Re: examining floating point data with 'print' on non floating point kernel

[Mark Kettenis <mark dot kettenis at xs4all dot nl>]

   Date: Thu, 14 Jul 2005 21:16:27 -0400
   From: Daniel Jacobowitz <drow@false.org>
	   version=3.0.2

   On Thu, Jul 14, 2005 at 08:08:22PM -0500, Claudia Salzberg wrote:
   > Hi,
   > 
   > I am trying to use the 'print' command on a function declared in
   > the program being debugged.  This function takes in 10 values and
   > prints them with a printf(%f).  The program prints the values
   > correctly if run independently of gdb.  If the program is run on
   > gdb normally (meaning with just run) they also print correctly.
   > However, when called with the print command the output is
   > unexpected and it prints either 0.00000 or nan where the various
   > floating point values should be.  This works fine on a power 3
   > ppc box but does not work on the 440GP.

   This means that your 440GP installation is using a different floating
   point model than GDB thinks it is.

Or the Linux kernel on the 440GP doesn't properly initialize the FPU
when gdb writes the floating-point registers before the program itself
has executed an FPU instruction.

Mark