(0) Previously, the imaginary axis was a branch cut of acosh.
This is no longer the case:
(1) The 24 trigonometric and inverse trigonometric functions now
directly use the Common Lisp functions for double floating
point evaluation. It's no longer neccessary to use rectform
for complex double float evaluation.
(%i4) cos(1+%i),numer;
(%o4) 0.83373002513114902-0.98889770576286506*%i
(%i5) acos(%);
(%o5) 0.99999999999999978*%i+1.0
(2) The algorithm that decides when to apply a reflection identity
no longer uses 'sign'; instead it uses 'great'.
(%i6) assume(a < 0)$
(%i7) [sin(a),cos(a)];
(%o7) [sin(a),cos(a)]
A 'sign' bug causes bug 1377516. The proposed 'great' based
algorithm eliminates this bug (but doesn't fix the sign bug)
(%i8) trigreduce(sin(x)*sin(sqrt(3)*x/2));
(%o8) cos((sqrt(3)*x)/2-x)/2-cos((sqrt(3)*x)/2+x)/2
(3) Taylor polynomials are more contagious:
(%i9) taylor(x,x,0,5);
(%o9) x+...
(%i10) sinh(%);
(%o10) x+x^3/6+x^5/120+...
(%i11) asinh(%);
(%o11) x+...
Barton