Re: [xml] More windows exports

[Igor Zlatkovic <igor zlatkovic com>]

On Mon, Apr 28, 2003 at 05:07:39AM -0400, Daniel Veillard wrote:
>   I just noticed one thing:
>
> #ifdef LIBXML_SCHEMAS_ENABLED
> xmlUCSIsAlphabeticPresentationForms
> #endif
>
>   all these symbols from the unicode map are actually dependent on
> LIBXML_REGEXP_ENABLED .

Eh? So the .def.src file was wrong all the time :-) Okay, I'll fix that
in an instant. I forgot to add a ChangeLog entry about the new file
anyway.

>   Okay, can I add the xsltproc command to regenerate it when the API
> is rebuilt ?

Sure. Regenerate it as often as you please :-)

>   Once done and checked doing the same for libxslt/libexslt is the
> next step, but should be quite easier since there is no conditional
> compilation.

Yes, I'll make the stylesheet for lib(e)xslt as well.

>   Found and fixed, it was hardcoded from the API generator, and left
> from the time where the schemas stuff was not part of the default build.

Okay, then I'll remove the hardcoded part in the stylesheet.

Another thing, not necessarily related to the topic, but it has shaken
my interest awake: 

The stylesheet basically walks through all functions and
variables, producing an entry for each in the .def.src file. For every
symbol, it 

1. runs few xsl:if tests to determine if there should be an #ifdef
2. it outputs the symbol name
3. outputs the #endif which corresponds to the #ifdef from step 1.

The step 1 can output more than one #ifdef, if a symbol depends on more
macros, or no #ifdef at all. The step 3 must accordingly output the same
number of #endifs.

What I don't like is the fact that the step 3 must run the same tests as
the step 1. In our example, the stylesheet is small and it does not
matter, but somehow it would be cool if there were a way for the step 1
to tell the step 3 simply how many #endifs it must output, to avoid
duplicate tests.

Do you have an idea? Or perhaps a hint for a better organisation of the
stylesheet, which would avoid duplicate tests?

Ciao,
Igor