[ REQ 5397]: hardy mystery

["Brad Barber" <bradb>]

Hi,

I'm releasing a new version of Qhull and ran into some very
strange behavior from hardy.  This happened whether I was
compiling with gcc or with cc and fat binaries.


This code==>

void qh_printpoint(FILE *fp, char *string, double *point) {
  int k,id;
  double *p;
  
  if (!point)
    return;
  fprintf (stdout, "point %x %2.2g %2.2g %2.2g\n", (int) point, point[0], point[1], point[2]);
  p= point;
  fprintf (stdout, "point %2.2g %x %2.2g %x %2.2g %x %2.2g %x\n", *p, (int)p, 
       *p++, (int)p, *p++, (int)p, *p++, (int) p);

  for(k= 3; k--; ) {
      fprintf(stdout, " %2.2g", *point++);
  }
  fprintf(fp, "\n");
} /* printpoint */

==== produces this result  ==>

point 75470  0  0 -0.5
point  0 75470  0 75478  0 75480 -0.5 75488
  0 -0.5  0

==============

the first two lines are right, the third is offset by one.  This
happens for every triple of doubles.  If I change the loop to
read:

  for(k= 3; k--; ) {
	p= point;
      fprintf(stdout, " %x %2.2g", (int) p, *point++);
  }

it works fine.  The code works fine on godel and gauss.  I'm stuck.
Any ideas?

					--Brad

P.S.: This is from ~bradb/qhull/Install/qhull_hardy/io.c
I was compiling with gcc .