Update Rust toolchain to nightly-2023-02-03

[tedinski]

Description of changes:

This PR is blocked on waiting for #2149 to be merged first.
Once that's merged, this PR should be rebased and force pushed.

The following upstream changes had impact this week:

• (Primitive::Pointer) abi: add AddressSpace field to Primitive::Pointer  rust-lang/rust#107248
• (ParamEnv) InstCombine away intrinsic validity assertions rust-lang/rust#105582
• (local_def_id) Remove HirId -> LocalDefId map from HIR. rust-lang/rust#107206
• (GeneratorWitnessMIR) Compute generator saved locals on MIR rust-lang/rust#101692
• (ConstEvalCounter) Use stable metric for const eval limit instead of current terminator-based logic rust-lang/rust#106227

This change caused all stubbing code to fail (CI will fail) and I could use help with it:

• make output_filenames a real query rust-lang/rust#106812

I think what we're seeing is that queries.global_ctxt()? is now "held" by rustc_driver across the after_expansion call, and consequently our use of queries.global_ctxt().unwrap().enter() fails here:

kani/kani-compiler/src/kani_middle/stubbing/annotations.rs

Line 43 in 4178958

queries.global_ctxt().unwrap().peek_mut().enter(|tcx| {

This might be a bug with that change in upstream rust? @celinval what do you think?

Resolved issues:

Resolves #2114

Testing:

• How is this change tested? ci

• Is this a refactor change? no

Checklist

• Each commit message has a non-empty body, explaining why the change was made
• Methods or procedures are documented
• Regression or unit tests are included, or existing tests cover the modified code
• My PR is restricted to a single feature or bugfix

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 and MIT licenses.

[celinval]

Yeah, it could've been a mistake on their side. @oli-obk was this restriction done on purpose?

We could work around it by moving the stubbing code to the after_analysis callback. My guess is that after_parsing might be too soon, since we deal with path resolution and the macro might've not been expanded yet.

[oli-obk]

Ah no, that's a bug on our end. I think we could just pass the TyCtxt into the after_analysis query, but short term we should just stop holding the lock on the rustc side

[celinval]

Ah no, that's a bug on our end. I think we could just pass the TyCtxt into the after_analysis query, but short term we should just stop holding the lock on the rustc side

@oli-obk do you think you can own the fix, or do you want us to submit a patch upstream?

[oli-obk]

I'll handle it, should be fixed in the next few days

[oli-obk]

The upstream patch has landed, but I think it didn't make today's nightly build, so you'll need to wait another day

[celinval]

Perfect! Thanks a lot @oli-obk.

[qinheping]

closed as #2291 created.