FreeTDS Development Group

["ZIGLIO, Frediano, VF-IT" <Frediano.Ziglio AT vodafone.com>]

> 
> Federico Alves wrote:
> > I open a ticket with DBD-Sybase and the author Michel Peppler thinks
> > there is an issue
> > with freetds, not with DBD-Sybase.
> > <URL: http://rt.cpan.org/Ticket/Display.html?id=34411 >
> > 
> > On Mon Mar 24 20:19:06 2008, falves1 wrote:
> > > I compiled DBD-Sybase 1.08 using DBI 1.604, and the perl 
> Makefile.PL 
> > > works fine, but then when I do a make I get this
> > > Sybase.xsi:275: error: invalid lvalue in assignment
> > > Sybase.xsi: In function âXS_DBD__Sybase__db_DESTROYâ:
> > > Sybase.xsi:334: error: invalid lvalue in assignment
> > > make: *** [Sybase.o] Error 1
> > 
> > This is a problem with FreeTDS.
> > 
> > The issue is that FreeTDS defines CS_UINT_TYPE, but not 
> CS_BIGINT_TYPE,
> > and Sybase never defines one without the other.
> 
> I'm not convinced this is a "problem" with FreeTDS, and I'm 
> not sure how
> to fix it.  
> 
> FreeTDS doesn't guarantee line-by-line compatibility with 
> Sybase's header
> files.  The project's header files describe the functions, 
> structures, and
> datatypes implemented by the library.  If CS_UINT_TYPE but not
> CS_BIGINT_TYPE is supported/used/implemented by the library, then only
> CS_UINT_TYPE *should* appear in the header file.  Anything else would
> mislead the developer: the translation unit might compile but fail to
> link, or link and not work.  
> 
> Michael, AIUI DBD::Sybase assumes CS_BIGINT_TYPE upon testing for
> CS_UINT_TYPE.  Wouldn't it be better to test for 
> CS_BIGINT_TYPE directly? 
> 
> 
> I'm always eager to make sure FreeTDS supports DBD::Sybase to 
> the extent
> possible, allowing for deficiencies in the implementation.  But adding
> something to the header file not implemented by the library could
> adversely affect other users.  
> 

I checked the code and it seems that CS_BIGINT_TYPE, CS_DATE_TYPE, 
CS_TIME_TYPE and CS_UINT_TYPE are not so supported. I also don't know how 
easy is implementing unsigned types and date/time. What happen for instance 
if we send unsigned data to ms sql? Perhaps it would be better to undefine 
these constants.

freddy77