FreeTDS Development Group

["James K. Lowden" <jklowden AT freetds.org>]

Frediano Ziglio wrote:
> > I remember from reading the ANSI C rationale that extern was the
> > source of some difficulty because there was more than one reasonable,
> > well established practice.  Some linkers want:
> > 
> >     file1.c:        char value[] = "hi";
> >     file2.c:        extern char value[];
> > 
> > others want:
> > 
> >     file1.c:        extern char value[] = "hi";
> >     file2.c:        extern char value[];
> > 
> > I chose #1.  You might try #2 (add extern to the definition).  
> > 
> > Funny we haven't bumped into this before.  
> 
> I think is related to symbol reduction used in Makefiles
> (-export-symbols-regex). 

Hmm.  Why does it work for me?  Are we using different versions of
libtool?  Does the nightly snapshot link for you?  

> I think however that it would be better to
> avoid ctlib.h inclusion in common.c and use cs_locale/ct_dt_info.

Sure, if we ever get there.  Someone who cares more than I do about ct-lib
will have to do that.  In any case, you always need a compiled-in default.
 All I did was centralize it.  

I thought about adding -e to tds_strftime() and decided against. Sybase
and Microsoft don't even format their dates the same way.  It's OK if you
get "Jan 03, 2008" on Unix and "Jan  3, 2008" on Win32.  If someone really
doesn't like that, they know where the patch goes.  

I didn't put STD_DATETIME_FMT in any .h file.  The .c files that need it
declare it themselves.  

How do you feel about a release?  

Regards, 

--jkl