"Stabilizable" prototype for expanded const generics

[nikomatsakis]

Metadata
Point of contact	@BoxyUwU
Team(s)	lang, types
Goal document	2025h1/min_generic_const_arguments

Summary

Experiment with a new min_generic_const_args implementation to address challenges found with the existing approach to supporting generic parameters in const generic arguments.

Tasks and status

• Discussion and moral support (lang, types )
• Implementation and mentoring (@BoxyUwU)
• Implementation (@camelid, @compiler-errors)
• Reviewer (@compiler-errors)

[nikomatsakis]

This issue is intended for status updates only.

For general questions or comments, please contact the owner(s) directly.

[BoxyUwU]

Key Developments: #125915 landed which is necessary to start implementation work on min_generic_const_args, a draft outlining the ideas behind the proposed feature has also been written
Blockers: no
Help Wanted: not currently in need of anything

[BoxyUwU]

The past ~month has been spent figuring out some implementation constraints arising from DefId creation of anon consts. My time has also been somewhat diverted to other things though that should have cleared up now :-)

[BoxyUwU]

camlid has started working on using the new lowering schema for more than just const parameters which once done will allow us to introduce a min_generic_const_args feature gate. compiler-errors has been working on removing the eval_x methods on Const that do not perform proper normalization and are incompatible with this feature

[BoxyUwU]

min_generic_const_args now exists as a feature gate (though without any functionality only some gated refactorings), shouldn't be long before it has actual functionality behind it 🤞 The refactoring to remove all the eval_x methods on ty::Const has been completed so it will now be possible to correctly implement normalization for constants.

[BoxyUwU]

(nothing this month)

[nikomatsakis]

This is a continuing project goal, and the updates below this comment will be for the new period 2025h1

[BoxyUwU]

camelid has a PR up which is ~fully finished + reviewed which enables resolving and lowering all paths under min_generic_const_args. It's taken a while to get this bit finished as we had to take care not to make parts of the compiler unmaintainable by duplicating all the logic for type and const path lowering.

[BoxyUwU]

camelids PR has been merged, we now correctly (to the best of my knowledge) lower const paths under mgca. I have a PR open to ensure that we handle evaluation of paths to consts with generics or inference variables correctly, and that we do not attempt to evaluate constants before they have been checked to be well formed. I'm also currently mentoring someone to implement proper handling of normalization of inherent associated constants under mgca.