Closed huonw closed 9 years ago
That is some edge case there. I'm guessing it's something to do with the shift-left implementation.
Seems like ar
and br
are computed correctly, but the number is normalised after doing the elementwise addition (i.e. ll::add
), but before push
ing the carry, so, in this case, the raw addition entirely overflows and leaves every limb 0 meaning normalization trims everything off. #46
This test fails:
(I'm debugging it now.)