vbvictor wrote: > This PR breaks that case, true. But I hope we can agree that the current > behaviour is still not ideal, right? We would like the check to work whether > you instantiate the template or not. That would require another PR > implementing heuristic name resolution (and I would be happy to write it!).
I think it's not perfectly ideal, but good enough for 99.9..% cases. I think we have dozens of other checks that operate with same logic involving templates. (To find exact number need grep `instantiate()` in test dir). So if in this case we get a lot of perf benefit - worth doing `IgnoreUnless..`, otherwise I personally wouldn't bother much about it. > I think it would be less work in the end to leave this PR as-is, go implement > the heuristic. This sounds good to me. ----------------------------- I don't like merging with just `IgnoreUnless` because it may brong regression to library template code that is usually instantiated to concrete types in the end e.g. in tests. https://github.com/llvm/llvm-project/pull/175121 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
