Re: [xml] Re: xmlLoadCatalogs behaviour

[Igor Zlatkovic <igor stud fh-frankfurt de>]

Aaah, right.

I reverted the first change, but the second one slipped through. I'll commit 
this today. Five minus-points for me :-)

Ciao
Igor


Karl Eichwalder wrote:
> Igor Zlatkovic <igor stud fh-frankfurt de> writes:
>
>
> > the KDE people cried about xmlLoadCatalogs not accepting the colon as a
> > directory separator in its parameter. They were right, because the newly
> > introduced behaviour which supports URI notation breaks the old habits and
> > all projects which relied on that.
> >
> > I have reverted the behaviour of xmlLoadCatalogs to the old way.
>
>
> Thanks, but it was not sufficient; Adrian worked out a patch as follows:
>
> -------------------------------------------------------------------
> Mon Nov  4 14:48:08 CET 2002 - adrian suse de
>
> - revert catalog separator change, accepting ":" again.
>   libxml2 is using anyway ":" internal, even when a " " separator
>   is given
>
> -------------------------------------------------------------------
>
> --- libxml2-2.4.25/catalog.c    2002-03-18 20:37:54.000000000 +0100
> +++ libxml2-2.4.26/catalog.c    2002-10-05 04:35:13.000000000 +0200
> @@ -2917,7 +2938,7 @@
>         while (IS_BLANK(*cur)) cur++;
>         if (*cur != 0) {
>             paths = cur;
> -           while ((*cur != 0) && (*cur != ':') && (!IS_BLANK(*cur)))
> +           while ((*cur != 0) && (!IS_BLANK(*cur)))
>                 cur++;
>             path = xmlStrndup((const xmlChar *)paths, cur - paths);
>             if (path != NULL) {
> @@ -2925,8 +2946,6 @@
>                 xmlFree(path);
>             }
>         }
> -       while (*cur == ':')
> -           cur++;
>      }
>  }
>  
>
>
> > This keeps compatibility, but means no URI notation and no usage on
> > Windows (where every path specification contains a colon).
> >
> > I would introduce a new function, xmlLoadCatalogList(const char* paths),
> > which has a full URI support and separates directories by a space
> > exclusively.
>
>
> Providing a new function sounds good to me.