Do not use non-wf input expectations from fudge when checking function calls

compiler/rustc_hir_typeck/src/fn_ctxt/checks.rs

@@ -215,6 +215,29 @@ impl<'a, 'tcx> FnCtxt<'a, 'tcx> {
             self.check_place_expr_if_unsized(fn_input_ty, arg_expr);
         }
 
+        let formal_input_tys_ns;
+        let formal_input_tys = if self.next_trait_solver() {
+            // In the new solver, the normalizations are done lazily.
+            // Because of this, if we encounter unnormalized alias types inside this
+            // fudge scope, we might lose the relationships between them and other vars
+            // when fudging inference variables created here.
+            // So, we utilize generalization to normalize aliases by adding a new
+            // inference var and equating it with the type we want to pull out of the
+            // fudge scope.
+            formal_input_tys_ns = formal_input_tys
+                .iter()
+                .map(|&ty| {
+                    let generalized_ty = self.next_ty_var(call_span);
+                    self.demand_eqtype(call_span, ty, generalized_ty);
+                    generalized_ty
+                })
+                .collect_vec();
+
+            formal_input_tys_ns.as_slice()
+        } else {
+            formal_input_tys
+        };
+
         // First, let's unify the formal method signature with the expectation eagerly.
         // We use this to guide coercion inference; it's output is "fudged" which means
         // any remaining type variables are assigned to new, unrelated variables. This
@@ -238,37 +261,16 @@ impl<'a, 'tcx> FnCtxt<'a, 'tcx> {
                     let origin = self.misc(call_span);
                     ocx.sup(&origin, self.param_env, expected_output, formal_output)?;
 
-                    let formal_input_tys_ns;
-                    let formal_input_tys = if self.next_trait_solver() {
-                        // In the new solver, the normalizations are done lazily.
-                        // Because of this, if we encounter unnormalized alias types inside this
-                        // fudge scope, we might lose the relationships between them and other vars
-                        // when fudging inference variables created here.
-                        // So, we utilize generalization to normalize aliases by adding a new
-                        // inference var and equating it with the type we want to pull out of the
-                        // fudge scope.
-                        formal_input_tys_ns = formal_input_tys
-                            .iter()
-                            .map(|&ty| {
-                                // If we replace a (unresolved) inference var with a new inference
-                                // var, it will be eventually resolved to itself and this will
-                                // weaken type inferences as the new inference var will be fudged
-                                // out and lose all relationships with other vars while the former
-                                // will not be fudged.
-                                if ty.is_ty_var() {
-                                    return ty;
-                                }
-
-                                let generalized_ty = self.next_ty_var(call_span);
-                                ocx.eq(&origin, self.param_env, ty, generalized_ty).unwrap();
-                                generalized_ty
-                            })
-                            .collect_vec();
-
-                        formal_input_tys_ns.as_slice()
-                    } else {
-                        formal_input_tys
-                    };
+                    // Check the well-formedness of expected input tys, as using ill-formed
+                    // expectation may cause type inference errors, see #150316.
+                    for &ty in formal_input_tys {
+                        ocx.register_obligation(traits::Obligation::new(
+                            self.tcx,
+                            self.misc(call_span),
+                            self.param_env,
+                            ty::ClauseKind::WellFormed(ty.into()),
+                        ));
+                    }
 
                     if !ocx.try_evaluate_obligations().is_empty() {
                         return Err(TypeError::Mismatch);

tests/ui/coercion/fudge-inference/expectated-input-not-satisfying-fn-bounds-issue-149881.current.stderr

@@ -0,0 +1,22 @@
+error[E0277]: the size for values of type `[{integer}]` cannot be known at compilation time
+  --> $DIR/expectated-input-not-satisfying-fn-bounds-issue-149881.rs:15:24
+   |
+LL |     <[_]>::into_vec(id(Box::new([0, 1, 2])));
+   |                     -- ^^^^^^^^^^^^^^^^^^^ doesn't have a size known at compile-time
+   |                     |
+   |                     required by a bound introduced by this call
+   |
+   = help: the trait `Sized` is not implemented for `[{integer}]`
+note: required by an implicit `Sized` bound in `id`
+  --> $DIR/expectated-input-not-satisfying-fn-bounds-issue-149881.rs:10:7
+   |
+LL | fn id<T>(x: Box<T>) -> Box<T> {
+   |       ^ required by the implicit `Sized` requirement on this type parameter in `id`
+help: consider relaxing the implicit `Sized` restriction
+   |
+LL | fn id<T: ?Sized>(x: Box<T>) -> Box<T> {
+   |        ++++++++
+
+error: aborting due to 1 previous error
+
+For more information about this error, try `rustc --explain E0277`.

tests/ui/coercion/fudge-inference/expectated-input-not-satisfying-fn-bounds-issue-149881.next.stderr

