Rust 1.26.1 point release

[pietroalbini]

Note: there is no decision on whether 1.26.1 should be released yet. This is just a wishlist.

There are a few bad bugs in the 1.26.0 release, and there is a discussion (#50646 (comment)) about making a point release with a few backports. This issue tracks the PRs we would like to backport, and if we should do the point release at all.

cc @rust-lang/release @rust-lang/core

---

• Filter out missing components from manifests #49705: Filter out missing components from manifests
  • Reason: rustup can't download 1.26.0 on both ARM and FreeBSD (tier 2 platform) rustup update fails on FreeBSD for nightly due to missing docs #49462
• Fix fn main() -> impl Trait for non-Termination trait #50656: Fix fn main() -> impl Trait for non-Termination trait
  • Reason: that syntax was stabilized by accident impl Trait is allowed in the return type of main #50595
• Update stable channel RLS and Rustfmt #50794
  • Reason: RLS doesn't work on Windows RLS interferring with compilation after 1.26 #50646
• Update stable channel RLS and Rustfmt #50794 (same as RLS)
  • Reason: rustfmt badly formats things Unexpected string literal indents rustfmt#2695
• n/a
  • Reason: missing 1.26.0 release notes (Update RELEASES.md for 1.26.0 #49523) on the stable artifacts
• Fix issue #50811 (NaN > NaN was true). #50812: NaN > NaN
  • Reason: Fixes bug in const-eval

[leoyvens]

I'd like to propose #50694 for an eventual point release, it makes impl Trait work with a bunch of common traits that have a Self default (impl PartialEq, impl Add, ...) rather than stack overflow.

[BatmanAoD]

It sounds like #50646 has already been fixed in the beta channel. Does the backport require a new PR?

[sanxiyn]

#49462 affects ARM as well as FreeBSD. I think the point release is warranted.

[smmalis37]

I'd like to strongly vote in favor of a point release containing the fix for #50646.

[est31]

We should also backport #49523. Right now the distributed source of the 1.26.0 release artifacts don't contain release notes for the 1.26.0 release, which is bad.

[nrc]

We're getting quite a lot of reports of problems on the RLS repo, so I'd be in favour of a point release for the RLS. I'll put together a PR.

I'd also like to update rustfmt due to rust-lang/rustfmt#2695

[nrc]

PR for RLS/Rustfmt: #50794

[kennytm]

I nominate #50811 to be included.

[nrc]

Discussed and approved at today's core team meeting. We wanted to include RLS/Rustfmt, termination trait, NaN in consts, and missing manifests (they were all decided either to be high reward or low risk enough to justify inclusion).

[leoyvens]

@nrc Did #50694 get discussed? It's already accepted and merged for beta, got 3 issue reports and the patch was minor.

[nrc]

Did #50694 get discussed? It's already accepted and merged for beta, got 3 issue reports and the patch was minor.

No. We missed that one. Ping @rust-lang/core - how do we feel about another inclusion?

[nrc]

iiuc it is not a stability concern, just a bug-fix since we currently stack overflow. It looks like a non-trivial patch to me, but I'm not in a great place to judge. Seems like we could live without it.

[eedahl]

Added some recent errors to #50646, currently happening on nightly, jic they're of relevance.

[Mark-Simulacrum]

The point release PR is up. I believe it includes all items we'd like to; if your desired PR wasn't included please let me know.

[sinesc]

Is this testable via rustup or download?

[pietroalbini]

There is no build yet, after the PR will be merged a build will appear in the dev-static bucker.

[Mark-Simulacrum]

Release artifacts are built, you can run RUSTUP_DIST_SERVER=https://dev-static.rust-lang.org rustup update stable to test it locally -- please report problems at https://internals.rust-lang.org/t/rust-1-26-1-prerelease-testing/ or here.