weird results about integer division

[jack dang]

*GMP 5.0.1*
*No particular configure options are used other than ./configure*
*gcc version 4.3.2
Linux 2.6.27.5-117.fc10.i686 #1 SMP*
*Test code*
*#include <stdio.h>
#include <stdlib.h>
#include <gmp.h>

main()
{
        mpz_t integ;
        mpz_t mask;
        mpz_t temp[25];
        int i;

        mpz_init (integ);
        mpz_init (mask);

        for(i = 0; i < 25; i++)
                mpz_init (temp[i]);

        mpz_set_ui (mask, 0x01f);
        for(i = 1; i < 25; i++)
        {
                mpz_add_ui(integ, integ,i);
                mpz_mul_2exp(integ, integ, 5);
        }

        mpz_add_ui(integ, integ, i);

        for(i = 0; i < 25; i++)
        {
                mpz_and(temp[i], integ, mask);
                mpz_tdiv_q_2exp(integ, integ, 5);       // use tdiv_ or
fdiv_, but cdiv will cause wrong results
        }

        for(i = 0; i < 25; i++)
        {
                mpz_get_ui(temp[i]);
                mpz_out_str(NULL, 10, temp[i]);
                printf("  ");
        }
        printf("\n");

        mpz_clear(integ);
        mpz_clear(mask);
    }*

   The program is to store integers 0 ~ 25, five bits for each number, in a
128-bit integer. Then print them out.
    However, if function *mpz_cdiv_q_2exp()* were used instead of *
mpz_fdiv_q_2exp()* or *mpz_tdiv_q_2exp()*, the results would be
    25  25  24  23  22  21  20  19  18  17  16  15  14  13  12  11  10  9
8  7  6  5  4  3  2
    I am not familiar with rounding modes, and don't know how this was
implemented in the library. Guess might be a bug or  I misused the library
in some way.
    The expected result is
     25  24  23  22  21  20  19  18  17  16  15  14  13  12  11  10  9  8
7  6  5  4  3  2  1

     Just compile the program as following
     gcc test.c -lgmp -o test

-- 
Wei Dang