FreeTDS Development Group

[Brian Bruns <camber AT ais.org>]

Hmmm...typdef'ing SQLCHAR to SYBCHAR probably could be done 
unconditionally, however attempting to use ODBC and dblib in the same 
.c file would suffer some incompatibility.  Incidentally, I run into 
this alot with DB2 (SQL/CLI based API) and ctlib/dblib so this is perhaps 
nothing new.  The stickler is our usage in dblib_handle_err_message which 
has no obvious work around, but could be promoted to a runtime check perhaps? 
As it stands now we can not have a single freetds installation for sybase 
derived dblib code and ms-dblib code. It would be nice to get at least 
that far. 

On Sun, 9 Feb 2003, Steve Langasek wrote:

> On Sun, Feb 09, 2003 at 11:00:03PM -0500, Brian Bruns wrote:
> 
> > I had been thinking about this since I put in the --enable-msdblib option 
> > in.  We could make the dbdaterec structure hold the values for both 
> > sybase 
> > and msdblib names and then dbdatecrack can fill in both values, great for 
> > source compatibility with both and if we put the sybase stuff first in 
> > the 
> > structure, binary compat should be no problem (assuming we are shooting 
> > for sybase binary compatibility, which is really the only option on *nix 
> > platforms).  Anyway, point is we could remove the --enable-msdblib option 
> > entirely.
> 
> That sounds like a good idea to me; then PHP can just check for
> whichever structure member it prefers to use at compile time, and build
> accordingly.
> 
> I'm afraid we won't be able to get rid of the MSDBLIB option entirely,
> as it appears people have found this useful in other parts of the code
> as well.  As for binary compatibility, it would be good if this change
> made the cut for 0.61: changing the structure size still implies some
> incompatibility, so it would be nice to get this change in for a release
> where we're already bumping the soname.
> 
> BTW, would it be worthwhile to use a set of defines or anonymous unions
> within the struct?  It would save space and simplify the code a little
> bit, I think; there's really no reason to keep two sets of values for
> those elements that have the same semantics between Sybase and MS.
> 
>