gribozavr added a comment. In D65877#1625493 <https://reviews.llvm.org/D65877#1625493>, @ymandel wrote:
> I was going to add a test for `Type`/`QualType` and realized that they don't > carry any source location info. Therefore, I don't think they belong as > top-level matchers for rewrite rules. Agreed. > Instead, users should use `typeLoc(loc(<type matcher here>)`. Therefore, the > logic of `getKind` can be eliminated in favor of using `K.getSupportedKind()` > directly. However, that means we'll need to keep them out of the collection > of matchers that we're processing. > > For that, I propose either we ban them in `buildMatchers()` or we detect them > and wrap them in `typeLoc(loc(<type matcher here>)`. I'm fine with either > one, but prefer the latter. WDYT? I'd prefer we ban them completely, and let the user wrap them manually if they need to. While some users would appreciate the ergonomics, I think AST matchers are complicated even without us adding more implicit behavior on top. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D65877/new/ https://reviews.llvm.org/D65877 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
