FreeTDS Development Group

["Brian Bruns" <camber AT ais.org>]

> > 
> > Ok, let's test this theory.  If you did something like:
> > 
> > DBINT i;
> > 
> > memcpy(&i, dbdata(dbproc, 2), sizeof(i));
> > 
> > and then use 'i' instead of dbdata() directly does the 
> > problem go away?
> > 
> > The information in the row buffer *should* be 4 byte aligned, 
> > I remember
> > at least one alignment patch in the 0.60 series, so it'd be 
> > worth a try. 
> > If not it's definately a bug on our part.
> >  
> 
> Arghh...
> 
> in token.c
> 
>               if (IS_TDS42(tds)) {
>                       remainder = info->row_size % align; 
>                       if (remainder)
>                               info->row_size += (align - remainder);
>               }
> 
> Why IS_TDS42 ??
> 
> freddy77

That is an excellent question.  Unfortunately I don't know.  It's in
tds_col_info which is a TDS 4.2 only function, so the protocol check is
worthless.  The problem lies in tds7_process_result() probably:

                /* actually this 4 should be a machine dependent #define
*/
                remainder = info->row_size % 4;
                if (remainder) info->row_size += (4 - remainder);

If we definately knew TDS 4.2 was working it'd help locating the problem.

Brian