Re: our libc included iconv library

[Takahiro Kambe <taca%back-street.net@localhost>]

In message 
<20090304.044208.48403892.tshiozak%anzu.bogus-hosts.imou.to@localhost>
        on Wed, 04 Mar 2009 04:42:08 +0900 (JST),
        T.SHIOZAKI <tshiozak%netbsd.org@localhost> wrote:
> > > As I said on source-changes -- I strongly object a blind removal of all
> > > // options.
> > I also don't agree with removing // options to solve iconv()'s problem.
> > It dosen't things better, but cause another confusion.
> > 
> > What do you think about it, tshiozak?
> 
> Anyway, we cannot resolve this problem unless our iconv emulates
> evilness of GNU libiconv completely.  The judgement whether to take
> such evilness into our base system is difficult.
So I think that src/lib/libc/iconv/iconv.c revision 1.9 should be reverted.

As Joerg said another thread, 1.8 => 1.9 change the behavior from:

        o Simply error due to unsupported encoding.

to:

        o Supported encoding but the result is much differ from what
          the application expected.

As a result, the application which badly use "//" might fail
mysteriously.

Best regards.


-- 
Takahiro Kambe <taca%NetBSD.org@localhost>/<taca%back-street.net@localhost>