Does buffered_channel really need a power of two capacity ?

[Romain-Geissler-1A]

Hi,

In #111 you did write that the implementation of buffered_channel does rely on the fact that the capacity is a power of two. In the meantime, the implementation has changed, and from what I see now we have a more classical ring buffer with two points for the push and the pop location. So I don't see where the assumption that we have a power of two capacity still stands.

Also the documentation says we can store only capacity - 1 elements in this channel (The channel can hold only capacity - 1 elements, otherwise it is considered to be full), but reading the code I do feel we can store exactly up to capacity elements.

If I am right, I guess the documentation shall be updated to reflect these lack of constraints and the check for the capacity power of two caracteristic in the constructor of buffered_channel. That would make this "container" more like a "normal" one: it can really hold at max capacity elements (and not n - 1 as documented now) and can have arbitrary size.