Hello list,
the last time I mentioned this bug (or misbehavior maybe?) I didn't get any feedback. Anyway, I hope I get some feedback now on this patch, which unfortunately makes complex number division more complex (;-) but fixes mistaken results in case of denominator overflow. Any ideas for optimizing this algorithm are welcome.
You may reproduce with:
writeLn(cstr((-8E+231 - 1E+230*i) / (-1E+144 + 4E+199*i)));without the patch it returns:
-0.000000000000000E+000
but with the patch:
-2.500000000000000E+030+2.000000000000000E+032i
Dimitris
