stabilize never type

[WaffleLapkin]

View all comments

This PR:

• stabilizes the never type (!!!)
• sets the never type fallback to ! on all editions (breaking change, see crater report analysis and refresher on never type fallback)
• removes dependency_on_unit_never_type_fallback lint (there is no more never type fallback to () so this lint can't be triggered)

Cat:

Tracking:

• Tracking issue for promoting ! to a type (RFC 1216) #35121
• Tracking Issue for never type fallback change #148922

Stabilization plan FCP:

• Edition 2024: Make ! fall back to ! #123508

Lint bump:

• Report never type lints in dependencies #141937

Rust 2024 edition change:

• Tracking Issue for Rust 2024: Make ! fall back to ! #123748
• Edition 2024: Make ! fall back to ! #123508

[rustbot]

⚠️ Warning ⚠️

• Some commits in this PR modify submodules.

  If this was not intentional, see I changed a submodule on accident in the rustc dev guide.

[WaffleLapkin]

@bors try

[rust-bors]

☀️ Try build successful (CI)
Build commit: 1c4199d (1c4199da70f816f2239f41c55168ff75a3663718, parent: b2f1ccf524a3a4cf9c34545167cc23b659cf1cbd)

[jackh726]

As a reviewer and attendee of the lang meeting yesterday:

There was some concern about the amount of breakage in the crater run for this (~3300 crates).

Some next steps:

• @WaffleLapkin said that they would run a similar crater run, but without changing fallback on edition < 2024
  • If there is less breakage, that might be a better stabilization target
  • There may be some "common pattern" in there that we could write a targeted hack for to smooth over the breakage for now (this would be on the types team to figure out)
• It would be good to try to get some of the crates with more dependents from the crater run to release patches for older versions that could fix a larger portion of the breakage
  • This can/should happen in parallel

Ultimately, I get the sense that we would accept this breakage if we have to, but it would be good to dot our I's and cross our T's on minimizing.

Given this, as a reviewer, marking as blocked. @WaffleLapkin let me know if there's anything you need to me to do to help move this forward.

[WaffleLapkin]

Current status on library backports:

• For picky-asn1-x509 (Backporting the fix for the never type fallback change Devolutions/picky-rs#479; ~45 crater issues), wl-clipboard-rs (Backporting the fix for the never type fallback change YaLTeR/wl-clipboard-rs#78; ~213 crater issues), and ashpd (Backporting the fix for the never type fallback change bilelmoussaoui/ashpd#434; ~380 crater issues) maintainers released backports
• For redis (Backporting the fix for the never type fallback change redis-rs/redis-rs#2054; ~527 crater issues) I worked on the backports and a maintainer released them
• sqlx (Backporting the fix for the never type fallback change transact-rs/sqlx#4247; ~839 crater issues) and solana-client (Backporting the fix for the never type fallback change anza-xyz/agave#12174; ~340 crater issues) maintainers refused backports
• bluer (Backporting the fix for the never type fallback change bluez/bluer#189; ~97 crater issues) maintainers intend to make backports, but haven't yet
• trie-db (Backporting the fix for the never type fallback change paritytech/trie#231; ~291 crater issues) maintainers haven't responded yet

[nihohit]

Hi, the maintainer of the redis-rs crate here, just wanted to add thanks for @WaffleLapkin for the fixes - it took the work on itself, and fixed old versions that were in quite a sorry state CI-wise regardless of this change, while being polite and communicative. Thanks for the exemplary behavior!

[traviscross]

After the last lang call, I checked in with @WaffleLapkin about this. There are some minor matters affecting the library side (Infallible = !) that are being addressed. Libs-api OKed the breakage via FCP. For details, see:

• Make Infallible = ! #155924

These are the pending bits of type system work that had been referenced in our call. These do not affect the breakage that's in front of us due to changing the never-to-any coercion fallback type in older editions.

From us, what @WaffleLapkin wants is our OK on this stabilization (including this breakage) subject to these planned fixes landing.

That makes sense to me.

Proposal: Let's accept this stabilization — stabilize the never type, switch the never-to-any coercion fallback type to ! in Rust 2015, 2018, and 2021, and remove the dependency_on_unit_never_type_fallback lint — accepting the resulting breakage in older editions due to the fallback change,¹ subject to the unrelated fixes described in #155924 landing before this PR is merged.

This completes the last step of the plan for stabilizing ! that we approved in #123508 (comment).

Never say never.

@rfcbot fcp merge lang

Footnotes

1. For analysis related to accepting this breakage, see my remarks on #146167. ↩

[rust-rfcbot]

@traviscross has proposed to merge this. The next step is review by the rest of the tagged team members:

• @joshtriplett
• @nikomatsakis
• @scottmcm
• @tmandry
• @traviscross

No concerns currently listed.

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

cc @rust-lang/lang-advisors: FCP proposed for lang, please feel free to register concerns.
See this document for info about what commands tagged team members can give me.

[nikomatsakis]

@rfcbot reviewed

[tmandry]

Thanks so much @WaffleLapkin for driving this forward!

@rfcbot reviewed

[joshtriplett]

Never shall never stable'd be until
Great Waffle'd work to high Rust hill
Shall come against it.

(with apologies to the Scottish play)

[Qix-]

Thank you for pushing this @WaffleLapkin. It's going to be a huge improvement to a lot of code. Very exciting :)

[BoxyUwU]

cc @rust-lang/types

[jackh726]

Yeah, thinking about it, types probably should have been on this FCP (because of changes to fallback). In practice, I don't care all that much - and I don't think anyone else on the team will either.