Bug in mpfr_pow() function

[Kees-Jan van der Kolk]

Hello,

I just ran into a bug in the mpfr_pow() function, in gmp version
4.1.2 (configured with --enable-mpfr). When mpfr_pow() is given a 
negative integer exponent, it treats the exponent as if it were 
positive:

    mpfr_pow(10, 3)    -> gives 1000, ok
    mpfr_pow(10, -3)   -> gives 1000, not ok

When using non-integer (floating point) arguments to mpfr_pow(), the
function seems to behave correctly.

A test-program is given below. Compilation and linking was done with:

    # gcc -o mpfrbug mpfrbug.c -lmpfr -lgmp

Gcc version 3.3 was used.

Of course, a workaround is easy to come up with, but it would be nice 
if the problem was fixed in a future release. Thank you in advance.

Kind regards,

Kees-Jan van der Kolk
Delft University of Technology
http://space.tudelft.nl


---[file: mpfrbug.c]--------------------------------------------------

#include <stdio.h>
#include <gmp.h>
#include <mpfr.h>

int main(int argc, char** argv)
{
    mpfr_t v1;
    mpfr_t v2;
    mpfr_t v3;

    mpfr_init(v1);
    mpfr_init(v2);
    mpfr_init(v3);

    mpfr_set_d(v1, 10, GMP_RNDN);
    mpfr_set_d(v2, -3, GMP_RNDN);
    mpfr_pow(v3, v1, v2, GMP_RNDN);

    fprintf(stdout, "should be 0.001: ");
    mpfr_out_str(stdout, 10, 0, v3, GMP_RNDN);
    fprintf(stdout, "\n");

    mpfr_clear(v1);
    mpfr_clear(v2);
    mpfr_clear(v3);

    return 0;
}

----------------------------------------------------------------------