Re: [xml] returning existing handler in xmlNewCharEncodingHandler

[Igor Zlatkovic <igor zlatkovic com>]

Daniel Veillard wrote:

> On Tue, Feb 10, 2004 at 11:50:02AM +0100, oliverst online de wrote:
>
> > patch against encoding.c version 1.77 as discussed with Igor on the list
> >
> > 1233a1234,1241
> >
> > >     * Check for existing handler
> > >     */
> > >    handler = xmlFindCharEncodingHandler(name);
> > >
> > >    if (handler != NULL)
> > >        return(handler);
> > >
> > >    /*
> >
> > So long
>
>
>   Hum, I might disagree with Igor on this, but it seems to me that
> it might be better to call xmlFindCharEncodingHandler() before
> calling xmlNewCharEncodingHandler(), rather than changing the semantic
> of the existing call. In any case in a multithreaded program it seems
> to me that you need to ensure some critical section when changing the
> default handlers for an encoding. Usually this is best done at program
> startup, but I may have missed something, sorry I couldn't participate
> to last week discussion, but after rereading the archive I'm not sure
> changing the behaviour is the best thing to do.

I agreed to the patch for the following reason.

The patch will change the behaviour of xmlNewCharEncodingHandler(), but 
not that of xmlFindCharEncodingHandler(). The idea was to prevent the 
registration of more than one handler for the same encoding name. That 
because only the handler which was registered first can ever be found by 
xmlFindCharEncodingHandler(). That's the whole reason, it has nothing to 
do with multithreading and related problems.

This of course cannot hold if there is code which relies on what 
xmlNewCharEncodingHandler() returns and does not use 
xmlFindCharEncodingHandler(). But the API suggests not to be used that 
way for the reasons above.

At the very end, I don't care particularily about that patch, it won't 
change anything for me. If it really breaks something somewhere, then 
better not to apply it.

Ciao,
Igor