OpenMP reduction issue with Cython

I am facing an unexpected behavior of Cython (v. 0.29.13) when compiling OpenMP parallel code that is expect to perform a reduction:

import cython
from cython.parallel import prange, parallel

cpdef omp_test1(n):
    cdef int sum = 0, i, imax
    imax = <int>n
    for i in prange(imax, nogil=True):
        sum += 1
    return sum


cpdef omp_test2(n):
    cdef int sum = 0, i, imax
    imax = <int>n
    with nogil, parallel():
        for i in prange(imax):
            sum += 1
    return sum

When calling both functions I would expect to have as a return value the input argument n. Instead, the returned value is n * num_threads. Surprisingly, on my Mac this unexpected behavior is observed only with GCC (v. 9.2.0_2); the clang (v. 11.0.0) returns the expected value n.

Is there something wrong I am doing or is there a problem with the generated Cython code or GCC compiler?