Bug in the integrate.



El Mar 16 Abr 2002 00:32, escribió:
> Just as a general tactic, one could run the
> same problem on commercial macsyma and Maxima
> and trace all functions that seem to matter.
>
> The first time a function returns a different answer
> suggests a place where the bug may exist.
>
...
> > I agree.  Doesn't Wolfgang Jenkner's patch correct the problem?  I
> > haven't tried it out yet.
> >
Well, I have already tried Wolfgang Jenkner's patch and it works. The only 
point to be made is that its results seem to lack a factor(%) on them.

C1) integrate(3^log(x),x);
				 1
			      (------ + 1) LOG(x)
			       LOG(3)
			     3
(D1) 			     --------------------
				1
			     (------ + 1) LOG(3)
			      LOG(3)
(C2) factor(%);
				LOG(x)
				------ + LOG(x)
				LOG(3)
			       3
(D2) 			       ----------------
				  LOG(3) + 1
(C3) risch(3^log(x),x);
				   LOG(3) LOG(x)
			       x %E
(D3) 			       -----------------
				  LOG(3) + 1

were the results I got with Wolfgang's patch and the sin algorithm as the way 
to perform the integration.

He caught the bug: 3^log(x) dx -> 3^y exp(y) dy -> 3^(y + y / log(3)) dy is 
what his patch does. The present code goes from 3^y exp(y) dy to 3^(y + y) dy 
as he explained in his message which I had not read when I started this 
thread.

On the other side, I have run into some problems while compiling today's 
snapshot. All of them are in the init-cl.lisp.
Basically in  maxima-data-path there is a variable *maxima-verpkgdatadir* 
which has not been defined previously and it causes problems when compiling 
with cmu-18d (previous defparameter?). There is also a 
(setq si::*info-paths* ...) declaration which I think should be only gcl 
stuff.
-- 
Juan Pablo Hierro Álvarez
hierro en ideafix punto litec punto csic punto es
Clave pública: 0xA8707ADF en pgp.rediris.es