[OpenACC] Stop trying to analyze invalid Var-Decls.

The code to analyze VarDecls for the purpose of ensuring a magic-static
isn't present in a 'routine' was getting confused/crashed because we
create something that looks like a magic-static during error-recovery,
but it is still an invalid decl.

This patch causes us to just 'give up' in the case where the vardecl is
already invalid.

Fixes: #140920

Author: erichkeane

clang/lib/Sema/SemaOpenACC.cpp

@@ -1756,7 +1756,7 @@ ExprResult SemaOpenACC::ActOnRoutineName(Expr *RoutineName) {
   return ExprError();
 }
 void SemaOpenACC::ActOnVariableDeclarator(VarDecl *VD) {
-  if (!VD->isStaticLocal() || !getLangOpts().OpenACC)
+  if (!getLangOpts().OpenACC || VD->isInvalidDecl() || !VD->isStaticLocal())
     return;
 
   // This cast should be safe, since a static-local can only happen in a

clang/test/SemaOpenACC/gh140920.cpp

@@ -0,0 +1,19 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+// Ensure that we are properly handling 'vardecl' when they are created during
+// error recovery. The errors themselves aren't really relevant/necessary to the
+// bug fix.
+struct Thing{ };
+struct pair {
+  // expected-error@+2{{no member named 'T1'}}
+  // expected-error@+1{{expected a qualified name after 'typename'}}
+  template <typename enable_if<Thing::template T1<int>() &&
+                                   !Thing::template T1<int>(),
+    // expected-error@+4{{non-friend class member 'type' cannot have a qualified name}}
+    // expected-error@+3{{type specifier is required}}
+    // expected-error@+2{{non-static data member 'type' cannot be declared as a template}}
+    // expected-error@+1{{no member named 'type' in the global namespace}}
+                               bool>::type = false>
+  // expected-error@+1{{expected '(' for function-style cast or type construction}}
+  void func(void);
+};