Only treat possible eval calls going through 'with' as special.

src/full-codegen/full-codegen.cc

Index: src/full-codegen/full-codegen.cc
diff --git a/src/full-codegen/full-codegen.cc b/src/full-codegen/full-codegen.cc
index 3af540b294c18491514fc7a4e6393aca6063e19d..43ba0b529f00cee2cbed41084f3d6274b409df37 100644
--- a/src/full-codegen/full-codegen.cc
+++ b/src/full-codegen/full-codegen.cc
@@ -1669,6 +1669,7 @@ void FullCodeGenerator::VisitCall(Call* expr) {
 
   switch (call_type) {
     case Call::POSSIBLY_EVAL_CALL:
+    case Call::POSSIBLY_EVAL_THROUGH_WITH_CALL:

[Toon Verwaest, 2016/11/09 14:13:08]

Shouldn't POSSIBLY_EVAL_CALL be moved to OTHER_CALL now? Here and the
ast-graph-builder?

[rmcilroy, 2016/11/09 17:02:29]

Nope because we also need to resolve the eval function for POSSIBLY_EVAL_CALL
which we don't do with OTHER_CALL (there is no possibly_eval branch below like
there is in ast-graph-builder / bytecode-graph-builder).

Given that we never actually go through FCG for functions with eval any longer
(since they will be optimized by Turbofan, so always go through Ignition after
the change which feeds all TF code via Ignition) I don't think it's worth doing
any work here - we will eventually rip support for eval out of FCG once we are
sure the TF via Ignition change is going to stick.

[Toon Verwaest, 2016/11/10 09:40:19]

I see. But it is still relevant in the ast-graph-builder though right?

       EmitPossiblyEvalCall(expr);
       break;
     case Call::GLOBAL_CALL: