cor3ntin wrote: > I'm not sure I understand, because MLTAL doesn't seem to own the arguments in > the first place, so copying it doesn't seem of any help.
Well, yes, we would have to create a copy of the argument list corresponding to the depth of the pack we are trying to deduce. There may be a better solution to implement https://eel.is/c++draft/temp.arg.explicit#9. Maybe @mizvekov would have an idea. But I suspect any such solution would be an even bigger refactor. So, unless we have a clever idea, I would like to see explored a solution that clones `MultiLevelTemplateArgumentList` instead of mutating it in place. (`TemplateInstantiator` would own a stack of `SmallVector<TemplateArgument>`) We could also add in `MultiLevelTemplateArgumentList::ArgList` an `optional<unsigned>` that tracks if a given template argument should be treated to be null, such that cloning a `MultiLevelTemplateArgumentList` just sets that optional. This would limit the changes to `TemplateInstantiator` https://github.com/llvm/llvm-project/pull/104687 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
