Fznamznon added a comment. In D151553#4380667 <https://reviews.llvm.org/D151553#4380667>, @erichkeane wrote:
> Like @cor3ntin I'm concerned about removing the transform call. I'm just as > concerned that it caused no regressions... > > If we have to transform the arguments, then this can be a dependent thing, > which means the callee should be possible to be dependent, right? Thus needs > to be transformed somewhere? Oh, I didn't get @cor3ntin 's question in the first place! The question was where it the callee instantiated, right? The callee is instantiated when the whole operator call expr is rebuilt in Sema::RebuildCXXOperatorCallExpr. From what I see, `Sema::RebuildCXXOperatorCallExpr` doesn't really care which callee is passed there. It checks which operation it represents and performs lookup depending on the arguments. This triggers instantiation. Please see the test I'm adding. There is several cases where the operator callee is dependent. They work fine with the change. I also see the instantiated function decl in the AST, just the same as before the change. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D151553/new/ https://reviews.llvm.org/D151553 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
