Re: [gtk-vnc-devel] disconnect signal

[Anthony Liguori <anthony codemonkey ws>]

Jonh Wendell wrote:
> Em Ter, 2007-09-11 às 20:05 +0100, Daniel P. Berrange escreveu:
>   
> > On Tue, Sep 11, 2007 at 03:55:13PM -0300, Jonh Wendell wrote:
> >     
> > > Hi, folks.
> > >
> > > What do you think about change the way the 'disconnect' signal is
> > > emitted? I think it should be raised only if the connection is lost, or
> > > if the server closes the connection.
> > >
> > > If i close the connection by myself, or if i close the whole program, i
> > > don't want to receive a 'you were disconnected' signal.
> > >
> > > Was i clear? I don't know if what i have in mind should be implemented
> > > like i said or by adding another type of signal...
> > >       
> > I intended the disconnect signal to be used to track the 'lifecycle' of
> > a connection and as such I think it should always be emitted, no matter
> > how the connection was closed.
> >     
>
> OK, no problem, but:
>
>   
> > One thing I do want to add though is a 'error' signal which would be emitted
> > whenever anything goes wrong in the client (ie anywhere the code which sets
> > the has_error flag to TRUE). In particular this signal would include some
> > kind of info about what failed - perhaps just the 'strerror(errno)' data or
> > a message generated by our own code for some semantic error eg "Unsupported
> > protocol version", or "Supported authentiction method", or 'Remote server
> > closed connection". Would a general 'error' signal solve the issue you want
> > to address ?  The 'error' signal would of course be immediately followed by
> > the 'disconnected' signal too. 
> >
> > Dan.
> >     
>
> Indeed again, but, in this case, the 'disconnected' signal is useless in
> my opinion.
>   

I agree with Dan.  I think simpler semantics is always better.  
Disconnected doesn't necessarily mean error.

> I was about to ask for a 'auth error' signal, but this 'error' signal is
> more generic and i think it solves my problems.
>   

I don't much like the idea of an "error" signal.  It means that you end 
up with a catch-all handler.

You definitely want a separate authentication-error signal.  For the 
case where the server prematurely closes the connection, and 
unexpected-end-of-file signal or something similar would be useful.  
Note that an unexpected EOF is not the same as disconnected.

> In this case, we should start i18n for gtk-vnc, in order to have the
> error messages translated.
>   

Why would there be string messages at the gtk-vnc level?  gtk-vnc should 
produce machine readable errors and then the user of gtk-vnc should 
translate that into a human-readable error message.

Regards,

Anthony Liguori

> Also, it would be nice having a table of error code available in
> vncdisplay.h, in case the application which receives the signal chooses
> not to display an specific error.
>
> Cheers,
>