FreeTDS Development Group

[entropy AT freetds.org]

jklowden AT schemamania.org wrote:
> On Sun, Sep 03, 2006 at 04:42:12PM -0400, entropy AT freetds.org wrote:
> > Norbert Sendetzky wrote:
> >
> > > I'm using a standard 32bit x86 machine.
> > >
> > > The problem I'm facing are lots of warnings when compiling the sybase backend 
> > > of my library (OpenDBX, http://www.linuxnetworks.de/opendbx/) against FreeTDS 
> > > instead of the Sybase ctlib.
> > >
> > > In Sybase' ctlib, CS_INT is defined as "long" and as "int" in FreeTDS which is 
> > > problematic when using printf() functions. This is because of the different 
> > > expectations gcc/printf have when handling "%d" and "%ld" modifiers.
> > The problem is that your code is making assumptions about what type a 
> > CS_INT is.  If it were reasonable for your code to do that, then there 
> > would be no such thing as a CS_INT, and instead it would just be "int" 
> > or "long".
> >
> > The most portable thing your code can do in this case is explicitly cast 
> > to a type, then use the printf(3) format for that type.
>
> That's good advice from a C perspective.  The application shouldn't assume 
> anything about opaque types.  
>
> > From the library's perspective, though, maybe we're making a mistake?  We 
> do a bunch of things in the header files -- e.g. use #define instead of enum
> -- just because that's how Sybase did it 20 years ago.  It wouldn't be 
> wrong to change it.  

Not sure what you mean.  In my copy of FreeTDS, CS_INT is a typedef.  I 
don't see how an enum would be helpful.

[550]entropy@zippy:~/freetds/include $ grep CS_INT cstypes.h | head -1
typedef tds_sysdep_int32_type CS_INT;

Even though I don't think it's relevant here, I'm opposed to the idea of 
sweeping changes where we would convert CPP macros to enums.  I don't 
think the religious benefits outweigh the practical lossage, such as 
breaking existing code using #ifdef on those macros.

--
Cheers,
entropy