Re: [xml] [patches] static linking on win32 with static cruntimes

[Daniel Veillard <veillard redhat com>]

On Fri, Jan 28, 2005 at 11:16:07PM -0500, Joel Reed wrote:
> snip
>
> > > also, while on topic of APIs, the fact that extension registeration
> > > functions don't take the hash to register the extension in as an 
> > > argument, but instead register to global xmlHashes means folks wanting
> > > to using plugins MUST dynamically link libxslt into the plugin.
> > > would we ever want to fix that? not sure...
> >
> >   I'm not sure I understand :-)
>
> sorry, let me try again. so a plugin init function looks like this:
>
> void EXSLTPUBFUN exslt_org_regular_expressions_init (void)
> {
>     xsltRegisterExtModuleFunction ((const xmlChar *) "match",
>                                    (const xmlChar *)
>                                    EXSLT_REGEXP_NAMESPACE,
>                                    exsltRegexpMatchFunction);
> }
>
> xsltRegisterExtModuleFunction in libxslt/extensions.c adds
> exsltRegexpMatchFunction to xsltFunctionsHash which is a global
> variable.
>
> now imagine xsltproc and exslt_org_regular_expressions.so, each
> statically linking to libxslt. each has a xsltFunctionsHash now.
> if they were both dynamically linked to libxslt, there'd be only
> one xsltFunctionsHash.
>
> so given the current API, exslt_org_regular_expressions.so ends
> up adding a new entry into its xsltFunctionsHash, and not
> the right one -> which is xsltproc's xsltFunctionsHash.

  Okay.

> it seems that one way to fix this would be to pass some context
> ptr to plugin init funcs, that would then just be passed along
> in some new xsltRegisterExtModuleFunction that besides the params
> above, also took a context pointer, and that in the implementation
> of this new func, exsltRegexpMatchFunction would be added not to 
> some global xsltFunctionsHash, but to context->xsltFunctionsHash.
> thereby fixing the issue with statically linked xsltproc's and plugins.

  okay, seems to make sense.

> whether this has other benefits, or is just too much change for
> a small set of users/configs i'm not sure. just describing how
> i think we could make it work.

  agreed there might be more benefits to passing the context down,
for example the access to the dictionnary.
  Send patch I will apply it before the next release (which got delayed,
sorry !)

Daniel

-- 
Daniel Veillard      | Red Hat Desktop team http://redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/