cg meeting 2021-03-02 (zulip thread)

What can we stabilize without having to supply generic params to anon consts?

loosen the ty/const param ordering restriction
const param defaults
- const evaluatable bounds?
  - start out by not supplying any generic params?
- should we allow const N: usize, const M: usize = N,
  - requires the hack already used for const args
- seems fine, requires stabilization report
  - mention lazy norm backcompat
const param types
- what types are allowed?
  - structural match
  - do we require explicit opt in? pcg issue
    - consensus seems to be yes
    - open question: are there worrying interactions with structural match?
- without val trees
  - c like enums? seems quite helpful and should already work
    - concern: might feel a bit arbitrary to users msg
- val trees ✨
- generic param types
  - requires changes to WithOptConstParam
- what types are allowed?
  - structural match
  - do we require explicit opt in? pcg issue
_ as array len and const argument
- rn always gets interpreted as a type
- add hir::GenericArg::Infer
- should we go ahead and solve this for paths/unit(()) at the same time?
  - not neccessarily, if only _ is easier

Not talked about during the meeting

inline consts
- typeck together with their parent
- const evaluatable bounds?
  - require them to not depend on params after typeck for now?