>Hi,
>
>I guess I'll let the problem rest, it is not of any real consequnece to me. If it were, I guess I would write my own BCDREALS subroutine. Funny, in the days of CP/M and MP/M, we used the Pascal/MT+ programming language which had a BCDREALS library routine - we even wrote our own version, to increase speed (8 bit 4MHz CPU) and save memory (64KB max).
>
>Rolf

Hi Rolf,

I didn't do any in-depth study other than vir** research that day (well.. I still a young guy when it was 80's. No I never write any vir** <*bg*>). But as far as I can recall, wasn't it the machine that doesn't have co-processor yet ?

Anyway, I wasn't trying to be mean or something like that, NO. I was just trying to give you a clarification because you mentioned "how can you trust VFP?".
About your problem, I think you can get away from it if you round the value to 13 decimals (52 bits mantisa).

Regards