https://issues.dlang.org/show_bug.cgi?id=24798
--- Comment #3 from kinke <[email protected]> --- To be clear, this is no double-destroy issue, as can be seen when printing the addresses of the destructed instances in the 2nd test case. The problem in the 2nd test case is that `MyMapResult.front()` returns a copy of `er.substs[0]`, with `er` already having been destructed (due to the actual issue), so it returns a copy of an `S(99)`, which then gets destructed, and the check then wrongly complaining about a double-destroy, while in reality it's a destruction of a copy of a destructed S instance. --
