On 14 Jul 2004 Norman Gray wrote:
> Fixed. I think. At any rate, this now produces plausible results on
> ix86 and PowerPC, and the new test I've added runs successfully on
> both. It's not a cast-iron test, but it would easily catch significant
> errors, I think.
It looks like it's fixed for reals, but I'll present an exercise for the
interested reader. Spot the flaw in PRM_PAR:
BYTE VAL__MAXUB
PARAMETER ( VAL__MAXUB = '00'X )
* [In decimal: 254]
Unless we've somehow entered a bizarro universe where 0 in hex is 254 in
decimal, something tells me this is wrong. They're coming out incorrect
for maximum non-bad value and maximum value for integers. The reals all
look fine, as do the minimum values for integers.
Cheers,
Brad.
|