Fix for sign extension buglet

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

And another small patch for 2.0.11.beta.  Guillaume Hanrot (thanks!)
noted that under some conditions, notably on DEC alphas, one line
in the cgcd() function in arith1.c would produce a (signed) C int, of
half the size of a signed long, with the most significant bit turned on,
which would then get converted to a (negative) C long by sign extension,
although a positive value was intended.  The fix is to tell the compiler
that the thing is a C long right from the start.  Two more potential
problems of this kind are secured below;  if anybody can find any other
similar places in the code, please do let us know.

Enjoy, Gerhard

Apply with patch -p0 -l from the pari-2.0.11.beta directory  (-l because
TABs/spaces otherwise might not match).

bash$ diff -u src/basemath/arith1.c.orig src/basemath/arith1.c
--- src/basemath/arith1.c.orig  Tue Aug  4 20:02:23 1998
+++ src/basemath/arith1.c       Tue Aug  4 20:02:25 1998
@@ -921,7 +921,7 @@
   if (a>b) { a %= b; if (!a) return b; }
   else { b %= a; if (!b) return a; }
   v=vals(a|b); a>>=v; b>>=v;
-  if (a==1 || b==1) return 1<<v;
+  if (a==1 || b==1) return 1L<<v;
   if (b&1)
     return ((long)ugcd((ulong)a, (ulong)b)) << v;
   else

bash$ diff -u src/basemath/arith2.c.orig src/basemath/arith2.c
--- src/basemath/arith2.c.orig  Tue Aug  4 21:30:38 1998
+++ src/basemath/arith2.c       Tue Aug  4 21:30:42 1998
@@ -316,7 +316,7 @@
     return 16384;
   else if (size <= 512)
     return (size-16) << 10;
-  return 1<<19;		/* Rho will generally be faster above this */
+  return 1L<<19;		/* Rho will generally be faster above this */
 }
 
 /********** about to become obsolete --GN **********/

bash$ diff -u src/basemath/ifactor1.c.orig src/basemath/ifactor1.c
--- src/basemath/ifactor1.c.orig        Tue Aug  4 21:31:57 1998
+++ src/basemath/ifactor1.c     Tue Aug  4 21:31:59 1998
@@ -620,7 +620,7 @@
   else if (size < 65500)
     c0 = size<<6;		/* back to linear; ECM taking over */
   else
-    c0 = 1<<22;
+    c0 = 1L<<22;
 				/* improbable, but size _could_ be
 				   very large indeed... */
   c = c0 << 4;			/* 16 iterations per round */