This bug is about supporting the new way of doing multi-head, via RANDR 1.2.
(The old way being Xinerema)
Rough time est.
This bug shouldn't be difficult considering both Xinerama and basic RANDR is in place, and we have only one output per CRTC.
We need to have a look at RandR 1.3 and see if panning is something that we want for VNC.
Clarification: This bug is about the server portions. Client is handled in bug 4229.
Implemented in r25475.
Tester also needs to check that old clients still work.
More outputs than necessary are being created. Need to have another look.
This is caused by how Gnome disables outputs. It not only removes the mode, but also the associated CRTC. This confuses our code.
Fixed in r25510.
We're doing something that upsets xfce's configuration tool. Need to have a look.
I'm getting a bogus "invalid screen layout" when using the OS X client. Seems to have something to do with very specific frame buffer sizes before and after the resize.
(In reply to comment #11)
> I'm getting a bogus "invalid screen layout" when using the OS X client. Seems
> to have something to do with very specific frame buffer sizes before and after
> the resize.
Fixed in r25655.
(In reply to comment #10)
> We're doing something that upsets xfce's configuration tool. Need to have a
Must have been fixed along the way. Cannot get any problems with it now.
I need some concrete testcases for this bug
Be creative! You're supposed to out-smart the guy implementing the bug ;)
- Test various randr tools (e.g. gnome's, xfce's and kde's display tools)
- Try various odd local monitor configurations and see that they are properly represented on the server.
- Reconnect and resize with various configurations to see that things transition smoothly.
- Look in xinit.log for randr related warnings/errors
- Test an old client and see how things behave when moving back and forth (e.g. reconnect to a multi-monitor session created by a new client)
Tested different layouts on client side (seperated, overlapped, in any direction) while connecting to a session, the session is successfully updated with the new layout requested, however i one changes the layout on serverside the client will not update its layout and therefore a mismatch will be introduced..
See new bug: #4419
Tested to connect against a session with multiple monitors with 3.4.0 client
which seems to work as expected, xrandr tells there is an additional VNC screen
which is not activated.
I failed to test with 3 monitors (hardware issue) as per request,
but i have done several attempts to make the server side break
and failed. I'm closing hereby this bug.