hlopko added a comment.
Let's continue the review at https://reviews.llvm.org/D76346.
================
Comment at: clang/lib/Tooling/Syntax/BuildTree.cpp:190
+ // Set role for the node that may or may not be delayed. Node must span
+ // exactly \p Range.
+ void markMaybeDelayedChild(llvm::ArrayRef<syntax::Token> Range, NodeRole R);
----------------
gribozavr2 wrote:
> Three slashes for docs.
Done.
================
Comment at: clang/lib/Tooling/Syntax/BuildTree.cpp:930
+ Builder.markChildToken(TemplateKW, syntax::NodeRole::IntroducerKeyword);
+ Builder.markMaybeDelayedChild(
+ TemplatedDeclaration,
----------------
gribozavr2 wrote:
> Why is this range maybe-delayed?
E.g because `template <class T> struct cls {};` is doesn't need delaying, but
template <class T> int var = 10;` does (SimpleDeclaration is processed before
Declarator). I'll try to come up with a more fitting design after I submit this
and https://reviews.llvm.org/D72446.
================
Comment at: clang/unittests/Tooling/Syntax/TreeTest.cpp:738
+ U foo();
+};
+ )cpp",
----------------
gribozavr2 wrote:
> Could you also add an out-of-line definition of X::foo? It will have two
> template parameter lists, which is a special case.
Example:
template <class T>
struct X {
template <class U>
U foo();
};
**template <class T>**
template <class U>
U X<T>::foo() {}
Discussed offline, RAV doesn't visit TemplateParameters and therefore we're
not creating a node for the first template parameter (in bold). We could
overcome this in the BuildTree.cpp, but more principled solution would be to
teach RAV to visit TemplateParameters. In a separate patch.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D72334/new/
https://reviews.llvm.org/D72334
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
