[no subject]

[RWG@MIT-AI]

a bug in expt (actually si:slow-binary-expt) which might
actually exist in ucode too:  straightforward ways to write the loop which
shifts the exp while squaring the base wind up squaring once more than
absolutely necessary, thus risking overflow even when exp is 1!
i munged lispm2;numer > accordingly.  this broke it completely,
due to being compiled in a decimal environment, so i stuck a base:8
in the top -*-line-*-.

also, is there some reason why ODDP is in qrand instead of numer?
also, why isn't it a macro, like bit-test?