[PATCH] D45476: [C++2a] Implement operator<=> CodeGen and ExprConstant

[John McCall via Phabricator via cfe-commits]

rjmccall added inline comments.


================
Comment at: lib/CodeGen/CGExprAgg.cpp:964
+    RHS = CGF.EmitAnyExpr(E->getRHS()).getAggregatePointer();
+    break;
+  case TEK_Complex:
----------------
EricWF wrote:
> EricWF wrote:
> > rjmccall wrote:
> > > It looks like we don't actually support any aggregate types here, which I think is fine because comparing those types is only sensible for things like calls.  If you do want to pave the way for that, or (probably more usefully) for supporting complex types, you should make EmitCompare take the operands as RValues and just use EmitAnyExpr here without paying any attention to the evaluation kind.
> > Initially I thought the same thing, but apparently member pointers are Aggregates under the Microsoft ABI.
> > 
> > I'll give  trafficking in `RValue`s, but all the functions `EmitCompare` calls use `Value*`, so it'll take some work.
> *I'll give trafficking in `RValue`s a shot, but ...*
Okay, this would be a *lot* cleaner with RValue.  You can break it down in your EmitCmp helper function instead of EmitCompare if you want, but you've basically just inlined EmitAnyExpr here.


================
Comment at: lib/CodeGen/CGExprAgg.cpp:924
+  }();
+  ArgTy->isAnyComplexType();
+  if (ArgTy->hasFloatingRepresentation())
----------------
Dead code?


https://reviews.llvm.org/D45476