[freetds] HP/UX and iconv

ZIGLIO Frediano Frediano.Ziglio at vodafone.com
Fri Dec 12 12:22:59 EST 2003


> 
> On Wed, 10 Dec 2003, "ZIGLIO Frediano" <Frediano.Ziglio at vodafone.com>
> wrote:
> > From man iconv(3)
> > 
> > If iconv() encounters a character in the input buffer that is legal 
> > but for which an identical character does not exist in the target 
> > codeset,
> > iconv() maps this character to a pre-defined character, called the
> > "galley character" that is defined at the time of table 
> generation. (See
> > genxlt(1)).
> > 
> > This mean that under such architecture invalid characters are not 
> > converted to '?' but to system depentend value (in case utf-8 -> 
> > iso8859-1 it's a '.') ...
> 
> OK, we should note that in our documentation.  Something 
> like, "Your iconv may behave somewhat differently, cf. iconv(3)."  
> 
> Does HP's (?) iconv(3) set errno in this case?  
> 

No.. it just returns success without setting errno...

Yesterday I tried --disable-libiconv with some problems (utf8_2
failed)... I updated iconv replacements adding needed UTF-8 support...

Now I got all unittests working under:
- HP/UX 10.20 PARISC (native iconv) mssql2k server
- Linux RH7.2 i386 (native and replacement iconv) mssql2k server, UTF-8
and ISO-8859-1 encoding
- Linux RH9.A i386 (native and replacement iconv) Sybase 11.9.2 server,
UTF-8 encoding
(just to update UG...)

Does Tru64 work OK ? 

It seems that if -D_REENTRANT is not defined some functions (like
strtok_r) are not defined causing some problems with platforms where
sizeof(int) != sizeof(void*) due to implicit conversions. Perhaps this
should be noted in UG... perhaps we should always define -D_REENTRANT
even if threadsafe is disabled ??

freddy77


More information about the FreeTDS mailing list