|
Bugzilla – Full Text Bug Listing |
| Summary: | “vncviewer -nocursorshape” makes Xorg, Xvnc, and vncviewr use all available CPU time | ||
|---|---|---|---|
| Product: | [openSUSE] openSUSE 11.0 | Reporter: | Mike Fabian <mfabian> |
| Component: | X11 Applications | Assignee: | Reinhard Max <max> |
| Status: | RESOLVED INVALID | QA Contact: | Stefan Dirsch <sndirsch> |
| Severity: | Normal | ||
| Priority: | P5 - None | ||
| Version: | Factory | ||
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Found By: | Development | Services Priority: | |
| Business Priority: | Blocker: | --- | |
| Marketing QA Status: | --- | IT Deployment: | --- |
|
Description
Mike Fabian
2008-04-11 16:22:16 UTC
I start a vncserver like
vncserver -geometry 1280x1024 :2
I made ~/.vnc/xstartup an empty file to exclude that the problem
is caused by anything I start automatically.
Then I connect with
vncviewer -nocursorshape :2
And now Xvnc, vncviewer, and Xorg use up all availlable CPU time:
16869 mfabian 20 0 71392 17m 2392 S 28.6 0.9 0:18.50 Xvnc
16896 mfabian 20 0 49448 7988 1920 R 28.6 0.4 0:18.34 vncviewer
646 root 20 0 289m 104m 8252 S 21.6 5.2 38:25.44 Xorg
It looks to me like this is what you get from disabling a performance improvement feature. See the manpage section on -nocursorshape in the vncviewer manpage. Why are you using -nocursorshape at all? I think I misread the man-page. I wanted to improve performance over a
low bandwidth line and thought that "-nocursorshape" makes things
uglier but faster.
I still find the text in the man-page confusing:
-nocursorshape
Disable cursor shape updates, protocol extensions used to handle
remote cursor movements locally on the client side (TightVNC-spe-
cific). Using cursor shape updates decreases delays with remote
cursor movements, and can improve bandwidth usage dramatically.
So that means that for best performance over a slow line I should
*not* use "-nocursorshape"? If yes, my bug report is INVALID.
I understood just the opposite.
And what about this option:
-x11cursor
Use a real X11 cursor with X-style cursor shape updates, instead of
drawing the remote cursor on the framebuffer. This option also dis-
ables the dot cursor, and disables cursor position updates in non-
fullscreen mode.
Should I set it to make things faster over a low bandwidth line or
not?
(In reply to comment #3 from Mike Fabian) > So that means that for best performance over a slow line I should > *not* use "-nocursorshape"? Yes, that's how I understand it. When strace'ing vncviewer -nocursorshape, I saw that its eating of CPU cycles is mostly due to the fact that it chatters with the server all the time. > If yes, my bug report is INVALID. Will do. (In reply to comment #4 from Mike Fabian) > Should I set it [-x11cursor] to make things faster over a low > bandwidth line or not? Don't know if this only means the client saves some CPU cycles on not having to draw the cursor or if it has a way to stop the server from sending the data at first place. |