[PATCH] D55492: Implement Attr dumping in terms of visitors

[Aaron Ballman via Phabricator via cfe-commits]

aaron.ballman added inline comments.


================
Comment at: include/clang/AST/AttrVisitor.h:21
+
+namespace attrvisitor {
+
----------------
steveire wrote:
> aaron.ballman wrote:
> > I don't think the namespace adds value.
> I think the point is to separate the implementation detail. I don't care either way, but the other visitors should be changed first and this one can follow the same pattern. 
Yeah, let's leave this in place and deal with it later.


================
Comment at: include/clang/AST/AttrVisitor.h:23-24
+
+template <typename T> struct make_ptr { using type = T *; };
+template <typename T> struct make_const_ptr { using type = const T *; };
+
----------------
steveire wrote:
> aaron.ballman wrote:
> > Oh, this is cute. We have StmtVisitor which does not use a namespace, but defines these functions. Then we duplicate these function definitions in each of the comment, decl, and attribute visitors under distinct namespaces so we don't get ODR violations.
> > 
> > I think we should add `llvm::make_const_ptr` to STLExtras.h so we can use the same definition from everywhere. We can use `std::add_pointer` in place of `make_ptr`, because that is a one-to-one mapping, so we can drop `make_ptr` entirely. However, I don't think we can use `std::add_pointer<std::add_const>` similarly because there's no way to bind that to the template template parameter directly. I'm happy to make these changes if you'd like.
> Yes, I've thought about this duplication too. If you deduplicate, I'll rebase this change.  
Sounds good, I'll try to get that in shortly.


================
Comment at: include/clang/AST/AttrVisitor.h:27
+/// A simple visitor class that helps create attribute visitors.
+template <template <typename> class Ptr, typename ImplClass,
+          typename RetTy = void, class... ParamTys>
----------------
steveire wrote:
> aaron.ballman wrote:
> > s/class/typename
> I think c++17 is the first to allow typename here. 
Huh, how neat. I'm fine leaving it as `class` if there are concerns about this being accepted by all our base compilers, it was a minor nit.


Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D55492/new/

https://reviews.llvm.org/D55492