Hi, > OK, so the odd thing then is the check for !remote_display earlier > on in the function (missing from the quote above) which seems to end > up initializing VNC (albeit with localhost) when CONFIG_VNC is > defined, but no other local display type has been selected.
remote_display is set in case either -vnc or -spice was given on the command line, and qemu will not start a local user display in that case (unless you explicitly ask for it via -display). > OK, so neither SPICE or VNC really fit into this patchset because > they are seen as additional to the display type, not the actual > display type - SPICE is even more special too given the dependency > on a guest VM driver to really function correctly. Hmm? spice works just fine with any display. > I wonder should there be a, similar, qemu_display_find_remote_default() > which would be used when no local display is defined. And that would > use a similar mechanism to permit selecting VNC or SPICE via > registration mechanisms over rather than the #ifdef style for remote > displays that will remain ever after this patchset (certainly could > be another patchset beyond this one). Well, historically qemu tried to do alot of stuff (like picking a display to use) automatically for the user. And we did stick to that for backward compatibility reasons. But this automagic causes quite some headache now and then, when changing things and the autmagic logic gets into the way. So I don't feel like adding more of this. Also vnc is in pretty much any qemu build, because it is enabled by default and doesn't need any external dependencies, so falling back to spice would be quite rare ... cheers, Gerd
