[PATCH] D100742: [clangd] Parameter hints for dependent calls

[Nathan Ridge via Phabricator via cfe-commits]

nridge added a reviewer: sammccall.
nridge added inline comments.


================
Comment at: clang-tools-extra/clangd/unittests/InlayHintTests.cpp:264
+        // FIXME: This one does not work yet.
+        A<T>::static_member($par3[[t]]);
       }
----------------
This is an interesting case. Clang builds a `CXXDependentScopeMemberExpr` for this callee, but `HeuristicResolver` [currently assumes](https://searchfox.org/llvm/rev/92880ab7a2b2145f0605f367cd6d53d6892903c3/clang-tools-extra/clangd/HeuristicResolver.cpp#108) that such expressions are only built for non-static member accesses (since, for static member accesses, clang usually builds a `DependentScopeDeclRefExpr` instead).

The `CXXDependentScopeMemberExpr` is created [here](https://searchfox.org/llvm/rev/92880ab7a2b2145f0605f367cd6d53d6892903c3/clang/lib/Sema/SemaTemplate.cpp#757), and I note the dependence on whether the //enclosing context// is an [instance method](https://searchfox.org/llvm/rev/92880ab7a2b2145f0605f367cd6d53d6892903c3/clang/lib/Sema/SemaTemplate.cpp#750). I guess it must think that, after instantiation, `A<T>` could turn out to be a base class, and thus this could be a "non-static member access with qualifier".

I don't see anything obvious on `CXXDependentScopeMemberExpr` that would let us tell apart "definitely a non-static member" from "maybe static, maybe non-static", so I guess the appropriate solution is to drop the `NonStaticFilter` [here](https://searchfox.org/llvm/rev/92880ab7a2b2145f0605f367cd6d53d6892903c3/clang-tools-extra/clangd/HeuristicResolver.cpp#108) altogether?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D100742/new/

https://reviews.llvm.org/D100742