FreeTDS Development Group

[James K. Lowden <jklowden AT speakeasy.org>]

On Sun, 4 Aug 2002 21:49:08 +0800, "Varley, David(CBorn at Alcoa)"
<David.Varley AT alcoa.com.au> wrote:

> Regarding the atoll() problem, a long is 64 bits on the alpha anyway,
> and atol() returns a 64 bit int, so atoll() is not required or provided.
> A simple #define atoll atol  would fix this, but I'm no autoconf guru,
> and I assume that's the place for a proper fix.

David, 

>> As a long-time lurker on this list, I have to break my silence on this,
>> although I hate to do so on a negative note after all the great work
>> Bill's been doing. 

Glad to make your acquaintance.  

I just committed a change to configure.in that checks for 64-bit ints and
uses atol() if atoll() is not provided.  Please let me know if it works
for you.  The configure script should fail to define HAVE_ATOLL on your
system, causing FreeTDS to rely on atol().  

Because the NetBSD system I happen to test with is a few years old, it
lacks both 64-bit ints and atoll().  I'm experimenting with it to see if I
can use it anyway, subject to the limitations imposed by the missing
features.  Partly for that reason, the configure script will *not* fail if
64-bit integers are unsupported; it is only a warning.  I want to see if
FreeTDS can be made to live within such limitations.  Ideally, it would
use 32-bit ints if 64 bit aren't available, and call the error handler in
situations where the width was insufficient.  

At the moment, however, char<->money is quite broken for me, see next
message.  

Regards, 

--jkl