@@ -0,0 +1,22 @@
+error[E0277]: the size for values of type `[{integer}]` cannot be known at compilation time
+  --> $DIR/expectated-input-not-satisfying-fn-bounds-issue-149881.rs:15:24
+   |
+LL |     <[_]>::into_vec(id(Box::new([0, 1, 2])));
+   |                     -- ^^^^^^^^^^^^^^^^^^^ doesn't have a size known at compile-time
+   |                     |
+   |                     required by a bound introduced by this call
+   |
+   = help: the trait `Sized` is not implemented for `[{integer}]`
+note: required by an implicit `Sized` bound in `id`
+  --> $DIR/expectated-input-not-satisfying-fn-bounds-issue-149881.rs:10:7
+   |
+LL | fn id<T>(x: Box<T>) -> Box<T> {
+   |       ^ required by the implicit `Sized` requirement on this type parameter in `id`
+help: consider relaxing the implicit `Sized` restriction
+   |
+LL | fn id<T: ?Sized>(x: Box<T>) -> Box<T> {
+   |        ++++++++
+
+error: aborting due to 1 previous error
+
+For more information about this error, try `rustc --explain E0277`.

tests/ui/coercion/fudge-inference/expectated-input-not-satisfying-fn-bounds-issue-149881.rs

@@ -0,0 +1,17 @@
+//@ revisions: current next
+//@ ignore-compare-mode-next-solver (explicit revisions)
+//@[next] compile-flags: -Znext-solver
+
+// FIXME(#149379): This should pass, but fails due to fudged expactation
+// types which are potentially not well-formed or for whom the function
+// where-bounds don't actually hold. And this results in weird bugs when
+// later treating these expectations as if they were actually correct..
+
+fn id<T>(x: Box<T>) -> Box<T> {
+    x
+}
+
+fn main() {
+    <[_]>::into_vec(id(Box::new([0, 1, 2])));
+    //~^ ERROR: the size for values of type `[{integer}]` cannot be known at compilation time
+}

tests/ui/coercion/fudge-inference/expectated-input-not-satisfying-fn-bounds-issue-89299.current.stderr

@@ -0,0 +1,16 @@
+error[E0277]: `dyn Trait + Send` cannot be unpinned
+  --> $DIR/expectated-input-not-satisfying-fn-bounds-issue-89299.rs:20:27
+   |
+LL |     let _x = Foo(Pin::new(&mut a));
+   |                  -------- ^^^^^^ the trait `Unpin` is not implemented for `dyn Trait + Send`
+   |                  |
+   |                  required by a bound introduced by this call
+   |
+   = note: consider using the `pin!` macro
+           consider using `Box::pin` if you need to access the pinned value outside of the current scope
+note: required by a bound in `Pin::<Ptr>::new`
+  --> $SRC_DIR/core/src/pin.rs:LL:COL
+
+error: aborting due to 1 previous error
+
+For more information about this error, try `rustc --explain E0277`.

tests/ui/coercion/fudge-inference/expectated-input-not-satisfying-fn-bounds-issue-89299.next.stderr

@@ -0,0 +1,16 @@
+error[E0277]: `dyn Trait + Send` cannot be unpinned
+  --> $DIR/expectated-input-not-satisfying-fn-bounds-issue-89299.rs:20:27
+   |
+LL |     let _x = Foo(Pin::new(&mut a));
+   |                  -------- ^^^^^^ the trait `Unpin` is not implemented for `dyn Trait + Send`
+   |                  |
+   |                  required by a bound introduced by this call
+   |
+   = note: consider using the `pin!` macro
+           consider using `Box::pin` if you need to access the pinned value outside of the current scope
+note: required by a bound in `Pin::<Ptr>::new`
+  --> $SRC_DIR/core/src/pin.rs:LL:COL
+
+error: aborting due to 1 previous error
+
+For more information about this error, try `rustc --explain E0277`.

tests/ui/coercion/fudge-inference/expectated-input-not-satisfying-fn-bounds-issue-89299.rs

@@ -0,0 +1,22 @@
+//@ revisions: current next
+//@ ignore-compare-mode-next-solver (explicit revisions)
+//@[next] compile-flags: -Znext-solver
+
+// FIXME(#149379): This should pass, but fails due to fudged expactation
+// types which are potentially not well-formed or for whom the function
+// where-bounds don't actually hold. And this results in weird bugs when
+// later treating these expectations as if they were actually correct..
+
+use std::pin::Pin;
+
+trait Trait {}
+
+impl Trait for i32 {}
+
+struct Foo<'a>(Pin<&'a mut (dyn Trait + Send)>);
+
+fn main() {
+    let mut a = 1;
+    let _x = Foo(Pin::new(&mut a));
+    //~^ ERROR: `dyn Trait + Send` cannot be unpinned
+}

tests/ui/coercion/fudge-inference/fn-ret-trait-object-propagated-to-inputs-issue-149379-1.rs

