Currently, units (ABT_unit, ABTI_unit, or unit_t) are manipulated through the right APIs (e.g., u_create_from_thread) or accessing directly they internal fields. This is wrong, since those units are exposed to the user, who can create their own structures. They are also internally overloaded, such as for use by synchronization functions outside the intended scope of usage: pool operations.

These issues can be addressed by:
1- Always manipulate units through APIs where necessary
2- Use different encapsulation methods to build waiting queues in synchronization constructs instead of overloading ABTI_unit