I just tried to calculate on my CASIO fx-991 ES the following equation:
$7F FF FF FE_{16} \cdot 2$ and
$7F FF FF FE_{16} + 7F FF FF FE_{16}$
My result was MATH ERROR.
I therefore expected I overshot the 32 bit limit of the base mode of my calculator. I concluded that if that happened in a processor, the "carry flag" would be set and continued to compare my result with given solution. I was wrong. After trying to figure out why and calculating this by hand, I realised the solution is $FF FF FF FC_{16}$. When I type this directly into my calculator it accepts the input without hesitation.
Does anyone have any idea what kind of math error I ran into if the result is clearly not too big for calculation? (By the way, if I take one digit away, it's fine with the calculation. But still, why does this happen before violating the max length restriction?)
Oh, and a final thought. This is apparently calculator specific. The windows calculator has no issue with the above input.
Best Answer
I don't know the inner workings of this particular calculator. But my wild guess theory is that your inputs were interpreted as a signed integers. Often the range of a signed integer is from $−80000000_{16}$ to $7FFFFFFF_{16}$, and this overflowed. With the latter input the calculator was forced to take it as unsigned.
You know your tool, and can cook up further tests that give more information :-)