Sema/AST: Fix a couple of protocol typealias validation order bugs

We break name lookup circularities by special-casing validation of
type aliases, resolving the underlying type of the type alias before
building the generic environment of the protocol that contains the
type alias.

However doing this gives the type alias an underlying type written
in terms of unresolved DependentMemberTypes which do not map to
associated types.

Attempting to substitute such a type fails.

We worked around this by re-validating type alias members of
protocols when we finally got around to building the protocol's
generic environment. Force this to happen earlier in
substMemberTypeWithBase(), because we need the fully resolved
underlying type in order to perform the substitution.

Also, fix TypeAliasDecl::setUnderlyingType() to only create a new
NameAliasType the first time it is called. In this special case
where it can be called twice with a resolved underlying type the
second time, we should not be creating a new NameAliasType.

There are better fixes possible here, requiring various levels
of refactoring. I'll investigate this soon, but for now this narrow
fix should address the problem with minimal risk.

Fixes <rdar://problem/33189068>.

Author: slavapestov

lib/AST/Decl.cpp

@@ -2421,14 +2421,19 @@ void TypeAliasDecl::setUnderlyingType(Type underlying) {
     underlying = mapTypeOutOfContext(underlying);
   UnderlyingTy.setType(underlying);
 
-  // Create a NameAliasType which will resolve to the underlying type.
-  ASTContext &Ctx = getASTContext();
-  auto aliasTy = new (Ctx, AllocationArena::Permanent) NameAliasType(this);
-  aliasTy->setRecursiveProperties(getUnderlyingTypeLoc().getType()
-      ->getRecursiveProperties());
-
-  // Set the interface type of this declaration.
-  setInterfaceType(MetatypeType::get(aliasTy, Ctx));
+  // FIXME -- if we already have an interface type, we're changing the
+  // underlying type. See the comment in the ProtocolDecl case of
+  // validateDecl().
+  if (!hasInterfaceType()) {
+    // Create a NameAliasType which will resolve to the underlying type.
+    ASTContext &Ctx = getASTContext();
+    auto aliasTy = new (Ctx, AllocationArena::Permanent) NameAliasType(this);
+    aliasTy->setRecursiveProperties(getUnderlyingTypeLoc().getType()
+        ->getRecursiveProperties());
+
+    // Set the interface type of this declaration.
+    setInterfaceType(MetatypeType::get(aliasTy, Ctx));
+  }
 }
 
 UnboundGenericType *TypeAliasDecl::getUnboundGenericType() const {

lib/Sema/TypeCheckType.cpp

@@ -3183,6 +3183,14 @@ Type TypeChecker::substMemberTypeWithBase(ModuleDecl *module,
   }
 
   if (auto *aliasDecl = dyn_cast<TypeAliasDecl>(member)) {
+    // FIXME: If this is a protocol typealias and we haven't built the
+    // protocol's generic environment yet, do so now, to ensure the
+    // typealias's underlying type has fully resoved dependent
+    // member types.
+    if (auto *protoDecl = dyn_cast<ProtocolDecl>(aliasDecl->getDeclContext()))
+      if (protoDecl->getGenericEnvironment() == nullptr)
+        validateDecl(protoDecl);
+
     if (aliasDecl->getGenericParams()) {
       return UnboundGenericType::get(
           aliasDecl, baseTy,

validation-test/compiler_crashers_2_fixed/0114-rdar33189068.swift

@@ -0,0 +1,19 @@
+// RUN: %target-swift-frontend %s -typecheck
+
+struct Bar : BarProtocol {
+    typealias Element = Int
+}
+
+struct Foo: FooProtocol {
+    typealias Things = Bar
+    func thing() -> Thing {}
+}
+
+protocol BarProtocol {
+    associatedtype Element
+}
+
+protocol FooProtocol {
+    associatedtype Things: BarProtocol
+    typealias Thing = Things.Element
+}