Why acquiring multiple mutexes manually (m1.lock(); m2.lock();) frequently causes circular deadlocks:

If Thread A locks m1 then waits for m2, while Thread B locks m2 then waits for m1, both threads halt forever.

The Fix in C++17:

void TransferFunds(Account& from, Account& to, double amount) {
    // std::scoped_lock acquires any number of mutexes deadlock-free using RAII
    std::scoped_lock lock(from.m_mutex, to.m_mutex);
    from.m_balance -= amount;
    to.m_balance += amount;
}

std::scoped_lock uses a deadlock avoidance algorithm (like std::lock) and unlocks automatically upon scope exit!

std::scoped_lock with class template argument deduction (CTAD) is one of the best additions in C++17 concurrency.

Eliminates manual lock ordering boilerplate and prevents multi-threaded deadlock bugs completely.