From: Kbwms <Kbwms AT aol DOT com>
Message-ID: <2cd092ba.353f5d4a@aol.com>
Date: Thu, 23 Apr 1998 11:24:57 EDT
To: djgpp AT delorie DOT com
Mime-Version: 1.0
Subject: Problems with paranoia.c and pow(0,0)
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7bit
Precedence: bulk

I ran paranoia.c today and obtained the following DEFECT printout:

Diagnosis resumes after milestone Number 90          Page: 7

Testing powers Z^i for small Integers Z and i.

* * * FLOATING-POINT ERROR 289 * * *
SR3980 failed for Z = 0.000000, Q = 0.000000!
DEFECT:  computing
	(0.00000000000000000e+00) ^ (1.02300000000000000e+03)
	yielded 0.00000000000000000e+00;
	which compared unequal to correct 1.00000000000000000e+00 ;
		they differ by -1.00000000000000000e+00 .
Similar discrepancies have occurred 1 times.

To continue, press RETURN

The program is actually trying to compute pow(0,0).  Where the number 1023
comes from is a mystery.

The following test program was written:

#include <float.h>
#include <math.h>
#include <stdio.h>
int main()
{
	double	x;
	_fpreset();
	x = pow(0.0,0.0);
	_fpreset();
	printf("pow(0,0) = %f\n", x);
	return 0;
}
This produced the following:

pow(0,0) = NaN

Note that the call to pow() had to be surrounded by calls to _fpreset()
to prevent an abort printout like this:

Exiting due to signal SIGFPE
Floating Point exception at eip=0000157b
eax=00007800 ebx=000c32c0 ecx=00000000 edx=0000033e esi=00000054 edi=0000c4d0
ebp=0007ffd4 esp=0007ffcc
program=C:\DJGPP.V2\TESTS\LIBC\ANSI\MATH\PARANOIA\TSTPOW.EXE
cs: sel=00af  base=807c5000  limit=000cffff
ds: sel=00b7  base=807c5000  limit=000cffff
es: sel=00b7  base=807c5000  limit=000cffff
fs: sel=0087  base=0001af10  limit=0000ffff
gs: sel=00c7  base=00000000  limit=ffffffff
ss: sel=00b7  base=807c5000  limit=000cffff

Call frame traceback EIPs:
  0x0000157b
  0x00001af2


Calculating 0^0 should produce 1.


K.B. Williams