Created attachment 131746 [details] [review]
initial unique application support

Small wart: currently, palimpsest seems to use gtk_dialog_run for some of its dialogs (e.g. the "sure you want to delete this partition" dialog). Thus, while the dialog is up, running a second instance of palimpsest will hang waiting for a response from the unique app (which wll come once the dialog is dismissed.

(In reply to comment #1)
> Created an attachment (id=131746) [edit]
> initial unique application support
> 

Committed. Thanks.

(In reply to comment #2)
> Small wart: currently, palimpsest seems to use gtk_dialog_run for some of its
> dialogs (e.g. the "sure you want to delete this partition" dialog). Thus, while
> the dialog is up, running a second instance of palimpsest will hang waiting for
> a response from the unique app (which wll come once the dialog is dismissed.

Mmm.. yeah. I don't really have any good ideas here. First of all we can't get rid of modal dialogs; we are messing with data on the hard disks so occasionally we really need to ask confirmation.

Btw, isn't it a libunique bug that it doesn't respond if gtk_dialog_run() is used? I mean, gtk_dialog_run() will still run the main loop and service the D-Bus connection...

Anyway, I think we should just dismiss dialogs (with GTK_RESPONSE_CANCEL) when getting a message from a 2nd palimpsest instance launched with --show-volume or --show-drive. This should also be documented in --help.

Thoughts?

We tentatively tracked this down to libdbus holding some lock while calling the pending-call-complete callback, preventing further dbus messages to be processed while we do the gtk_dialog_run. So, this is a dbus bug, and probably not worth working around here.