Re: pari-2.0.10.beta

[Gerhard Niklasch <nikl@mathematik.tu-muenchen.de>]

In response to:
> Message-Id: <19980713181407.H28256@io.txc.com>
> Date: Mon, 13 Jul 1998 18:14:07 -0400
> From: Igor Schein <igor@txc.com>
> To: "cohen@math.u-bordeaux.fr" <cohen@math.u-bordeaux.fr>,
>         "pari-dev@list.cr.yp.to" <pari-dev@list.cr.yp.to>
> 
> On Mon, Jul 13, 1998 at 08:25:20AM -0400, cohen@math.u-bordeaux.fr wrote:
> > 
> > As a temporary (?) fix to the bug described by M. Atria with 0-parameter
> > functions,
> > line 1349 of anal.c replace cgeti(nparam) by cgeti(max(1,nparam)).
> > 
> > H. Cohen
> 
> This still doesn't fix Gerhard's variation on the same theme:
> 
> mysqrt()=sqrtint(1)
> 1.-mysqrt()
> 
> It doubles the stack a few times, in my case up to 64M
> and then SEGFAULTs.

The real problem is that cgeti(x) is getting called with x==0, which makes
a royal mess of the PARI stack.

Karim is working on it but the mail system at U Paris Sud is STILL
down after three days, so he may have severe difficulties communicating
his findings.

Workaround:
If you omit the empty parentheses when calling an argumentless gp
user function, this particular bug doesn't seem to be exercised.
(I.e., invoke it as
gp > 1.-mysqrt
.)

Gerhard