Limit incorrectly believes that log is continuous along = the negative
real axis---this gives rise to the bug. Consider(%i7) limit(log(-1 + %i * x),x,0,minus);
(%o7) LO= G(-1)(%i8) limit(log(-1 + %i * x),x,0,plus);
(%o8) = LOG(-1)(%i9) trace(?logarc)$
(%i10) limit(atan(x)/(exp(1)= -exp(-x)),x,inf,plus);
Enter ?LOGARC [ATA= N,x]
Exit ?LOGARC %I*LOG((1-%I*x)/(= %I*x+1))/2
(%o10) %E^-1*%I*LOG(-1)/2This isn't logarc's f= ault. I don't have a fix.Barton=