[es6] do not add caller/arguments to ES6 function definitions

src/contexts.h

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef V8_CONTEXTS_H_
 #define V8_CONTEXTS_H_
 
 #include "src/heap/heap.h"
 #include "src/objects.h"
 
@@ skipping 564 matching lines @@
   }
 
   static int FunctionMapIndex(LanguageMode language_mode, FunctionKind kind) {
     if (IsGeneratorFunction(kind)) {
       return is_strong(language_mode) ? STRONG_GENERATOR_FUNCTION_MAP_INDEX :
              is_strict(language_mode) ? STRICT_GENERATOR_FUNCTION_MAP_INDEX
                                       : SLOPPY_GENERATOR_FUNCTION_MAP_INDEX;
     }
 
     if (IsConstructor(kind)) {
-      return is_strong(language_mode) ? STRONG_CONSTRUCTOR_MAP_INDEX :
-             is_strict(language_mode) ? STRICT_FUNCTION_MAP_INDEX
-                                      : SLOPPY_FUNCTION_MAP_INDEX;
+      // Use strict function map (no own "caller" / "arguments")
+      return is_strong(language_mode) ? STRONG_CONSTRUCTOR_MAP_INDEX
+                                      : STRICT_FUNCTION_MAP_INDEX;
     }
 
     if (IsArrowFunction(kind) || IsConciseMethod(kind) ||
         IsAccessorFunction(kind)) {
-      return is_strong(language_mode) ? STRONG_FUNCTION_MAP_INDEX :
-             is_strict(language_mode) ?
-                 STRICT_FUNCTION_WITHOUT_PROTOTYPE_MAP_INDEX :
-                 SLOPPY_FUNCTION_WITHOUT_PROTOTYPE_MAP_INDEX;
+      return is_strong(language_mode)
+                 ? STRONG_FUNCTION_MAP_INDEX
+                 : is_strict(language_mode) || !IsAccessorFunction(kind)

[arv (Not doing code reviews), 2015/04/07 16:18:39]

I don't understand why accessor functions are special cased here?

Maybe restructure this code to make it clearer what is going on.

[caitp (gmail), 2015/04/07 16:29:57]

It was to avoid breaking the old behaviour.

But I'm not sure, it looks like lexically declared accessors are always
MethodDefinitions now, so maybe that's a break from ES5? I'm not really sure if
they should be considered "new syntactic forms" or what

[arv (Not doing code reviews), 2015/04/07 17:00:11]

Accessors are considered Methods in ES6 so they have no prototype property.

I just don't understand why we would use a sloppy map for a strict mode accessor
function?

My suggestion was to rewrite this ternary into if statementest to make the logic
easier to follow.

[caitp (gmail), 2015/04/07 17:10:36]

Oh I see what you're saying. So I guess I was trying to keep compat, and wound
up breaking it, my bad.

But my question is, should this be treated as a new syntactic form (like other
methods), or should the legacy behaviour be kept (where caller/arguments are
added in sloppy mode)
I try to follow the conventions used in the rest of the code, but I have no
problem with rewriting it as an if statement for clarity.

[arv (Not doing code reviews), 2015/04/07 18:22:03]

I think sloppy accessors should be treated as sloppy FunctionExpressions but it
is not clear to me if this is specified.

Maybe file a bug on ES6 and hope that Andre Bargul answers you :-)

+                       ? STRICT_FUNCTION_WITHOUT_PROTOTYPE_MAP_INDEX
+                       : SLOPPY_FUNCTION_WITHOUT_PROTOTYPE_MAP_INDEX;
     }
 
     return is_strong(language_mode) ? STRONG_FUNCTION_MAP_INDEX :
            is_strict(language_mode) ? STRICT_FUNCTION_MAP_INDEX
                                     : SLOPPY_FUNCTION_MAP_INDEX;
   }
 
   static const int kSize = kHeaderSize + NATIVE_CONTEXT_SLOTS * kPointerSize;
 
   // GC support.
@@ skipping 15 matching lines @@
   static bool IsBootstrappingOrGlobalObject(Isolate* isolate, Object* object);
 #endif
 
   STATIC_ASSERT(kHeaderSize == Internals::kContextHeaderSize);
   STATIC_ASSERT(EMBEDDER_DATA_INDEX == Internals::kContextEmbedderDataIndex);
 };
 
 } }  // namespace v8::internal
 
 #endif  // V8_CONTEXTS_H_