Bugzilla – Bug 5760
Scrolling in using a trackpad is too sensitive
Last modified: 2018-07-19 10:59:28
You need to
before you can comment on or make changes to this bug.
We have had a report that scrolling is too sensitive in the HTML5 with a
trackpad on Mac OS X. Scrolling with a normal mouse scrool-wheel also seems a
bit too sensitive.
Could perhaps be related to Mac having a different resolution on the trackpad
which we aren't handling properly.
(In reply to comment #0)
> Scrolling with a normal mouse scrool-wheel also seems a bit too sensitive.
Disregard this, I have verified that this is not the case.
The issue with a trackpad remains unverified though. A trackpad from Apple will
I have reproduced the problem with the Apple Magic Trackpad using the HTML5
client running in Google Chrome and Safari on OSX 10.11.
Scrolling with the trackpad is extremely sensitive, it feels like maybe 10
times as sensitive in the HTML5 client compared to locally.
However, I can't reproduce the problem running the HTML5 client in Firefox on
Also seems to happen on Chromebook
(In reply to comment #5)
> Also seems to happen on Chromebook
Using the built-in touchpad that is.
Scrolling using the trackpad is way to sensitive on my laptop when running
Windows as well. No problem in Linux on my laptop however.
This issue is not limited to Web Access. It stems from the nature of the RFB
protocol where mouse wheel events are sent as clicks with a mouse up followed
by a mouse down. The protocol does not have support for wheel delta or speed.
Currently both client implementations (TigerVNC's vncviewer and noVNC) send one
mouse wheel event to the server for each wheel event no matter how small the
wheel delta is. Trackpads seem to have a tendency for sending a lot of small
delta events (delta ~= 0.01-1) while a regular mouse sends fewer larger delta
events (delta ~= 4-60). VNC usage is only suitable for the latter case really.
One way to mitigate the problem with trackpads is to accumulate a number of
small delta events in the client before sending an event to the server. A
proposed patch that does just this for noVNC can be found here: