use std::atomic for ControlValueAtomicBase

[m0dB]

fixes #13895 , using std::atomic, and replaces a pre-c++17 construction to terminate if the atomic or the ringbuffer implementation should be using by using std::atomic::is_always_lock_free

[Swiftb0y]

Note that there are similar changes in main already. #13574

[Swiftb0y]

also @m0dB you have pushed this branch (and another similar one) to upstream instead of fork. Was that intentional?

[m0dB]

also @m0dB you have pushed this branch (and another similar one) to upstream instead of fork. Was that intentional?

No, that was accidental. I normally manage to avoid this mistake, but occasionally it goes wrong.

[m0dB]

Note that there are similar changes in main already. #13574

Oh, too bad I didn't see that. At a first glance, it looks like we used to same approach, but I will do a diff.

[m0dB]

You went a bit further than I did (makes sense, since I merely wanted to fix the TSAN warnings, and you went for code quality).

I propose we cherry-pick your changes for 2.5.

[daschuer]

We are at the end of beta in the 2.5 branch and after 2.4.2 release this is the new stable.
Since this is basically only a warning fix. Let's not introduce it without a testing period.

Can we close this?

[m0dB]

Sure, I can move all further tsan work to main, though I can imagine that it would be nice to have it in a 2.5 dot release.