Re: Possible pari-2.0.5/gp bug

[Karim Belabas <karim@mpim-bonn.mpg.de>]

Roland Dreier wrote (msg 55):
>
> ? factor(3*s^4-140*s^2-4537/9*s-4900/9)
>   ***   the PARI stack overflows !!!

The code in factpol() mistakenly assumed the polynomial to have integer
coefficients AND leading term +/- 1 (that's two bugs in one line, the first
one being fatal).  (this was introduced in a 2.0 pre-release while making
sure that factors always have positive leading coeff.)

Instead of Louis patch (msg 56) which fixes only the fatal one, apply the
patch below.

Karim.

===========================  patch 13 (2.0.5.alpha) ===========================

*** src/basemath/polarit2.c.orig	Wed Feb 18 16:15:52 1998
--- src/basemath/polarit2.c	Fri Feb 13 13:43:07 1998
***************
*** 437,443 ****
    p1 = cgetg(1,t_VEC); fa=cgetg(lx,t_VEC);
    for (i=1; i<lx; i++) fa[i] = (long)p1;
    d=content(x);
!   if (!gcmp1(leading_term(x))) d = negi(d);
    if (!gcmp1(d)) x=gdiv(x,d);
    vv=varn(x); xp=deriv(x,vv); t=ggcd(x,xp);
    v=gdiv(x,t); w=gdiv(xp,t); j=0; f=1; nbfac=0;
--- 437,443 ----
    p1 = cgetg(1,t_VEC); fa=cgetg(lx,t_VEC);
    for (i=1; i<lx; i++) fa[i] = (long)p1;
    d=content(x);
!   if (gsigne(leading_term(x)) < 0) d = gneg(d);
    if (!gcmp1(d)) x=gdiv(x,d);
    vv=varn(x); xp=deriv(x,vv); t=ggcd(x,xp);
    v=gdiv(x,t); w=gdiv(xp,t); j=0; f=1; nbfac=0;
***************
--
Karim Belabas                          e-mail:
Max-Planck-Institut fuer Mathematik       karim@mpim-bonn.mpg.de
Gottfried-Claren-Str. 26               tel:
53225 Bonn (Germany)                      (00 49 228) 402-245