`-Znext-solver` allow `ExprKind::Call` for not-yet defined opaques #145993

lcnr · 2025-08-29T09:34:27Z

Based on #146329. Revival of #140496. See the comment on OpaqueTypesJank. I've used the following document while working on this https://hackmd.io/Js61f8PRTcyaiyqS-fH9iQ.

Fixes rust-lang/trait-system-refactor-initiative#181. It does introduce one subtle footgun we may want to handle before stabilization, opened rust-lang/trait-system-refactor-initiative#230 for that. Also cc rust-lang/trait-system-refactor-initiative#231 for deref and index operations

r? @BoxyUwU

rustbot · 2025-08-29T09:34:33Z

BoxyUwU is currently at their maximum review capacity.
They may take a while to respond.

compiler/rustc_hir_typeck/src/callee.rs

bors · 2025-08-30T05:35:20Z

☔ The latest upstream changes (presumably #146023) made this pull request unmergeable. Please resolve the merge conflicts.

tests/ui/impl-trait/non-defining-uses/method-call-on-infer.rs

compiler/rustc_next_trait_solver/src/solve/assembly/mod.rs

lcnr · 2025-09-15T12:39:33Z

@bors try @rust-timer queue

`-Znext-solver` allow `ExprKind::Call` for not-yet defined opaques

rustbot · 2025-09-15T12:39:39Z

Some changes occurred to the core trait solver

cc @rust-lang/initiative-trait-system-refactor

compiler/rustc_hir_typeck/src/callee.rs

BoxyUwU · 2025-09-17T14:39:46Z

compiler/rustc_hir_typeck/src/callee.rs

                return None;
            }

+            ty::Infer(ty::TyVar(vid)) => {


feels subtle to me that we are specifically checking for TyVar instead of just all ty::Infer? How does the rest of the code handle int/float vars

can never impl Fn and Deref, so 🤷 we generally accept int/float infer vars, e.g. structurally_resolve_type also does so

BoxyUwU · 2025-09-17T14:40:05Z

compiler/rustc_hir_analysis/src/autoderef.rs

+        // To do so we don't eagerly bail if the current type is the hidden type of an
+        // opaque type and instead return `None` in `fn overloaded_deref_ty` if the
+        // opaque does not have a `Deref` item-bound.
+        if let &ty::Infer(ty::TyVar(vid)) = self.state.cur_ty.kind()


similar here, I realise we previously only checked is_ty_var, but how does the rest of this codepath handle int/float vars

they don't implement Deref so they are the final self type and users of autoderef have to handle them if necessary

compiler/rustc_hir_typeck/src/op.rs

BoxyUwU · 2025-09-17T14:44:17Z

compiler/rustc_hir_typeck/src/place_op.rs

+            imm_tr,
+            base_ty,
+            opt_rhs_ty,
+            TreatNotYetDefinedOpaques::AsInfer,


i guess its the same reasoning here?

BoxyUwU · 2025-09-17T14:44:21Z

compiler/rustc_hir_typeck/src/place_op.rs

+            mut_tr,
+            base_ty,
+            opt_rhs_ty,
+            TreatNotYetDefinedOpaques::AsInfer,


i guess its the same reasoning here?

compiler/rustc_next_trait_solver/src/solve/eval_ctxt/mod.rs

compiler/rustc_type_ir/src/solve/mod.rs

lcnr · 2025-09-18T11:01:03Z

@bors r=BoxyUwU :3

bors · 2025-09-18T11:01:07Z

📌 Commit 9913c47 has been approved by BoxyUwU

It is now in the queue for this repository.

bors · 2025-09-18T13:49:56Z

⌛ Testing commit 9913c47 with merge 4cd91ef...

bors · 2025-09-18T16:53:57Z

☀️ Test successful - checks-actions
Approved by: BoxyUwU
Pushing 4cd91ef to master...

github-actions · 2025-09-18T16:57:01Z

What is this?

This is an experimental post-merge analysis report that shows differences in test outcomes between the merged PR and its parent PR.

