> > The correct unicode codepoint would be U+00FC, not U+FC000000. 'accum'
> > already contained the correct value _before_ byteswapping.
> 
> Code review and testing confirmed that this analysis is completely
> correct.  So i have committed this part of the patch.
Thanks!

> > 2. The code in preconv_encode() assumes 'accum' to be a 32 bit integer,
> > but the type declared is 'unsigned int'. While int is 32 bit on all
> > platforms that OpenBSD supports, it is not guaranteed by the standard.
> 
> It is true that neither the ANSI C89 nor the ISO C99 standard
> requires more than 16 bit, but Henri Kemppainen correctly pointed
> out to me in private mail that POSIX requires at least 32 bit.
> Thanks for that cluestick.
Oh interesting, good to know.

cheers,
natano

Reply via email to