[Spice-devel] [PATCH] Adding new ioctl for updating Vdagent state

[Christophe Fergeau]

On Tue, Aug 09, 2016 at 05:00:37PM +0300, Dmitry Fleytman wrote:
> 
> > On 9 Aug 2016, at 16:03 PM, Christophe Fergeau <cfergeau at redhat.com> wrote:
> > 
> > Hey,
> > 
> > On Tue, Aug 09, 2016 at 03:22:52PM +0300, Dmitry Fleytman wrote:
> >> 
> >>> On 9 Aug 2016, at 14:23 PM, Christophe Fergeau <cfergeau at redhat.com> wrote:
> >>> 
> >>> On Tue, Aug 09, 2016 at 09:22:15AM +0300, Sameeh Jubran wrote:
> >>>> Hi,
> >>>> 
> >>>> What do you think that needs to be done?
> >>>> The patches should be applied?
> >>>> The interface of the device should be extended to support broadcasting the
> >>>> mouse mode to the driver? ( I think would require to modify the spice
> >>>> protocol )
> >>>> The interface of spice-gtk should be extended to support broadcasting the
> >>>> mouse mode to vdagent?
> >>> 
> >>> The filtering of cursor commands depending on server/mouse mode is
> >>> already done in spice-server in
> >>> https://cgit.freedesktop.org/spice/spice/tree/server/cursor-channel.c#n438
> >>> so I don't think we need to know too much about client/server mode in
> >>> the guest driver. Maybe some VGA/QXL distinction will be needed to avoid
> >>> some VGA-specific issues, but when using QXL, removing EnablePointer()
> >>> altogether should do the right thing both in client and server mode (did
> >>> not test).
> >> 
> >> Hi Christophe,
> >> 
> >> For some reason mouse movements are less smooth when we do not filter out mouse events on driver level.
> >> I suspect this was the reason Sandy changed EnablePointer() behaviour of the driver.
> >> 
> >> Do you have any idea why could this happen?
> > 
> > Nope no idea, maybe too many events get sent? Adding the new ioctl
> > while asking me this question at the same time sounds like we are just
> > papering over an issue we don't understand.
> 
> Our investigation shows that this happens due to too many events indeed.
> I thought, perhaps you have seen similar issues on other platforms (pre-win10/linux)
> and have additional information.

Nope, adding a bunch of printfs to spice-server or spice-gtk (eg when
you receive a motion event from the client, and when you receive
QXL_CURSOR_xxx commands from the guest) to compare the amount of events
sent by the guest should tell you whether there are differences between
the other drivers and this one. I'd expect one QXL_CURSOR_SET message
per SPICE_MSGC_INPUTS_MOUSE_MOTION message or something like that (did
not check).

Christophe
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/spice-devel/attachments/20160809/8b19fc6a/attachment.sig>