Hi, On 01/13/2013 03:51 PM, Marc-André Lureau wrote:
Hi Why not be consistant with QXLMonitorsConfig and add an id field to VDAgentMonitorConfig? That's why spice-gtk didn't care much about QXLMonitorsConfig id, since VDAgentMonitorConfig didn't had one (since it was meant to reflect real hw I suppose), it couldn't set it back (unless the agent learned to deal with "sparse" monitors in some ways). I agree it could have been done in the first place. Patch looks good otherwise.
Because extending VDAgentMonitorConfig breaks the agent protocol, granted we could have 2 versions of the struct and select which version to use depending on the capability being present, but that is a pain, also how does the agent know which version the client is using? We would need to specify that the new version of VDAgentMonitorConfig only gets send when *both* sides have the capability, and that it then *must* be the new version, but that is a lot of complication for something which simply stuffing 0x0 sized monitors in the array for disabled heads solves very nicely. Also note that stuffing 0x0 sizes monitors in the VdAgentMonitor array is exactly what remote-viewer expects of the monitors property of the display-channel, so there is already precedent for doing things this way. Regards, Hans' _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel