odd bug in integrate: using x(t) instead of x get answer

>>>>> "Barton" == Barton Willis <willisb at unk.edu> writes:

    Barton> Substituting x --> z[0] also shows this bug. Also
    Barton> (%i1) s:  1/sqrt(log(x)+a*x+b)$

    Barton> (%i2) integrate(subst(x=z[0],s),z[0]);
    Barton> (%o2) -sqrt(%pi)*%i*%e^(-b-z[0]*a)*erf(%i*sqrt(b+z[0]*a+log(z[0])))

    Barton> (%i3) integrate(subst(x=a+b,s),a+b);
    Barton> (%o3) a/sqrt(log(b+a)+(a+1)*b+a^2)

    Barton> (%i4) integrate(subst(x=%i,s),%i);
    Barton> Maxima encountered a Lisp error:

    Barton> (%i5) integrate(subst(x=sqrt(5),s),sqrt(5));
    Barton> (%o5) (2*sqrt(b+sqrt(5)*a+log(5)/2))/a

This looks wrong.  I remember sometime ago that integrate was changed
to allow the integration variable to be more than just a plain
symbol.  Perhaps to support using things like z[0] as the dummy
variable?  But this seems to be too loose.

Ray