Hi, folks

My test environment is GCC (4.7.1), GDB (7.6) and uclibc (0.9.33.2 commit 
ff22cca4).

The testcase is:

        $ cat long_long_simple.c
        #include <stdio.h>

        int main()
        {
                long long val = 01234567123456701234567LL; /* 21+ octal digits 
*/
                printf("%.17g\n", val);
                return 0;
        }

Debug the program with UCLIBC_HAS_OBSTACK *disabled*:

        sh-4.2# ./long_long_simple
        -5.9822653797615731e-120

        sh-4.2# gdb ./long_long_simple
        GNU gdb (GDB) 7.6
        [...]
        (gdb) b 6
        Breakpoint 1 at 0x84f0: file long_long_simple.c, line 6.
        (gdb) r
        Starting program: /tmp/long_long_simple

        Breakpoint 1, main () at long_long_simple.c:6
        6       long_long_simple.c: No such file or directory.
        (gdb) p val
        $1 = -6399925985474168457
        (gdb) p/o val
        $2 = 01234567123456701234567
        (gdb) p/f val
        $3 = -5.9822653797615723e-120

Debug the program with UCLIBC_HAS_OBSTACK *enabled*:

        sh-4.2# ./long_long_simple
        -5.9822653797615731e-120

        sh-4.2# gdb long_long_simple
        GNU gdb (GDB) 7.6
        [...]
        (gdb) b 6
        Breakpoint 1 at 0x84f0: file long_long_simple.c, line 6.
        (gdb) r
        Starting program: /tmp/long_long_simple

        Breakpoint 1, main () at long_long_simple.c:6
        6       long_long_simple.c: No such file or directory.
        (gdb) p val
        $1 = -6399925985474168457
        (gdb) p/o val
        $2 = 01234567123456701234567
        (gdb) p/f val
        $3 = -5.9822653797615731e-120

>From the value of `$2', we could see that GDB get the correct value of `val'.
But if we format the value as `p/f', the two values of `$3' are not the same:
$3 = -5.98226537976157*23e*-120    UCLIBC_HAS_OBSTACK disabled, this is the 
expected value
$3 = -5.98226537976157*31e*-120    UCLIBC_HAS_OBSTACK enabled, this value is 
not correct

I have no idea if the problem has anything to do with 
http://lists.uclibc.org/pipermail/uclibc/2015-May/048932.html
Did anyone ever hit this case? Any suggestion is appreciated.

Thanks,
Sheng

_______________________________________________
uClibc mailing list
[email protected]
http://lists.busybox.net/mailman/listinfo/uclibc

Reply via email to