Re: Input won't accept floats

[Francisco Jesús Martínez Serrano]

On 5/18/06, John W. Eaton wrote:
> On 18-May-2006, Francisco Jesús Martínez Serrano wrote:
>
> How can we possibly help given the above information?  Your code could
> be doing anything, but you haven't shown it to us, so the best we can
> do is guess.  My guess is your code does something to alter the output
> format state for Octave, but again, that's just a guess.

Sorry for being so imprecise. It's a compiled function yes, but in
fact it's a whole 10000+ lines standalone visualization program
written in C using libglade and OpenGL, to which i added some hook
functions using matwrap, to be able to start it as a separate thread
and then use octave as scripting language to set parameters and get
data to make 2D plots and so on.

Interesting bits of code:

pthread_t glthread;

void irhys_start()
{
       pthread_create(&glthread,0,init_irhys_thread,0);
}

void *init_irhys_thread(void* foo)
{
       [...]
       gtk_init(&argcp, &argv);
       gtk_gl_init (&argcp, &argv);
       [...]
}

I make the irhys_start function available with matwrap (all the hooks
inside the same header file):

#ifdef __cplusplus
extern "C" {
#endif

void irhys_start(void); //This function causes the problem
void irhys_print_options(); //This dumb function (just printfs) is fine

[the rest of the hooks that work correctly]
#ifdef __cplusplus
}
#endif


Previously the program was written using glut and this was not
happening, so I assume it is some interaction between octave and
libglade/gtk/you name it, though in principle they should be
completely separated (i used to believe the visualization program was
reentrant, with GLUT it was working fine).

It is indeed a real input problem, the output is not just messed:

octave:5> 1/0.5
warning: division by zero
ans = Inf

however if one works with variables:

octave:6> a=1/2
a = 0.50000
octave:7> 1/a
ans = 2

Anyway, since I've never looked inside the octave code I just thought
that it might be a recognisable problem for those that are used to
deal with its internals.

I guess i'll have to recompile Octave in debug mode and do some debug
(maybe also gtk+).

I'll report if i find something that might be of interest to the
octave community.

Thanks.