Boost block indirect sort segmentation fault

Following is the code:

template<int N>
void para_sort(std::vector<bfs_node<N>>& res) {
    #pragma omp parallel
    {
        boost::sort::block_indirect_sort(res.begin(), res.end(), compare_node<N>);
    }
}

The code runs without any exception when running with only one OMP thread. However, the code throws in an exception on running with multiple threads.

The code works fine with multiple threads when Boost::Parallel_stable_sort used.

I suspected it would be the strict weak ordering requirement that caused it raise exception. But in that case it should be raised with parallel_stable_sort and std::sort (sequential version).

Is boost::block_indirect_sort demands more stricter or weaker ordering constraint.

template<int N>
bool compare_node(const bfs_node<N>& lhs, const bfs_node<N>& rhs) {
    if (lhs.id == rhs.id) return false;
    for (int i = 0; i < N * 64; ++i) {
        bool lflag = in_set(lhs.id, i);
        bool rflag = in_set(rhs.id, i);
        if (lflag == rflag ) { continue; }
        if (lflag) return true;
        return false;
    }
    return false;
} // 

where id field here represents a unordered_set type data structure. For ex: id represents the set {1,2,3} and in_set(id, 1) will return true and otherwise false

Exception raised: Invalid read of size 16 ==2125== at 0x46C59F: void boost::sort::pdqsort_detail::pdqsort_loop<__gnu_cxx::__normal_iterator<bfs_node<1>*, std::vector<bfs_node<1>, std::allocator<bfs_node<1> > > >, std::less<bfs_node<1> >, false>(__gnu_cxx::__normal_iterator<bfs_node<1>*, std::vector<bfs_node<1>, std::allocator<bfs_node<1> > > >, __gnu_cxx::__normal_iterator<bfs_node<1>*, std::vector<bfs_node<1>, std::allocator<bfs_node<1> > > >, std::less<bfs_node<1> >, int, bool)