Almost sixty years since Landauer linked the erasure of information with an increase of entropy, his famous erasure principle and byproducts like reversible computing are still subjected to debates in the scientific community. In this work we use the Liouville theorem to establish three different types of the relation between manipulation of information by a logical gate and the change of its physical entropy, corresponding to three types of the final state of environment. A time-reversible relation can be established when the final states of environment corresponding to different logical inputs are macroscopically distinguishable, showing a path to reversible computation and erasure of data with no entropy cost. A weak relation, giving the entropy change of $k ln 2$ for an erasure gate, can be deduced without any thermodynamical argument, only requiring the final states of environment to be macroscopically indistinguishable. The common strong relation that links entropy cost to heat requires the final states of environment to be in a thermal equilibrium. We argue in this work that much of the misunderstanding around the Landauers erasure principle stems from not properly distinguishing the limits and applicability of these three different relations. Due to new technological advances, we emphasize the importance of taking into account the time-reversible and weak types of relation to link the information manipulation and entropy cost in erasure gates beyond the considerations of environments in thermodynamic equilibrium.