On Fri, Sep 15, 2006 at 03:54:12PM +0200, Georg Baum wrote:
> Enrico Forestieri wrote:
> 
> > On Fri, Sep 15, 2006 at 11:21:39AM +0200, Georg Baum wrote:
> >> Enrico Forestieri wrote:
> >> 
> >> > The attached patch fixes the problem reported by Uwe. I am afraid but
> >> > with mingw/cygwin the setlocale() thing doesn't really work, so I
> >> > always assume iso-8859-1 (latin1).
> >> 
> >> dt2dv/dv2dt should not be aware of any locale. They simply should process
> >> the raw data without interpretation.
> >> IMO the isprint test is wrong and should be dropped completely. At least
> >> for our usage we know that everything we put in the dtl file is OK.
> > 
> > Sorry Georg, I didn't really study the dtl sources. If the author felt
> > necessary testing if a character was a printable one, I have to trust him
> > that this is the correct thing to do.
> 
> No, you don't ned to trust him, since in our case the input data is
> generated by dv2dt. We know for sure that it will not output anything that
> was not in the original DVI file.
> Testing for non-printable chars makes sense if the input comes from an
> untrusted source, but that is not the case here.

Maybe.

> >> Could you prepare & commit a patch that does this, including a comment
> >> why we do so?
> > 
> > Hmm, without a full understanding of what could go wrong if a control
> > character is let in, I will not do that. Notice that the occurrence of
> > a non-ascii character in the body of a document has never been a problem.
> > The problem reported by Uwe was due to the use of non-ascii characters as
> > arguments of the \pdfauthor and \pdfsubject macros.
> 
> The full understanding statement applies to your fix as well IMO. How do you
> know that dt2dv can cope with non-ascii chars at all?

Because isprint would not fail in a locale aware environment when the
locale is iso-8859-1 and its argument is 246 (รถ).
However, I found that not all non-ascii chars cause problems and I can't
understand why.

-- 
Enrico

Reply via email to