Problem: difference between 32 and 64-bit TFM

[Ron Aaron]

Attached is a small sample program which I've compiled against a 32-bit and a 64-bit version of TFM.  Both versions of TFM were compiled with FP_MAX_SIZE of 8192 instead of 4096.

The 64-bit version works as expected, the 32-bit gives an incorrect value (for 989 factorial); I would expect to get the same results from either.

Is there something I'm doing wrong in the TFM configuration?  I see this on all platforms (Windows, Linux, OS X)

N.B.: is there a way to detect overflow in TFM?

Thanks!

[Ron Aaron]

Nobody has ideas?