Today, Dan Gildea has checked in a correction for the function absarg in
rpart.lisp to handle correctly the absolute value of a power with
complex numbers. This correction is much more useful than my first
attempt to change cabs to $cabs in simpabs, because now the algorithm is
correct.
Thus, I am wondering if the usage of $rectform in the function $cabs is
only a workaround to get correct results from cabs (before the
revision).
I have tried to remove the $rectform from $cabs and I have got no
problems with the testsuite. The only change I have found is that for a
complex symbol we would get cabs(z) --> cabs(z) and not
sqrt(realpart(z)^2+imagpart(z)^2).
A lot of results would get more simple without a $rectform to $cabs.
Perhaps the call to $rectform can be removed from $cabs.
Dieter Kaiser