Code review comment for lp:~alan-griffiths/mir/client-dies-without-server

> I suppose sigterm is an ok temporary solution (and better than hanging)...
> still problematic though, because the io service has died, a properly written
> client would try to release its surfaces and disconnect on SIGTERM. A client
> can't tell whether the signal came from a mir library or something like "kill
> -SIGTERM".

Yes, and those calls now fail gracefully. (C.f. 153 etc.)

> What's final plan for informing the client the server died?
> callback in the api maybe?

Adding a lifecycle notification callback that the server connection died.