clang-plugin: Enable RecursiveASTVisitor approach by default.

tools/clang/plugins/FindBadConstructsConsumer.cpp

Index: tools/clang/plugins/FindBadConstructsConsumer.cpp
diff --git a/tools/clang/plugins/FindBadConstructsConsumer.cpp b/tools/clang/plugins/FindBadConstructsConsumer.cpp
index c79a764eb953be319cb5e83fe766cee2bbaa8a3c..a6af112bdb2ec69dd4531a751bdfb5fcc6b379e7 100644
--- a/tools/clang/plugins/FindBadConstructsConsumer.cpp
+++ b/tools/clang/plugins/FindBadConstructsConsumer.cpp
@@ -279,7 +279,7 @@ void FindBadConstructsConsumer::CheckCtorDtorWeight(
       emitWarning(record_location,
                   "Complex class/struct needs an explicit out-of-line "
                   "constructor.");
-    } else {
+    } else if (options_.check_implicit_copy_ctors) {

[dcheng, 2016/02/16 23:03:50]

Since this flag is about implicit constructors, and this else branch seems to
handle lots of things, would it make sense to check this flag on line 293
instead?

Also, do the plugin tests need to be updated? I know that we hardcode
with-ast-visitor for the tests right now: should we update this flag?

[vmpstr, 2016/02/16 23:13:27]

The move ctor branch would have the same issues as the copy ctor branch (I
couldn't think of a better flag name though)

The branch on 319 also generated additional errors with the RAV approach. Let me
dig in a bit more to see what those were. 
Oops, done.

       // Iterate across all the constructors in this file and yell if we
       // find one that tries to be inline.
       for (CXXRecordDecl::ctor_iterator it = record->ctor_begin();