The intention of this line is "push all parent nodes of a l-th & (r-1)-th node" and if (((r >> i) << i) != r) is just a pruning for decreasing # of push.

So push((r - 1) >> i) is more intuitive for me. Is there any realistic merit (e.g. speed) in push(r >> i)?

Thank you for your quick reply and explanation.
I understand.

I tried speed test on AtCoder server and Library Checker server, but this removal does not affect execution time.

I would like to ask you why, in lazysegtree.hpp,
is the line 59 ( if (((r >> i) << i) != r) push(r >> i); )
different from
the line 91 (if (((r >> i) << i) != r) push((r - 1) >> i);) ?

If the operations of these lines have the same meaning, we should unify the writing.
Or, do these operations have different meanings?

I would like to ask you why, in lazysegtree.hpp,
is the line 59 ( if (((r >> i) << i) != r) push(r >> i); )
different from
the line 91 (if (((r >> i) << i) != r) push((r - 1) >> i);) ?

Nice catch, these lines do the same operation, and we should fix this.

The intention of this line is "push all parent nodes of a l-th & (r-1)-th node" and if (((r >> i) << i) != r) is just a pruning for decreasing # of push.

I agree with this intension and I think we should rewrite the line 59.
I'll create new pull request.