Re: embedding octave - floating point exception

[Marco Driusso]

On 2009/3/11 Jason Riedy wrote:
> Oy.  The tcl init code in ns-2 explicitly turns on trapping /
> interrupts for division by zero.  There's a bizarre comment
> before that code, but it looks like a conscious decision on their
> part.

Are you talking about this one?

static inline void
setup_floating_point_environment()
{
[....]
    /*
     * In general we'd like to catch some serious exceptions (div by zero)
     * and ignore the boring ones (overflow/underflow).
     * We set up that up here.
     * This depends on feenableexcept which is (currently) GNU
     * specific.
     */
    int trap_exceptions = 0;
[....]
    trap_exceptions |= FE_DIVBYZERO;
[....]
    trap_exceptions |= FE_INVALID;
[....]
    trap_exceptions |= FE_OVERFLOW;
[....]
//    trap_exceptions |= FE_UNDERFLOW;
[....]
    feenableexcept(trap_exceptions);
[....]
} 

in the .../ns-2.33/common/tkAppInit.cc ? If so, I've tried to change
these exceptions and remove the wrapping around the calls to octave,
but I got the older floating point exception. Debugging I realized that
running my simulation in ns-2 I don't call this function. And so, could
you please tell me the file  are you talking about, if is not this one?
Thank you so much for the collaboration, you're giving me a big help.

Marco