Thanks for the comment! Your example with new is valid and until recently I also thought that was the only way to cause a memory leak with emplace(). Until I read the following SO thread https://stackoverflow.com/questions/37060326/at-which-point-does-mapemplace-create-an-object, where T.C. explains that this could occur even with std::make_unique() due to underspecfied emplace(). Tthis appears to be the case when you are trying to emplace() an object where the key is already present.

Following the Clang-Tidy documentation as well as Meyers’s book, the memory leak could occur when a temporary object A is created but emplace() could not allocate space for a new item. We do not possess a pointer to the temporary object A so we could not destroy the object A after it is passed to emplace() function and the emplace() fails. The memory leak could occur in such example:

std::map<int, std::unique_ptr> m;
m.emplace(1, new("Ann",63));

Regarding the example code you showed us, I think there will be no memory leak at all because the temporary object is now an std::unique_ptr which has the capability of destroying itself when the program goes out of the scope. So even if the object A is successfully created, stored in std::unique_ptr and the emplace() fails, std::unique_ptr will destroy this object and prevent the memory leak.

Is there anything that I am missing in your example?