not_a_number for MinGW build (gcc-4.5.0)

[Tatsuro M. <tma...@ya...>]

Hello

For current windows binaries,
gnuplot> pr NaN
0.0

This is not correct.

For the cygwin and the djgpp (modified recently) give 
gnuplot> pr NaN
NaN

I have googled NaN for the MinGW, I have not obtained good information yet.
So I have written the test code the below

**************************************
#include <stdio.h>
#include <math.h>


int main(void){
    double dbl;
	unsigned long uldbl[2];

    printf("test 0.0/0.0\n");
    dbl = 0.0/0.0;
    *( double* )uldbl = dbl;
    printf("%f \n",dbl);
    printf("isnan = %d \n",isnan(dbl));
    printf("%x %x\n", uldbl[0], uldbl[1]);
    printf("\n");

    printf("test 0xffffffff 0x7fffffff \n");
    uldbl[0]=0xffffffff; uldbl[1]= 0x7fffffff;
    dbl = *( double* )uldbl;
    printf("%f \n",dbl);
    printf("isnan = %d \n",isnan(dbl));
    printf("%x %x\n", uldbl[0], uldbl[1]);
    printf("\n");

    printf("test atof(\"NaN\") \n");
    dbl = atof("NaN");
    *( double* )uldbl = dbl;
    printf("%f \n",dbl);
    printf("isnan = %d \n",isnan(dbl));
    printf("%x %x\n", uldbl[0], uldbl[1]);
    printf("\n");

    printf("test atof(\"-1.#IND00\") \n");
    dbl = atof("-1.#IND00");
    *( double* )uldbl = dbl;
    printf("%f \n",dbl);
    printf("isnan = %d \n",isnan(dbl));
    printf("%x %x\n", uldbl[0], uldbl[1]);
    printf("\n");

    printf("test atof(\"1.#QNAN0\") \n");
    dbl = atof("1.#QNAN0");
    *( double* )uldbl = dbl;
    printf("%f \n",dbl);
    printf("isnan = %d \n",isnan(dbl));
    printf("%x %x\n", uldbl[0], uldbl[1]);
    printf("\n");

    printf("test 0 \n");
    dbl = 0;
    *( double* )uldbl = dbl;
    printf("%f \n",dbl);
    printf("isnan = %d \n",isnan(dbl));
    printf("%x %x\n", uldbl[0], uldbl[1]);
    printf("\n");

    return 0;
}
*************************
The results are

$ ./nan
test 0.0/0.0
-1.#IND00
isnan = 1
0 fff80000

test 0xffffffff 0x7fffffff
1.#QNAN0
isnan = 1
ffffffff 7fffffff

test atof("NaN")
2293504.000000
isnan = 0
0 41417f80

test atof("-1.#IND00")
2293504.000000
isnan = 0
0 41417f80

test atof("1.#QNAN0")
2293504.000000
isnan = 0
0 41417f80

test 0
0.000000
isnan = 0
0 0

#******************

>From above it seems to be

	unsigned long lnan[2]={0xffffffff, 0x7fffffff};
    return *( double* )lnan;

is applicable as for the MSC and djgpp.

So I changed from
#if defined (__MSC__) || defined (DJGPP) || defined(__DJGPP__) 
to
#if defined (__MSC__) || defined (DJGPP) || defined(__DJGPP__) || defined (__MinGW32__)
but pr NaN gives 0.0.

This is the current situation.

Any suggestions?

Regards

Tatsuro

--------------------------------------
Learn more about breast cancer - Pink Ribbon Campaign 2010
http://yj.pn/JAy9L7