Re: addfrac() + Karatsuba = bug

[Michael Somos <somos@grail.cba.csuohio.edu>]

Christian Cornelssen mentioned his testing of the bug code. However,
I should mention that I did extensive testing and debugging before
I posted the message. I found that the 'avma = (long)z;' in three
places near the end of 'addfrac()' leads to 'quickmulii()' eventually
clobbering 'delta' in 'addfrac()' near the end. I suggest using the
alternate 'avma = (long)delta;' in these threee cases. I did not have
time to fully test if this is the correct solution. There may be lot
of other issues involved here with interaction of code with respect
to setting of 'avma'. Certainly it would have saved lots of debugging
time if more effective memory allocation monitoring were in place. I
don't see any such capability right now. That seems to be why small
changes in one routine was able to clobber memory used by another.
Shalom, Michael