make copy_within #[inline]

[oconnor663]

@briansmith suggested this in
#54236 (comment).

[rust-highfive]

r? @sfackler

(rust_highfive has picked a reviewer for you, use r? to override)

[sfackler]

We generally do not add #[inline] to methods that are already inlined.

[oconnor663]

@briansmith what do you think here? I don't know enough about the compiler to have a real opinion.

[briansmith]

I don't know under what circumstances the compiler automatically inlines functions. It seems reasonable to me to think that, although the compiler might inline something without a hint today, it might choose to not do so tomorrow. OTOH if the compiler is guaranteed to always inline the function (why?) then I agree it is redundant.

[sfackler]

it might choose to not do so tomorrow.

This is the compiler. We know exactly how it behaves, and if it stops behaving that way, we'll deal with this at that time. There are literally thousands of functions that we expect to be inlined without an explicit #[inline].

[oconnor663]

I'll close this PR on that basis. I'm curious to know more, though: Is inlining something that's determined only by looking at the inlined function itself? Like if the compiler inlines foo in one case, should I expect it to inline foo in all cases?

[hudson-ayers]

FWIW, I have run into several places where copy_within is not inlined when using opt-level=s. I suspect this may be related to the addition of #[track_caller] to the method.