cor3ntin wrote: > I _think_ the new map should be unnecessary: `FunctionTemplateDecl` and > `VarTemplateDecl` already have a list of specializations that you can walk > and check the point of instantiation and template specialization kind to see > if there's a pending instantiation.
I don't think that's viable for non-template but templated entities like member function in a class template. Am I missing something? > If you want/need to keep the map approach, you'll need to also deal with the > case where the template is declared in a PCH and then defined somewhere that > includes the PCH, which will mean serializing and deserializing the map > contents. I did not consider that :( https://github.com/llvm/llvm-project/pull/73463 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits