precision in zetak

[Pierre.Charollois@math.u-bordeaux1.fr]

Dear Pari users,

I found, like other people before, the following trouble with the zetak
command :

it works properly with low precision, replying the right rational special
value :
1/30 in my case.
But when I asked for more digits, the answer turns out to be
completely wrong.

The good thing is that low precision is enough to recognize the rational
appearing.
But it took me a while to fix  this problem
 in my whole long and high precision  routine using this command.

As I understood from the mailing list (dev), there is no medication for this
bug yet.
Hence this e-mail is more like a Warning than a Question.


Cheers,

Pierre Charollois

Here is the session :



                  GP/PARI CALCULATOR Version 2.2.11 (released)
                i686 running cygwin (ix86 kernel) 32-bit version
               compiled: Oct 20 2005, gcc-3.4.1 (cygming special)
                (readline v5.0 enabled, extended help available)

parisize = 4000000, primelimit = 500000
(10:09) gp > S=zetakinit(x*x-5);
(10:09) gp > T=zetak(S,-1)
%2 = 0.03333333333333333333333333333
(10:09) gp > \p 240
   realprecision = 240 significant digits
(10:09) gp > allocatemem
  *** allocatemem: Warning: doubling stack size; new stack = 8000000 (7.629
Mbytes).
(10:10) gp > S=zetakinit(x*x-5);
(10:10) gp > T=zetak(S,-1)
%4 = 722489817858621912736514810078790.3782681214653231834976085766586244077995
3962166197128137792893080920179588508779993018946793133971088708805520549113318
2944763943049172772593092241460415893167896301794187079177346794106871173943395
95468210