Re: FreeTDS dbwillconvert discrepancy

Joe Losco wrote:
> 
> My problem that I've narrowed it down to is with this stub of code
> if (SYBCHAR != pcol->type) {           
>                 pcol->size = dbwillconvert(pcol->type, SYBCHAR);
>             }
> 
> I've placed  "printf("name:%s type:%d size:%d\n", pcol->name, 
> pcol->type, pcol->size);" around this piece of code so that I can see 
> the before and after effects.  In the command line pure C application 
> dbwillconvert returns the size of the would be data converted to the 
> correct type.  This then lets the memory be allocated properly in the 
> later calloc that uses pcol->size+1, as well as everything else to work 
> as intended. 

Yes, that's my code.  

db-lib as specified and implemented by the vendors offers no way to
discover the printed length of a non-character value.  FreeTDS extends
dbwillconvert() to do that: instead of a simple boolean, it returns the
number of bytes needed for a character representation.  I'm sorry it's not
better documented.  :-(

> However, in the Objective C/GUI version it appears to return a Bool.  It
> 
> seems to return 1 if it is convertable and 0 if it is not.  This 
> obviously messes up the rest of the allocations in the sample resulting 
> in truncation and overflow errors.  

My first instinct was to wonder if in fact you're linking to FreeTDS and
not one of the vendor libraries, or perhaps to a version prior to 0.82.  I
assume none of those is the case.  

My second guess is to wonder if Objective C is looking at the header file
(or something like that), interpreting each nonzero return code as "true",
and returning only 0 or 1.  Is there glue code that allows Objective C to
call a C library?  Is there some preprocessing that gets done to make C
functions available to it?  

If so, the solution might be to define the function as returning DBINT
instead of DBBOOL.  Or something like that.  

> I then looked into this farther and 
> found http://www.freetds.org/reference/a00275.html#ga24 which from what 
> I can tell is that the desired effect of that function is to return the 
> bool. 

Ouch.  Yes.  I'll update that.  Thanks for pointing it out.  

--jkl