Issue with `lazysegtree`

[guoyiz23]

#include <atcoder/lazysegtree>

using namespace std;
using namespace atcoder;

using lint = long long;

lint op(lint a,lint b){return a+b;}
lint e(){return 0;}

lint mapping(lint f, lint x) { return x+f; }
lint composition(lint f, lint g) { return f+g; }
lint id() { return 0; }

int main() {
    #define si(x) int(x.size())
    vector<lint> a{2,5,5};
    lazy_segtree<lint, op, e, lint, mapping, composition, id> seg(a);
    auto res{0ll};
    for (int i=0; i<si(a); i++) {
        seg.apply(0,si(a),-a[i]);
        auto lower{seg.prod(0,i)}, higher{seg.prod(i+1,si(a))};
        printf("lower = %lld, higher = %lld\n", lower, higher);
        assert(lower <= 0);
        assert(higher >= 0);
        res -= lower;
        res += higher;
        seg.apply(0,si(a),a[i]);
    }

    return 0;
}

The above code prints

lower = 0, higher = 6
lower = -3, higher = 0
lower = 2, higher = 0

and fails the assert. This code is meant to compute the sum of absolute value of differences between all pairs of elements in a sorted array.

When I remove the if condition in https://github.com/atcoder/ac-library/blob/master/atcoder/lazysegtree.hpp#L135 as well as the one on L136, the code works on this case. I won't go into full detail for now, but basically, there is a node the value of which needs to be updated (to be equal to the sum of those of its two children), and that if condition is getting in the way.

I was led to this by https://atcoder.jp/contests/abc351/tasks/abc351_e.

Moreover, even after I removed that aforementioned if, the value of res at the end for a = {3, 3, 4} is 2 when it should be 4.

I actually examined the case of a = {2, 5, 5} closely (ran thru with many debug print statements) but have yet to look into that of a = {3, 3, 4}, would be happy to discuss. Feel free to contact me via email at guoyiz at yandex.ru. Thank you!

[guoyiz23]

Per yosupo06/library-checker-problems#1137 (comment) looks like I had used it wrong. Sorry about that. Surely, it was very difficult for me to believe that there could actually be something wrong with such a commonly used library.