@@ -1,11 +1,9 @@
 //@ revisions: current next
 //@ ignore-compare-mode-next-solver (explicit revisions)
 //@[next] compile-flags: -Znext-solver
+//@ check-pass
 
-// FIXME(#149379): This should pass, but fails due to fudged expactation
-// types which are potentially not well-formed or for whom the function
-// where-bounds don't actually hold. And this results in weird bugs when
-// later treating these expectations as if they were actually correct..
+// A regression test for https://github.com/rust-lang/rust/issues/149379.
 
 fn foo<T>(x: (T, ())) -> Box<T> {
     Box::new(x.0)
@@ -14,6 +12,4 @@ fn foo<T>(x: (T, ())) -> Box<T> {
 fn main() {
     // Uses expectation as its struct tail is sized, resulting in `(dyn Send, ())`
     let _: Box<dyn Send> = foo(((), ()));
-    //~^ ERROR mismatched types
-    //~| ERROR the size for values of type `dyn Send` cannot be known at compilation time
 }

tests/ui/coercion/fudge-inference/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.current.stderr

@@ -1,72 +1,5 @@
-error[E0277]: the size for values of type `dyn Send` cannot be known at compilation time
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:32:55
-   |
-LL |     let _: Box<dyn Send> = field_to_box1(Foo { field: 1, tail: () });
-   |                                                       ^ doesn't have a size known at compile-time
-   |
-   = help: the trait `Sized` is not implemented for `dyn Send`
-note: required by an implicit `Sized` bound in `Foo`
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:10:12
-   |
-LL | struct Foo<T> {
-   |            ^ required by the implicit `Sized` requirement on this type parameter in `Foo`
-help: you could relax the implicit `Sized` bound on `T` if it were used through indirection like `&T` or `Box<T>`
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:10:12
-   |
-LL | struct Foo<T> {
-   |            ^ this could be changed to `T: ?Sized`...
-LL |     field: T,
-   |            - ...if indirection were used here: `Box<T>`
-
-error[E0308]: mismatched types
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:32:55
-   |
-LL |     let _: Box<dyn Send> = field_to_box1(Foo { field: 1, tail: () });
-   |                                                       ^ expected trait object, found integer
-   |
-   = note: expected trait object `dyn Send`
-                      found type `{integer}`
-
-error[E0277]: the size for values of type `dyn Send` cannot be known at compilation time
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:32:42
-   |
-LL |     let _: Box<dyn Send> = field_to_box1(Foo { field: 1, tail: () });
-   |                            ------------- ^^^^^^^^^^^^^^^^^^^^^^^^^^ doesn't have a size known at compile-time
-   |                            |
-   |                            required by a bound introduced by this call
-   |
-   = help: the trait `Sized` is not implemented for `dyn Send`
-note: required by an implicit `Sized` bound in `field_to_box1`
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:19:18
-   |
-LL | fn field_to_box1<T>(x: Foo<T>) -> Box<T> {
-   |                  ^ required by the implicit `Sized` requirement on this type parameter in `field_to_box1`
-help: consider relaxing the implicit `Sized` restriction
-   |
-LL | fn field_to_box1<T: ?Sized>(x: Foo<T>) -> Box<T> {
-   |                   ++++++++
-
-error[E0277]: the size for values of type `dyn Send` cannot be known at compilation time
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:36:38
-   |
-LL |     let _: &dyn Send = field_to_box2(&Bar { field: 1 });
-   |                        ------------- ^^^^^^^^^^^^^^^^^ doesn't have a size known at compile-time
-   |                        |
-   |                        required by a bound introduced by this call
-   |
-   = help: the trait `Sized` is not implemented for `dyn Send`
-note: required by an implicit `Sized` bound in `field_to_box2`
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:23:18
-   |
-LL | fn field_to_box2<T>(x: &Bar<T>) -> &T {
-   |                  ^ required by the implicit `Sized` requirement on this type parameter in `field_to_box2`
-help: consider relaxing the implicit `Sized` restriction
-   |
-LL | fn field_to_box2<T: ?Sized>(x: &Bar<T>) -> &T {
-   |                   ++++++++
-
 error[E0308]: mismatched types
-  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:38:38
+  --> $DIR/fn-ret-trait-object-propagated-to-inputs-issue-149379-3.rs:34:38
    |
 LL |     let _: &dyn Send = field_to_box3(&(1,));
    |                        ------------- ^^^^^ expected `&(dyn Send,)`, found `&({integer},)`
@@ -81,7 +14,6 @@ note: function defined here
 LL | fn field_to_box3<T>(x: &(T,)) -> &T {
    |    ^^^^^^^^^^^^^    --------
 
-error: aborting due to 5 previous errors
+error: aborting due to 1 previous error
 
-Some errors have detailed explanations: E0277, E0308.
-For more information about an error, try `rustc --explain E0277`.
+For more information about this error, try `rustc --explain E0308`.