[PATCH] D112707: [clangd] IncludeCleaner: Be more conservative in marking RecordDecl usage

[Sam McCall via Phabricator via cfe-commits]

sammccall added a comment.

What's the purpose of this patch at a high level? Was it triggered by a real example?
IIUC it's avoiding false negatives, where we're using a class and an otherwise-unused header forward-declares that class.
Avoiding false negatives isn't a high priority at this point, unless it's a *really* common case.
As Kadir says this is subtle and risks introducing false positives.

My inclination is that we shouldn't spend time making to make these heuristics more precise/complicated right now, but I'm willing to be convinced...


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D112707/new/

https://reviews.llvm.org/D112707