Displaying IEEE 754 floating point representation in GDB?

Question

When I ask GDB to print a real in binary, I get this:

(gdb) p/t 5210887.5
$1 = 10011111000001100000111

According to this,

0 10010101 00111110000011000001111

is the expected value.

Aligning them,

         1 0011111000001100000111
0 10010101 00111110000011000001111

And it looks like GDB is giving me the integer representation after rounding. Thing is it was doing this with a variable at work as well. Using a declaration in a C program -

main()
{
  float f = 5210887.5;
}

and debugging it -

$ gcc -g -O0 floatdec.c -o floatdec
$ gdb floatdec
GNU gdb (GDB) Red Hat Enterprise Linux (7.2-60.el6_4.1)
Copyright (C) 2010 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-redhat-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /home/ /floatdec...done.
(gdb) b main
Breakpoint 1 at 0x804839a: file floatdec.c, line 3.
(gdb) r
Starting program: /home/ /floatdec 

Breakpoint 1, main () at floatdec.c:3
3         float f = 5210887.5;
(gdb) s
4       }
(gdb) p f
$1 = 5210887.5
(gdb) p/t f
$2 = 10011111000001100000111

same thing, it is showing me the integer representation. Is there no way to have GDB show me the floating point format?

Answer 1

The t format does indeed convert the value to an integer. From gdb: Output Formats:

t
   Print as integer in binary. The letter ‘t’ stands for “two”.

This is backed up by the code in gdb's print_scalar_formatted:

if (options->format != 'f')
    val_long = unpack_long (type, valaddr);

where unpack_long converts various types of values, including float, into a long.

A workaround is to take the address of the variable, cast it to an (int32 *), and dereference that.

(gdb) p f
$2 = 5210887.5
(gdb) p/t *(int32_t *)&f
$3 = 1001010100111110000011000001111
(gdb) p/t {int32_t}&f
$4 = 1001010100111110000011000001111

Or, at the language-independent level, using the x command:

(gdb) x/tw &f
0xbffff3f4: 01001010100111110000011000001111

This was done on an x86. Other-endian systems may produce a different bit pattern.