> On Apr 4, 2025, at 12:40, Jakub Jelinek <ja...@redhat.com> wrote: > > On Fri, Apr 04, 2025 at 04:25:11PM +0000, Qing Zhao wrote: >>> It is not just the parser, it is also the human reader who should be >>> able to clearly distinguish this (which is why I still prefer >>> designators syntax because this makes it perfectly clear). >> >> Yes, the forward declaration approach is just a compromised approach in >> order to work with >> C++ in the future. > > I thought the forward decl approach is there just for C.
If it’s only for C, I think that the new keyword __self approach should be the best one. However, since Apple people insisted that the same attribute with the same syntax will be used in C++ as well (I think that this the root reason for this whole discussion). Then a syntax can be accepted by both C and C++ is needed here. This forward decl approach is a compromised approach for this purpose. Looks reasonable for both C and C++. (Do I miss anything here?) > C++ doesn't really need it, it has tons of other parts of the class parsed > only when the class is complete (e.g. in class method definition bodies, > NSDMIs, ...). So it wouldn't be really strange to let the attribute > argument be parsed for C++ only when the class is complete too. > And IMHO it should use the this->member syntax there. If different syntaxes of the counted_by attribute for C and C++ is accepted (which I prefer -:). Then for C, I prefer __self approach. Qing > > Jakub >
