Switch from Chalk to the next trait solver #20329

jackh726 · 2025-07-28T14:07:21Z

At this point, opening up for general review, though I don't think this is fully ready to merge.

Some checklist (non-exhaustive) of things to do before/after merge (feel free to edit):

Before:

Fix the tidy lints (currently, I see that we need to add doc tests to the new files)
Squash (or just rework) commits
Review all the changed tests, open an issue to track the "real" regressions
Go through the todo!()s and FIXMEs and decide which are needed/useful to fix prior to merge
Decide what (if anything) should get upstreamed in a separate PR(s) first
Switch with_db calls to ScopedDb::set_db (or decide on a better API)

After (before stable):

Test across different projects, find things that are "blocking" (probably after merge)
Figure out where memory opts can be made

After (any time):

Transition things to from chalk_ir to rustc_type_ir
- Inference table blocks a lot of things (e.g. normalization)
- Some queries are "easy" like layout_of_ty

Mixed:

General cleanup of things
- Reduce/remove allow(unused)

ChayimFriedman2 · 2025-07-31T15:40:42Z

crates/query-group-macro/src/queries.rs

@@ -48,7 +48,8 @@ impl ToTokens for TrackedQuery {
                quote!(#(#options),*)
            })
            .into_iter()
-            .chain(self.lru.map(|lru| quote!(lru = #lru)));
+            .chain(self.lru.map(|lru| quote!(lru = #lru)))
+            .chain(Some(quote!(unsafe(non_update_return_type))));


That should be changed before merge, right?

Sure, if someone can point me to how to fix it.

ChayimFriedman2

First round of review, I think that's what I will be able to do today.

crates/hir-ty/src/next_solver/interner.rs

crates/hir-ty/Cargo.toml

crates/hir-ty/src/tls.rs

ChayimFriedman2 · 2025-07-31T15:53:33Z

crates/hir-ty/src/lib.rs

@@ -29,20 +47,24 @@ mod infer;
 mod inhabitedness;
 mod interner;
 mod lower;
+mod lower_nextsolver;


Are all those X_nextsolver modules supposed to replace their original before merge?

I think that's pretty much impossible if we want this to land anytime soon. There is so much that relies on the non-next solver modules that goes beyond trait solving.

ChayimFriedman2 · 2025-07-31T15:58:51Z

crates/hir-ty/src/db.rs


    #[salsa::invoke(crate::drop::has_drop_glue)]
    #[salsa::cycle(cycle_result = crate::drop::has_drop_glue_cycle_result)]
    fn has_drop_glue(&self, ty: Ty, env: Arc<TraitEnvironment>) -> DropGlue;
+
+    #[salsa::invoke(crate::layout_nextsolver::layout_of_adt_query)]


The fully qualified names everywhere (crate::X_nextsolver::X) make the code harder to read, but I suppose we'll get rid of that once we remove the Chalk types.

Yeah, if there are specific places that make sense to remove them, let me know. But unfortunately a lot of them are kind of necessary.

crates/hir-ty/src/next_solver/interner.rs

jackh726 · 2025-08-01T15:43:29Z

Squashed and rebased version here: jackh726@5fa8436

Will switch this PR after a bit of review, just so people can see commits if they need.

Veykril

Just a quick glance

crates/hir-ty/src/consteval.rs

crates/hir-ty/src/next_solver/consts.rs

crates/hir-ty/src/next_solver/interner.rs

…ng, map unevaluated

crates/hir-ty/src/layout_nextsolver.rs

crates/hir-ty/src/tests/regression.rs

ChayimFriedman2 · 2025-08-07T17:01:10Z

crates/hir-ty/src/next_solver/interner.rs

+        if GLOBAL_CACHE.is_set() {
+            GLOBAL_CACHE.with(call)
+        } else {
+            GLOBAL_CACHE.set(&RefCell::new(GlobalCache::default()), || GLOBAL_CACHE.with(call))


Setting the TLS without context seems like it may cause (hard-to-debug) problems by sharing the cache across revisions, when things have changed.

rustbot added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jul 28, 2025

ChayimFriedman2 reviewed Jul 31, 2025

View reviewed changes

rust-cloud-vms bot force-pushed the next-trait-solver-querify branch from 0c5be1c to 57c7b7d Compare July 31, 2025 21:36

Implement next trait solver

5fa8436

Veykril reviewed Aug 1, 2025

View reviewed changes

Remove duplicate traits

f185653

rust-cloud-vms bot force-pushed the next-trait-solver-querify branch from 8c0dd60 to eb38567 Compare August 2, 2025 01:15

Use phantomdata instead of &'db ()

ab8258d

rust-cloud-vms bot force-pushed the next-trait-solver-querify branch from eb38567 to ab8258d Compare August 2, 2025 01:18

jackh726 and others added 5 commits August 2, 2025 01:40

Fix doc string tidy lint

e98f85d

Use explicit lifetimes in db

2eb10de

Map ConstKind::Error

cbefae0

Typo

2c2aad3

Clean up DbInterner

fbf45f7

Veykril force-pushed the next-trait-solver-querify branch from 93688bb to fbf45f7 Compare August 2, 2025 07:23

jackh726 added 4 commits August 2, 2025 22:01

Use chalk branch to fix const error tys, bless tests, fix setup_traci…

f5d4d98

…ng, map unevaluated

Map consts better

15481ec

Implement Pattern. Remove a couple todos.

e750ec6

Call variances_of directly

ec03a91

ShoyuVanilla mentioned this pull request Aug 5, 2025

Merge Trait and TraitAlias handling #20376

Open

jackh726 added 4 commits August 5, 2025 20:46

Remove some todos

89389a9

Cleanup a bit of mapping

3478c7f

Remove rest of todos

0b98208

Fix remaining clippy warnings

ca14780

ShoyuVanilla reviewed Aug 6, 2025

View reviewed changes

crates/hir-ty/src/layout_nextsolver.rs Outdated Show resolved Hide resolved

ShoyuVanilla reviewed Aug 6, 2025

View reviewed changes

crates/hir-ty/src/tests/regression.rs Show resolved Hide resolved

jackh726 added 2 commits August 6, 2025 14:07

Add some normalization code from rustc

1f1cedc

Forward layout_of_ty_query to layout_of_ty_ns

5fac002

jackh726 added 10 commits August 6, 2025 14:25

Call new_with

61e9b8c

Change all uses of DbInterner::new to DbInterner::new_with

1817cb9

Typo check unstalled -> not_stalled

baea230

Add FIXME comment for hint

da12891

Manual debug impl for ParamConst

36d7737

Fix clippy lints

197c29d

Go through FIXMEs for next solver stuff

e60303e

Don't intern trait_solve query. Implement compute_goal_fast_path.

a0d37f1

Clippy fix

63719bf

Upgrade chalk. And add salsa::attach until it works.

7dc4d72

ChayimFriedman2 reviewed Aug 7, 2025

View reviewed changes

jackh726 added 2 commits August 7, 2025 18:29

Remove tls reentrant map hack

abf2a9e

Clippy fix

16ba92f

Switch from Chalk to the next trait solver #20329

Are you sure you want to change the base?

Switch from Chalk to the next trait solver #20329

Uh oh!

Conversation

jackh726 commented Jul 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ChayimFriedman2 Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

jackh726 Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

ChayimFriedman2 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ChayimFriedman2 Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

jackh726 Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

ChayimFriedman2 Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

jackh726 Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jackh726 commented Aug 1, 2025

Uh oh!

Veykril left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ChayimFriedman2 Aug 7, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jackh726 commented Jul 28, 2025 •

edited

Loading