Keep in mind to mention PR12354936 in the diff sum up so people can grep for it easily.

Thanks

If there is a race condition, then the proper solution is a lock or something lock free using atomics, not a null check that has no ordering constraint.

Note that you find yourself blocked because you made this depend on a test, however, the test doesn't seems to be testing any of that code.

I think we may want to keep this around for QUIC.

While the change isn't bad per se, this is actually not helping. This code is maintained today by the Core developers and taking ownership of it just increase our maintenance burden. Without a direct benefit, this is hard to justify.

Fix diff title/description.

As The GuessVerificationProgress doc says:

Clearly, size_t was the wrong type.

rebase

Remove unecessary dependency

Rebase

Nits in the test

fix various style things

I'm not sure why so much parameters are passed down as environment variables, that sounds fragile to me.

Where is the rest of it?

Backport the doc with the behavior change. These are not separable.

Check the test suite.

What is this doing? Why is this necessary? Clearly, this isn't removing unnecessary NONNEGATIVE_SIGNED, because these never were there in the first place.

The description give no real information as to what this is actually doing or testing.