[PATCH] D140828: [C++] Implement "Deducing this" (P0847R7)

Mon Aug 28 01:30:06 PDT 2023

cor3ntin added inline comments.

================
Comment at: clang/lib/AST/DeclCXX.cpp:841
       const auto *ParamTy =
-          Method->getParamDecl(0)->getType()->getAs<ReferenceType>();
+          Method->getNonObjectParameter(0)->getType()->getAs<ReferenceType>();
       if (!ParamTy || ParamTy->getPointeeType().isConstQualified())
----------------
aaron.ballman wrote:
> cor3ntin wrote:
> > cor3ntin wrote:
> > > aaron.ballman wrote:
> > > > Under what circumstances should existing calls to `getParamDecl()` be converted to using `getNonObjectParameter()` instead? Similar for `getNumParams()`.
> > > everytime you want to ignore (or handle differently) the explicit object parameter. I'll do another survey at some point, to be sure i didn't miss a spot
> > I found a bug https://github.com/cplusplus/CWG/issues/390 ! (`AreSpecialMemberFunctionsSameKind`) - So far nothing else but I'll do several passes
> This adds an unfortunate amount of complexity to the compiler because now we have to explicitly remember to think about this weird scenario, but I'm not seeing much of a way around it. I suspect this will be a bit of a bug factory until we're used to thinking explicitly about this.
> 
> Be sure to double-check things like attributes that take arguments which represent an index into a function parameter list (like the `format` attribute), as those have to do a special dance to handle the implicit `this` parameter. I'm guessing the static analyzer and clang-tidy will both also run into this in some places as well.
https://cplusplus.github.io/CWG/issues/2787.html 
Do we want to implement that resolution now?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D140828/new/

https://reviews.llvm.org/D140828