Remove inline_const from INCOMPLETE_FEATURES

[ghost]

#![feature(inline_const)] should be fully implemented now.

I didn't run tests locally because the compiler ran out of memory on my poor device, but I think it's fine because inline_const is only used in a few places.

r? @oli-obk
@rustbot label: F-inline_const

[oli-obk]

oh wow, we have no compilation failure tests for inline_const? We should definitely remedy that, but yea, not necessary in this PR.

I didn't run tests locally because the compiler ran out of memory on my poor device,

I'll wait for CI to pass, then I'll r+

[ghost]

oh wow, we have no compilation failure tests for inline_const? We should definitely remedy that, but yea, not necessary in this PR.

I found a feature gate test, but it doesn't contain anything related to #![allow(incomplete_features)].

[oli-obk]

@bors r+

Thanks for checking. I still think we should add some tests that enable the flag but cause compilation errors with const blocks, I have no clue how good/bad/weird such errors are right now.

[bors]

📌 Commit 4de0e8c has been approved by oli-obk

[lcnr]

we still have #78174 and #78132, both of which seem like quite fundamental challenges to me.

I think we should still keep inline_const in INCOMPLETE_FEATURES until at least #78174 is fixed.

[lcnr]

gonna @bors r- this for now. If you disagree with my assessment feel free to approve this again.

[ghost]

we still have #78174 and #78132, both of which seem like quite fundamental challenges to me.

I think we should still keep inline_const in INCOMPLETE_FEATURES until at least #78174 is fixed.

@lcnr #78174 should fail to compile, so it does not prevent correct uses of the feature. I think it should not block this PR? Also I think INCOMPLETE_FEATURES means the feature "may not be safe to use and/or cause compiler crashes", which is not the case of #78132.

Edit: Mentioned the wrong issue. Sorry.

[lcnr]

#![feature(inline_const)]

fn main() {
    const { &1 };
}

causes an ICE and is supposed to compile. https://play.rust-lang.org/?version=nightly&mode=debug&edition=2018&gist=50e7b8f6374b04372127fa8adc6f360c

[ghost]

#![feature(inline_const)]

fn main() {
    const { &1 };
}

causes an ICE and is supposed to compile. https://play.rust-lang.org/?version=nightly&mode=debug&edition=2018&gist=50e7b8f6374b04372127fa8adc6f360c

Oh, indeed. My bad. #78174 needs updating.

[lcnr]

To explain myself a bit better here:

I consider INCOMPLETE_FEATURES to be features which I would not recommend people to use except when experimenting. In my view inline_const is not yet working well enough for me to use it in any serious projects as it is quite easy to either break the compiler or get unexpected behavior when using it.

[ghost]

To explain myself a bit better here:

I consider INCOMPLETE_FEATURES to be features which I would not recommend people to use except when experimenting. In my view inline_const is not yet working well enough for me to use it in any serious projects as it is quite easy to either break the compiler or get unexpected behavior when using it.

Fair point. Maybe I'm too zealous about inline_const. 😕

[RalfJung]

I consider INCOMPLETE_FEATURES to be features which I would not recommend people to use except when experimenting. In my view inline_const is not yet working well enough for me to use it in any serious projects as it is quite easy to either break the compiler or get unexpected behavior when using it.

Now I am somewaht concerned that the standard library uses some of these features...

library/stdarch/crates/core_arch/src/lib.rs
5:#![allow(incomplete_features)]

library/alloc/src/lib.rs
73:#![allow(incomplete_features)]

library/core/src/lib.rs
65:#![allow(incomplete_features)]