Fix #54224 (const promotion regression)

[oli-obk]

r? @eddyb

[TimNN]

Ping from triage @eddyb / @rust-lang/compiler: This PR requires your review.

[oli-obk]

r? @RalfJung

[RalfJung]

I am confused. #54224 says it's not a bug, and it also doesn't mention anything about this being a regression. And then the fix for an issue that something doesn't get promoted is to make more things NOT_CONST?

[oli-obk]

This is a regression introduced in #51990 (comment)

We'll want to relax this in the future, but for now we should not change behaviour

[RalfJung]

Ah, so right now we accidentally accept some of the code you added in that test case?

[RalfJung]

There still is a Qualif::NOT_CONST below a self.mode test at

rust/src/librustc_mir/transform/qualify_consts.rs

Line 738 in 670d07e

self.add(Qualif::NOT_CONST);

Is that correct? If yes, could you add a comment saying why this is only rejected for Fn but not for the other modes?

Same for

rust/src/librustc_mir/transform/qualify_consts.rs

Line 762 in 670d07e

self.add(Qualif::NOT_CONST);

[RalfJung]

Oh I see, the difference is that in the other cases we always emit a feature gate in the else. That's what went wrong here. Thanks!

@bors r+

[bors]

📌 Commit 76f8a90 has been approved by RalfJung

[bors]

⌛ Testing commit 76f8a90 with merge e9e27e6...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: RalfJung
Pushing e9e27e6 to master...