1 files changed, 30 insertions, 0 deletions
diff --git a/tests/ui/const-generics/generic_const_exprs/const_kind_expr/relate_ty_with_infer_1.rs b/tests/ui/const-generics/generic_const_exprs/const_kind_expr/relate_ty_with_infer_1.rs
new file mode 100644
index 000000000..1e2484118
--- /dev/null
+++ b/tests/ui/const-generics/generic_const_exprs/const_kind_expr/relate_ty_with_infer_1.rs
@@ -0,0 +1,30 @@
+// check-pass
+#![feature(generic_const_exprs)]
+#![allow(incomplete_features)]
+
+// issue #107899
+// We end up relating `Const(ty: size_of<?0>, kind: Value(Branch([])))` with
+// `Const(ty: size_of<T>, kind: Value(Branch([])))` which if you were to `==`
+// the `ty` fields would return `false` and ICE. This test checks that we use
+// actual semantic equality that takes into account aliases and infer vars.
+
+use std::mem::size_of;
+
+trait X<T> {
+    fn f(self);
+    fn g(self);
+}
+
+struct Y;
+
+impl<T> X<T> for Y
+where
+    [(); size_of::<T>()]: Sized,
+{
+    fn f(self) {
+        self.g();
+    }
+    fn g(self) {}
+}
+
+fn main() {}