zyn0217 wrote: > > @cor3ntin Not much, we just need to move the handling of the friend > > declaration outside of `getTemplateInstantiationArgs()` :) > > > > > > Do you see any other outstanding issues? If not, can we go ahead with the > > status quo so we have sufficient timeframe in clang 20 cycle in case > > something else goes wrong? > > > > I don't think I'd be a fan of removing the friend declaration outside of > `getTemplateInstantiationArgs`. That is supposed to be the 'single source' > of truth for that, and should work on just about any declaration that could > potentially be templated. So I think we need to do the work to improve it > isntead.
Yeah, I admit that the current approach is suboptimal, as earlier iterations demonstrated that the patch could be cleanly built on top of the getTemplateInstantiationArgs() refactoring, without requiring any special-casing outside of that function. So I doubt if there’s much value in improving the old version of the function, especially since the refactoring work can be reapplied at any time. https://github.com/llvm/llvm-project/pull/102857 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
