C++ concurrency in action

Question

This is a question about the memory order listed in 'c++ concurrency in action'.

First, in listing 5.7; the consequence can be:

Thread A : x.store(true, release)

Thread B : x.load(acquire) returns true, y.load(acquire) returns false

Thread C : y.load(acquire) returns true, x.load(acquire) returns false

Thread D : y.store(true, release)

On the other hand, by 'release sequence':

• thread A's store -> thread B's x_load and
• by release sequence, -> thread C's x_load
• so thread C's x_load must be true, if thread B's y_load is false.

So I think it should be

Thread B : x.load(acquire) returns true, y.load(acquire) returns true

Thread C : y.load(acquire) returns true, x.load(acquire) returns false

or

Thread B : x.load(acquire) returns true, y.load(acquire) returns false

Thread C : y.load(acquire) returns true, x.load(acquire) returns true

Another question:

In listing 7.12 there is a paragraph,

'the easy way to do this is to make the fetch_add() in the successful-return branch use std::memory_order_release and the fetch_add() in the loop-again branch use std::memory_order_acquire. However, this is still overkill: only one thread does the delete , so only that thread needs to do an acquire operation. Thankfully, because fetch_add() is read-modify-write operation, it forms part of the release sequence, so you can do that with an additional load().'

I don't know what this paragraph means; where is the 'release sequence' ?