184442 2018-04-09 22:55:30 -0700 Bmalloc should have its own lockers instead of lock_guard and unique_lock 2018-04-09 22:55:46 -0700 1 1 1 Unclassified WebKit JavaScriptCore Safari Technology Preview Unspecified Unspecified NEW P2 Normal --- 184176 1 saam saam benjamin fpizlo ggaren gskachkov jfbastien keith_miller mark.lam msaboff rmorisset ticaiolima ysuzuki oldest_to_newest 1413206 0 saam 2018-04-09 22:55:30 -0700 std::unique_lock and std::lock_guard are incompatible types. With bug #184176, I need to use unique_lock in just a few places since we may drop it to wait on a condition variable. However, since we pass this locker to things we call, I need to change the function of a bunch of signatures that don't actually need the API unique_lock provides. Same with things that call this API, they now need to pass in unique_lock. A lot less code would be changed if we just had a class hierarchy like this: ``` // I'm just picking same names as std, but we could do something different. template <typename Lock> class LockGuard { public: LockGuard(Lock& lock) : m_lock(lock) { m_lock.lock(); } ~LockGuard(Lock& lock) { if (m_locked) m_lock.unlock(); } protected: Lock& m_lock; bool m_locked { true }; }; template <typename Lock> class UniqueLock : public LockGuard<Lock> { using Base = LockGuard<Lock>; public: UniqueLock(Lock& l) : Base(l) { } void lock() { m_locked = true; m_lock.lock(); } void unlock() { m_locked = false; m_lock.unlock(); } }; ``` Or something along those lines.