DBL_MIN on egcs-1.0.3a alphaev56-dev-linux-gnu

[Rob Whitehurst <rob at physics dot unimelb dot EDU dot AU>]

The following code causes problems with egcs-1.0.3a on an alpha;

#include <float.h>
#include <stdio.h>
#define DBL_TWO 2.2250738585072018772E-308

int main() {
    double one = 1.0,
           jollySmall = DBL_MIN,
           justSmall = DBL_TWO;

     printf("A jolly small number is %.20G\n"
            "Not quite so small is   %.20G\n", jollySmall, justSmall);
     fflush(stdout);

    return (one < jollySmall);
}


The output from gdb-4.17 is as follows;

This GDB was configured as "alphaev56-unknown-linux-gnu"...
(gdb) r
Starting program: /home/rob/Lagrange/Reflexion/small
A jolly small number is 2.225073858507200889E-308
Not quite so small is   2.2250738585072018772E-308

Program received signal SIGFPE, Arithmetic exception.
0x1200007ec in main () at small.c:14
14          return (one < jollySmall);
(gdb) p one < jollySmall
Erroneous arithmetic operation.
(gdb) p one < justSmall
$1 = 0
(gdb)

Compiled with -mieee it doesn't dump core but the value
reported for the variable jollySmall is still wrong and beneath
the value of DBL_MIN recorded in float.h: (and gdb still grumbles)

(gdb) r
Starting program: /home/rob/Lagrange/Reflexion/small
A jolly small number is 2.225073858507200889E-308
Not quite so small is   2.2250738585072018772E-308

Breakpoint 1, main () at small.c:14
14          return (one < jollySmall);
(gdb) p one < jollySmall
Erroneous arithmetic operation.
(gdb) c
Continuing.

Program exited normally.
(gdb)

  Anybody know what is going on here?

   Rob Whitehurst