On Wed, 2018-10-03 at 13:14 +0200, Jiri Denemark wrote: > On Tue, Oct 02, 2018 at 17:25:34 +0200, Andrea Bolognani wrote: > > This is the best I could come up with on my own; hopefully other > > developers such as yourself will contribute to the discussion and > > we'll collectively come up with something much better :) > > Technically you could introduce a new event and report warnings as > events. That would much cleaner IMHO. Of course, apps would have to run > an event loop and register for getting such events, but not doing so is > not too different from just ignoring them when they are forcefully > pushed on clients by some ugly black magic. Your approach only works for > stuff that can be detected at a client library level, but useless in > case you'd want to deprecate something in an XML, for example. Our RPC > can only transfer a success or an error, but not both. A new event would > be useful for all cases and you could even report several warning for a > single API call. Yeah, I've run into the RPC issue and mention it in the cover letter as something that needs to be addressed before this approach can be realistically considered. The idea of using events to report warnings sounds neat! I wonder much of a limitation the fact that you'd need to have an event loop running would be in practice... How likely is it that someone will write/has written a useful, non-helloworld libvirt application that doesn't have one? Could we detect whether an even loop is running from library code and, much further down the line, refuse to execute commands unless it is? Of course then you get into the fun of warning developers that they're not going to receive warnings :) -- Andrea Bolognani / Red Hat / Virtualization -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
