of this bug. ***
These issues were actually caused by the heap overflow protection from bug 6153.
Clicking many times very fast on Chrome could cause a crash that has now also been fixed. Upstream report: https://github.com/novnc/noVNC/issues/779
Here is a summary of what i have retested: - Chrome on Windows, Scroll webpage while playing youtube hangs updates until mouse moves... - Chrome Desktop. Scale down window and scrollbars appears even if not necessary due to session resize. - Chrome Desktop, Rigth click brings context menu of browser instead of being sent into session. This breaks subsequential mouse clicks both left and right. - Chrome Windows, Mouse cursor is not changed to hand when dragging window into a session. (local cursor) - Internet Explorer, if session is scaled down in two steps using xrandr, the canvas centering is offsetetd due to confusion about size. - Edge, resize window to smaller size eg. triggering automatic session resize interrupts the connection eg. Xvnc crash. No problems found testing the above issues using browsers on Windows 10. Also, Thomas have tested click, resize and mouse on MacOS X
Regression since r32196 we get an error in Chrome on any touch devices when using two fingers on the screen: > Uncaught TypeError: Cannot read property 'dispatchEvent' of null Steps to reproduce: 1. Put one finger on the screen 2. While keeping the first finger on the screen, touch with a second finger 3. Move the first finger away
Fixed now. In addition to reproducing the issue mentioned in comment #17 and verifying that it is fixed now, the tester should verify that mouse input to the session and the control bar handle still works. Due to the nature of the changes, it is especially important to verify that clicking inside the session or on the handle and then dragging outside the browser window still works as expected. We need to verify that the above still works in a lot of combinations of browsers and platforms: iOS --- * Safari Android ------- * Chrome * FF Chromebook ---------- * Chrome Windows 10 touch ---------------- * IE * Edge * Chrome * FF Windows 10 ---------- * IE * Edge * Chrome * FF macOS ----- * Safari
iOS --- ✓ Safari
Android ------- ✓ Chrome ✓ FF Chromebook ---------- ✓ Chrome Windows 10 touch ---------------- ✓ IE ✓ Edge ✓ Chrome ✓ FF Windows 10 ---------- ✓ IE ✓ Edge ✓ Chrome ✓ FF macOS ----- ✓ Safari
