[Spice-devel] two monitors with linux and libvirtd

[David Mansfield]

On 09/26/2013 04:16 AM, Christophe Fergeau wrote:
> Hey,
>
> On Wed, Sep 25, 2013 at 01:16:21PM -0400, David Mansfield wrote:
>> Here is exactly what isn't working:
>>
>> Environment:
>> F19 host, F19 guest both fully updated, with or without virt-preview
>> repository.  F19 remote-viewer. I've also tried adjusting ram and
>> vram setting a few times but currently it's ram=131072 and
>> vram=65536.
>>
>> Issues:
>> When opening the "second" monitor via the remote-viewer menu, it
>> stays "waiting for graphics server" (or some such)
>>
>> If, in a terminal window in the vm, I run xrandr with no arguments,
>> the second "monitor" begins to display the entire desktop (two
>> monitors worth), albeit somewhat shrunken to fit into the
>> remote-viewer window.
>>
>> If I open a third monitor via remote-viewer window, the second
>> "monitor" begins to show the correct data, but the third shows the
>> entire desktop (three monitors worth), shrunken to fit.
> Fwiw, this is working fine with a f19 client/f19 host/f20 alpha live cd
> guest, could be worth testing this to try to narrow which part of the
> equation is having issues.
I've now  tried F19 and Centos6 vm/host/client in all configurations. I 
also tried the F20 alpha as you suggested.

In F20 alpha livecd opening a "monitor" works much better, but resizing 
them with the in-vm "displays" control panel still fails pretty miserably.

This crashes spice server on F20 livecd.  After you get to the desktop, 
open second "monitor" from remote-viewer menu, then:

xrandr --output Virtual-0 --mode 1920x1200
xrandr --output Virtual-1 --mode 1920x1200 --right-of Virtual-0

(Actually the above works with resolution 1600x1200 so there is 
something strange going on. I've again tried tweaking "ram" and "vram" 
setting to no avail).
>> The actual mouse cursor position is offset (by about -8 x -8) from
>> where the displayed mouse cursor is (you must move your mouse right
>> and down from where you want to click).
> I've seen that in some VMs, I think there's a bug for it in bugzilla.

This one seems isolated to F19 VM, whatever host or client is being used.

>
>> More Information:
>> I have subsequently tried doing this exact same test with Centos6
>> host, Centos6 guest, Centos6 client.  This basically works (maybe
>> some rough edges which are completely unrelated).
I retract my previous statement, BTW.  Things are "stable" in centos6 - 
but quite often when switching resolution it switches but snaps back to 
the original resolution almost immediately and it's impossible to 
consistently get a specific configuration to  work.

The goal is to get the monitors to match the resolution of my physical 
monitors, obviously.

>>
>> I have then tried F19 remote-viewer connecting to centos6 guest,
>> Centos6 host - it doesn't work, but it's better.  Changing
>> resolutions (using the "monitors" control panel in the guest)
>> produces strange and incorrect overlaps of displayed desktop area /
>> mouse positions etc., but if you set things up using centos6 client
>> then connect from F19 things basically work.  Hard to really explain
>> the problems here except that clearly the guest and client are not
>> communicating "monitor" dimensions and offsets correctly.
> This is a bit similar to what is being discussed in
> https://www.redhat.com/archives/virt-tools-list/2013-September/msg00111.html
> There's a bug describing the behaviour you see I think.
Very similar, except that xrandr seems to display "correct" values is my 
case, but the viewer is confused.

If it would be helpful, I'd absolutely be willing to enable some 
debugging logs and capture the details of various failure scenarios - I 
have a growing number of "work at home" users and I'm trying to see if 
we can make spice work for us.

Thanks,
David Mansfield
Cobite,INC.