https://github.com/labath commented:
Okay, I think I understand now. The question is about what to do with pointer to data structures with child providers (e.g. `std::vector<T>*`). The original "frame var" indexed the pointer (meaning `v[0]` was the same as dereferencing the vector -- handy because "frame var" didn't allow you to insert a `(*...)` at a random point in the expression). The DIL implementation interpreted it as "returning the first element of the pointed-to vector" (essentially, transparently dereferencing the object). After your first version of the patch changed that to match "frame var", it broke the ObjC test because ObjC actually expected the transparent dereferencing behavior (which sort of makes sense as all ObjC objects are pointers). This version fixes that by taking over the ObjC special-casing code from the original implementation. This kind of special casing is something we wanted to avoid, but I think it's acceptable for now as it is just taking over existing code. We probably ought to fix it at some point though. https://github.com/llvm/llvm-project/pull/154269 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
