nridge added inline comments.
================
Comment at: clang-tools-extra/clangd/XRefs.cpp:1150
+ if (llvm::isa<UsingDecl>(D) || llvm::isa<UnresolvedUsingValueDecl>(D)) {
+ Decls = getDeclAtPosition(AST, *CurLoc,
+ Relations | DeclRelation::Underlying);
----------------
hokein wrote:
> I think it should not happen in practice (Decls just have 1 element in most
> cases), but the code feels hacky, we are throwing other decls if one of the
> `Decls` is a using decl.
>
> I suppose if we're using the same workaround as `locateASTReferent`, then we
> should follow that way by only adjusting the UsingDecl/UnresolvedValueDecl
> results and keeping others.
>
>
> In general, I think we probably need to remove this workaround (see the
> `FIXME` in `locateASTReferent`) by refining TargetDecl API. The current
> `DeclRelation::Underlying` enum is not enough to support our use case where
> we only want underlying decls for *non-renaming* alias. One rough idea to fix
> it is to split the `Underlying` to two `RenameAliasUnderlying` and
> `RemainingUnderlying` -- this would need some API design work, so no need to
> do it in this patch.
>
>
>
I don't think we're actually throwing out the other results: by calling
`getDeclAtPosition()` with `Relations | Underlying`, where `Relations` is the
original flags, the call should find the other results again. If we only
replaced the UsingDecl/UnresolvedValueDecl results, I think the other results
would appear in duplicate.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D87225/new/
https://reviews.llvm.org/D87225
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
