On Sun, Oct 25, 2015 at 11:17:15AM +0100, Georg Baum wrote:
> Scott Kostyshak wrote:
>
> > -% increase link area for cross-references and autoname them,
> > -\AtBeginDocument{\renewcommand{\ref}[1]{\mbox{\autoref{#1}}}}
> > -\addto\extrasfrench{%
> > - \renewcommand*{\equationautorefname}[1]{}%
> > - \renewcommand{\sectionautorefname}{sec.\negthinspace}%
> > - \renewcommand{\subsectionautorefname}{sec.\negthinspace}%
> > - \renewcommand{\subsubsectionautorefname}{sec.\negthinspace}%
> > +\@ifpackageloaded{babel}{
> > + % increase link area for cross-references and autoname them,
> > + \AtBeginDocument{\renewcommand{\ref}[1]{\mbox{\autoref{#1}}}}
> > + \addto\extrasfrench{%
> > + \renewcommand*{\equationautorefname}[1]{}%
> > + \renewcommand{\sectionautorefname}{sec.\negthinspace}%
> > + \renewcommand{\subsectionautorefname}{sec.\negthinspace}%
> > + \renewcommand{\subsubsectionautorefname}{sec.\negthinspace}%
> > + }
> > }
>
> If you switch between babel and polyglossia you will suddenly get different
> references in the PDF, which I would not expect at all as a user.
>
> If some features of a document are implemented in a babel specific way (even
> if the feature would in theory not need babel), then we should rather
> declare that explicitly like in 78c60de9bb IMHO, than to pretend that
> polyglossia can be used, but silently change the output.
Ah this is a good point. Dang it. In this case I think it makes even
more sense now to remove the babel-specific preamble code, which there
seems to be support for from others as well. There are several benefits
to removing the babel-specific preamble code. One of them that I'm
currently focused on is that if we use "Always babel" we will miss out
on a lot of tests. Basically, any document that Uwe has edited will
contain babel-specific preamble code and thus cannot be used as a test
of XeTeX or LuaTeX with non-TeX fonts.
I would prefer to leave the babel conditional in because of my reason
above, until we come to a decision and action on whether the
babel-specific code should be removed (e.g. fixing #9817?). That said, I
do understand that two wrongs doesn't make a right so let me know if you
believe strongly that I should revert.
Scott
