My retry-integrate replacement for meval (list '($ev) result '$nouns))) does seem to fix the multiple evaluation integration bug.
But maybe it's not the right thing to do.
The risch code does other sketchy things. The function trigin1 indiscriminately calls realpart or rectform on the output of
hypertrigint1. Along with option variables generate-atan2 and triginverses, the realpart or rectform call seems to
be needed to signal the definite integration code that the antiderivative might not be continuous everywhere the integrand is
continuous. Ugh.
Opps (see http://sourceforge.net/p/maxima/bugs/2590/)
(%i66) integrate(sqrt(1-cos(x)),x);
1" Enter "?hypertrigint1[sqrt(1-cos(x)),x,false]
1" Exit "?hypertrigint1(-sqrt(2)*%i*%e^(%i*x)-sqrt(2)*%i)/sqrt(-%e^(%i*x))
(%o66) (-sqrt(2)*cos(x)-sqrt(2))*sin((atan2(sin(x),cos(x))+%pi)/2)+sqrt(2)*sin(x)*cos((atan2(sin(x),cos(x))+%pi)/2)
(%i67) diff(%,x)- sqrt(1-cos(x))$
(%i68) wxplot2d(%,[x,0,14]);
I belive that %o66 is wrong on (%pi,2%pi) union (3 %pi, 4 %pi), ...
(defun retry-integrate (e)
(let (($opsubst t))
($substitute #'(lambda (&rest l) (apply #'$integrate l)) '%integrate e)))
--Barton