Auto merge of #141681 - compiler-errors:fast-path-stalled, r=<try>

Fast path for stalled obligations on self ty

If we see that the `self` type of a goal is an infer var, then don't try to compute the goal at all, since we know that it'll be forced ambiguous.

This is currently only implemented when there are no opaques in the environment. We could extend it to check that the self type is not related to any already defined opaques via subtyping, but I'll leave that as a follow-up.

---

Also stall coerce and subtype predicates if both of their vars are not resolved to concrete types.

---

~~Also, we don't care if the goal is higher-ranked for the sized and copy/clone fast path.~~ pulling this out into another PR.

r? lcnr

Author: rust-bors[bot]

compiler/rustc_infer/src/infer/opaque_types/table.rs

@@ -53,7 +53,7 @@ impl<'tcx> OpaqueTypeStorage<'tcx> {
         assert!(entry.is_some());
     }
 
-    pub(crate) fn is_empty(&self) -> bool {
+    pub fn is_empty(&self) -> bool {
         let OpaqueTypeStorage { opaque_types, duplicate_entries } = self;
         opaque_types.is_empty() && duplicate_entries.is_empty()
     }

compiler/rustc_trait_selection/src/solve/delegate.rs

@@ -64,6 +64,16 @@ impl<'tcx> rustc_next_trait_solver::delegate::SolverDelegate for SolverDelegate<
         span: Span,
     ) -> Option<Certainty> {
         if let Some(trait_pred) = goal.predicate.as_trait_clause() {
+            if self.shallow_resolve(trait_pred.self_ty().skip_binder()).is_ty_var()
+                // We don't do this fast path when opaques are defined since we may
+                // eventually use opaques to incompletely guide inference via ty var
+                // self types.
+                // FIXME: Properly consider opaques here.
+                && self.inner.borrow_mut().opaque_types().is_empty()
+            {
+                return Some(Certainty::AMBIGUOUS);
+            }
+
             if trait_pred.polarity() == ty::PredicatePolarity::Positive {
                 match self.0.tcx.as_lang_item(trait_pred.def_id()) {
                     Some(LangItem::Sized)
@@ -115,6 +125,17 @@ impl<'tcx> rustc_next_trait_solver::delegate::SolverDelegate for SolverDelegate<
 
                 Some(Certainty::Yes)
             }
+            ty::PredicateKind::Subtype(ty::SubtypePredicate { a, b, .. })
+            | ty::PredicateKind::Coerce(ty::CoercePredicate { a, b }) => {
+                if self.shallow_resolve(a).is_ty_var() && self.shallow_resolve(b).is_ty_var() {
+                    // FIXME: We also need to register a subtype relation between these vars
+                    // when those are added, and if they aren't in the same sub root then
+                    // we should mark this goal as `has_changed`.
+                    Some(Certainty::AMBIGUOUS)
+                } else {
+                    None
+                }
+            }
             _ => None,
         }
     }

tests/ui/traits/next-solver/coherence/coherence-fulfill-overflow.stderr

@@ -1,13 +1,10 @@
-error[E0119]: conflicting implementations of trait `Trait` for type `W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<_>>>>>>>>>>>>>>>>>>>>>`
+error[E0119]: conflicting implementations of trait `Trait` for type `W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<_>>>>>>>>>>>>>>>>>>>>>>>`
   --> $DIR/coherence-fulfill-overflow.rs:12:1
    |
 LL | impl<T: ?Sized + TwoW> Trait for W<T> {}
    | ------------------------------------- first implementation here
 LL | impl<T: ?Sized + TwoW> Trait for T {}
-   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ conflicting implementation for `W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<_>>>>>>>>>>>>>>>>>>>>>`
-   |
-   = note: overflow evaluating the requirement `W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<_>>>>>>>>>>>>>>>>>>>>>: TwoW`
-   = help: consider increasing the recursion limit by adding a `#![recursion_limit = "20"]` attribute to your crate (`coherence_fulfill_overflow`)
+   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ conflicting implementation for `W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<W<_>>>>>>>>>>>>>>>>>>>>>>>`
 
 error: aborting due to 1 previous error