Re: GSString problem

[Andreas Hoeschler]

Hi,

> > > after upgrading to a recent version of GNUstep my app throws the
> > > following exception.
> > >
> > > NSCharacterConversionException REASON: Can't get cStringLength from
> > > Unicode string
> >
> > I've had other reports of c-string related conversions failing when they
> > didn't before, and I suspect that:
> >
> > 2004-02-17  Bruno Haible  <bruno@clisp.org>
> >
> > * Source/Additions/Unicode.m (GetDefEncoding): Use the result of
> > nl_langinfo(CODESET) before falling back on ISO-8859-1.
> >
> > is causing it by setting the c-string encoding to ASCII when it used to
> > be set to iso-8859-1. If so, setting the environment variable
> > GNUSTEP_STRING_ENCODING to NSISOLatin1StringEncoding should fix it.
> >
> > Do we really ever want to set the c-string encoding to
> > NSASCIIStringEncoding instead of NSISOLatin1StringEncoding? (Especially
> > since parts of Unicode.m is pretending that ascii==iso-8859-1 anyway.)
>
> I guess we do if the user wants to ...
> ie if GNUSTEP_STRING_ENCODING specifies ASCII
>
> However, I think it's probably sensible to use ISO-8859-1 if GNUSTEP_STRING_ENCODING
> is not defined and nl_langinfo(CODESET) says ASCII
>
> I wasn't aware that Unicode.m ever pretended ascii==iso-8859-1 ... I don't think it should.

Using NSISOLatin1StringEncoding instead of NSASCIIStringEncoding in Unicode.m  did fix the problem for me.

      if (defEnc == GSUndefinedEncoding)
{
/* Encoding not set */
#if HAVE_LANGINFO_CODESET
/* Take it from the system locale information.  */
encoding = nl_langinfo(CODESET);
if (strcmp(encoding, "ANSI_X3.4-1968") == 0 /* glibc */
|| strcmp(encoding, "ISO_646.IRV:1983") == 0 /* glibc */
|| strcmp(encoding, "646") == 0 /* Solaris NetBSD */)
//	    defEnc = NSASCIIStringEncoding;
defEnc = NSISOLatin1StringEncoding;	// <------------ this is better
		...
}

Regards,

Andreas