Comparing 32e3d9f (parent) -> 4cd91ef (this PR)

Test differences

Show 45 test diffs

Stage 1

[ui] tests/ui/impl-trait/non-defining-uses/ambiguous-ops.rs#current: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/ambiguous-ops.rs#next: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/call-expr-incorrect-choice-diagnostics.rs#current: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/call-expr-incorrect-choice-diagnostics.rs#next: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/deref-constrains-self-ty.rs#current: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/deref-constrains-self-ty.rs#next: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/function-call-on-infer.rs#current: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/function-call-on-infer.rs#next: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/impl-deref-function-call.rs#current: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/impl-deref-function-call.rs#next: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/shex_compat-regression-test.rs#current: [missing] -> pass (J1)
[ui] tests/ui/impl-trait/non-defining-uses/shex_compat-regression-test.rs#next: [missing] -> pass (J1)

Stage 2

[ui] tests/ui/impl-trait/non-defining-uses/ambiguous-ops.rs#current: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/ambiguous-ops.rs#next: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/call-expr-incorrect-choice-diagnostics.rs#current: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/call-expr-incorrect-choice-diagnostics.rs#next: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/deref-constrains-self-ty.rs#current: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/deref-constrains-self-ty.rs#next: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/function-call-on-infer.rs#current: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/function-call-on-infer.rs#next: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/impl-deref-function-call.rs#current: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/impl-deref-function-call.rs#next: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/shex_compat-regression-test.rs#current: [missing] -> pass (J0)
[ui] tests/ui/impl-trait/non-defining-uses/shex_compat-regression-test.rs#next: [missing] -> pass (J0)

Additionally, 21 doctest diffs were found. These are ignored, as they are noisy.

Job group index

Test dashboard

Run

cargo run --manifest-path src/ci/citool/Cargo.toml -- \
    test-dashboard 4cd91ef8223ef54111d21aa9e9e71b3b26477dd3 --output-dir test-dashboard

And then open test-dashboard/index.html in your browser to see an overview of all executed tests.

Job duration changes

aarch64-apple: 5458.3s -> 6428.0s (17.8%)
dist-android: 1571.8s -> 1301.8s (-17.2%)
x86_64-gnu-llvm-20: 2374.1s -> 2766.1s (16.5%)
aarch64-gnu-debug: 3726.1s -> 4337.8s (16.4%)
dist-x86_64-apple: 8144.5s -> 9396.2s (15.4%)
i686-gnu-1: 7141.0s -> 8050.5s (12.7%)
x86_64-gnu-miri: 4400.0s -> 4929.4s (12.0%)
dist-aarch64-windows-gnullvm: 4717.7s -> 5249.8s (11.3%)
arm-android: 5683.0s -> 6272.9s (10.4%)
aarch64-gnu-llvm-20-1: 3284.3s -> 3616.9s (10.1%)

How to interpret the job duration changes?

Job durations can vary a lot, based on the actual runner instance
that executed the job, system noise, invalidated caches, etc. The table above is provided
mostly for t-infra members, for simpler debugging of potential CI slow-downs.

rust-timer · 2025-09-18T18:05:39Z

Finished benchmarking commit (4cd91ef): comparison URL.

Overall result: ❌✅ regressions and improvements - no action needed

@rustbot label: -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.1%	[0.1%, 0.1%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.2%	[-0.3%, -0.0%]	3
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (secondary 3.8%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	4.6%	[4.1%, 5.0%]	7
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.0%	[-2.0%, -2.0%]	1
All ❌✅ (primary)	-	-	0

Cycles

Results (primary -2.1%, secondary -2.6%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	3.0%	[3.0%, 3.0%]	1
Improvements ✅ (primary)	-2.1%	[-2.1%, -2.1%]	1
Improvements ✅ (secondary)	-3.7%	[-4.6%, -2.6%]	5
All ❌✅ (primary)	-2.1%	[-2.1%, -2.1%]	1

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 471.73s -> 473.718s (0.42%)
Artifact size: 387.92 MiB -> 389.93 MiB (0.52%)

support opaque types in method selection See my notes in https://hackmd.io/4ILASx3mQ3u_gW9r1JyqCw. This PR builds on #145993 and allows not-yet defined opaque types as self types in the `method_autoderef_chain`. E.g. for `Box<impl Deref<impl Foo>>` this results in the autoderef chain `Box<impl Deref> -> ?deref_hidden_ty -> ?foo_hidden_ty`. Method selection stays ambiguous if the final autoderef step is still an infer var unless that var is an opaque. TODO: treating opaques as rigid jank. r? `@BoxyUwU`

rustbot assigned BoxyUwU Aug 29, 2025

lcnr mentioned this pull request Aug 29, 2025

Incompletely allow overloaded call from opaque when self type bottoms out in infer #140496

Closed

lcnr changed the title ~~-Znext-solver allow ExprKind::Call on not-yet defined opaques~~ -Znext-solver allow ExprKind::Call for not-yet defined opaques Aug 29, 2025