[SOLVED] How often to derive from boost::signals::trackable?

How often to derive from boost::signals::trackable?

When using Boost.Signals, boost allows you to derive from boost::signals::trackable in order to ease object/connection lifetime management (See the boost documentation).

I am in an early stage of my project and I am thinking, whether to derive from boost::signals::trackable in

every new class I write that might use Boost.Signals in the future
or only in classes I am sure that they will need the functionality of the trackable bas e-class

The main reason for th first approach would be to prevent me to forget deriving from boost::signals::trackable.

Also double deriverations like

class Foo : public Base, public boost::signals::trackable
{
};

get unnecessary.

On the other side, preventing memory-leaks shouldn't be a main design-aspect. Testing and profiling tools like valgrind should be used to detect memory leaks.

Which approach is more suitable for growing projects?

Solution

Note that Boost.Signals2 supersedes Boost.Signals. It has much more flexible and powerful tracking mechanism.

Although the library aims to provide a thread-safe solution for multi-threaded programs, the locking overhead can be avoided in a single-threaded environment by setting boost::signals2::dummy_mutex as signal's internal mutex.