Re: [xml] RelaxNG validation error reporting

[Petr Pajas <pajas ufal ms mff cuni cz>]

Daniel Veillard <veillard redhat com> writes:

> > 1) I was quite surprised, that setting xmlRelaxNGSetValidErrors error
> > handlers to fprintf versus setting it to my custom routine (which in
> > turn only wraps vsprintf or similar), produce different error
> > messages: with fprintf I get full information, including filename and
> > line numbers, while with any other custom routine I only get the
> > 'body' of the error message (no filename, line no.). My question is,
> > is it intentional (seems so according the code in error.c), and is
>
>   yes intentional, so there is only one callback per error which was
> a highly requested change. This also mean I can't make a callback 
> for the context, and then a callback for the error informations.
> The application can then grab the informations from the context or
> the structured error handler.

ok I see. 
> ...
> > (I'm doing some work on XML-LibXML perl binding and at this point, I
> > just want to fix things, not rework, preserving compatibility with
> > 2.5.x versions as long as possible; some rearranging may come later).
>
>   understood, sorry about that but that part seriously changed, and there is
> no easy workaround.

Ok, I guess I'll try to switch to structured errors with libxml2.6.1.
It has some indisputable benefits too and if it has to be done, it'll
be done :-)

> > 2) sometimes, libxml2 only reports, that the document didn't validate
> > without giving any specific reason (just returns 1 from
> > xmlRelaxNGSetValidErrors but doesn't raise any error via the error
> > handlers). In these cases, xmllint only says 'fails to validate'.  I
> > have no 'real-life' examples for this, but try for example
> >
> > xmllint --noout --relaxng tutorA.rng demo.xml
> >
> > (files from libxml2's test/relaxng/). I would expect something like
> > "expecting element got TEI.2". Is it a bug? If not, are there some
> > typical situations, for this to happen?
>
>   Then it's a bug, please bugzilla it !

Done as http://bugzilla.gnome.org/show_bug.cgi?id=126093

Thanks,

-- Petr