tracking issue for "chalkification"

[nikomatsakis]

This is a place holder for various refactorings going on in the trait system aimed, ultimately, at improving efficiency and correctness. Some of the key goals include lazy normalization, better caching, and removing many of the trait-related overflow errors. I plan to flesh out details better here as we go.

• Remove the "leak check" in favor of "universes"
  • @sgrif plans to do this; a lot of it has been done in an older branch
  • The plan is in this gist.
  • The primary goal here is to get the code in place so we can do a crater run, since this is not an entirely backwards compatible change (though it fixes some soundness bugs).
• Canonicalization + normalization query
  • @nikomatsakis is working on this presently, see introduce canonical queries, use for normalization and dropck-outlives #48411
• Integrate universes into canonicalization (handle universes in canonicalization #48696)
• re-remove the leak-check, which has been re-added (eventual goal: re-remove leak-check from compiler  #59490)
• implement lazy normalization (Tracking issue for "Lazy normalization" #60471)

[mark-i-m]

#48411 is done

[shepmaster]

For my own searchability, I'm adding the plain word chalk.

[tmandry]

See also the WG-traits meeting doc which has been updated more recently.

[ebkalderon]

For reference, some substantial work on universes has recently been merged: #55517.

[mark-i-m]

What's left here? It looks like all the checkboxes are ticked.

[vi]

Until the question is answered, this tracking issue no longer tracks the issue.

[pnkfelix]

Until @nikomatsakis corrects me otherwise, I'm going to assume that #59490 blocks this and thus this tracking issue once again tracks the issue.

[syntacticsugarglider]

Now that #65232 is merged, what is the status of this issue? What's the effective status of #59490?

[nikomatsakis]

Hard to summarize. We're working on making an approachable roadmap and summary on the https://github.com/rust-lang/wg-traits repository, but it's not quite there yet.

[joshlf]

Ping for status 🙂 Very excited for this (especially for implied bounds)!

[tmandry]

I recently looked and as I recall, roughly ~60% of (ui) tests were passing with -Zchalk.

[pnkfelix]

Visiting for backlog bonanza.

@rustbot label: S-tracking-needs-summary

[pnkfelix]

@rustbot label: S-tracking-impl-incomplete

[jackh726]

Visiting for types triage. We feel like this issue doesn't buy us much. It feels like a general "things to make rustc behave like chalk", which is probably better tracked through individual issues + a project. We're currently pursuing a-mir-formality, which will have implications here anyways.