Enable concurrent optimization test after migrating scopes to Thread*

runtime/vm/flow_graph_compiler.cc

Index: runtime/vm/flow_graph_compiler.cc
diff --git a/runtime/vm/flow_graph_compiler.cc b/runtime/vm/flow_graph_compiler.cc
index c7d3215bd9473da59ef811dad3fb5c816185140f..dbf03c0fe3c165bd7817a80826487ed75109a38c 100644
--- a/runtime/vm/flow_graph_compiler.cc
+++ b/runtime/vm/flow_graph_compiler.cc
@@ -175,7 +175,7 @@ FlowGraphCompiler::FlowGraphCompiler(
         block_info_(block_order_.length()),
         deopt_infos_(),
         static_calls_target_table_(GrowableObjectArray::ZoneHandle(
-            GrowableObjectArray::New())),
+            GrowableObjectArray::New(Heap::kOld))),

[srdjan, 2015/08/19 18:29:34]

This is a temporary fix. Please add: TODO(srdjan): allocate in zone instead of
old-space.

[koda, 2015/08/19 18:44:28]

OK, although it doesn't look temporary to me. It's converted to an Array (using
MakeArray) and seems to be unconditionally stored in the Code object in
FinalizeStaticCallTargetsTable. (I suppose if it's usually empty, it's wasteful
to always allocate, and we could leave the field 'null', provided the rest of
the code can handle that.)

[srdjan, 2015/08/19 19:46:50]

There is still 'garbage' created by the GrowableObjectArray descriptor and by
the Grow operation.

         is_optimizing_(is_optimizing),
         may_reoptimize_(false),
         intrinsic_mode_(false),
@@ -516,7 +516,7 @@ void FlowGraphCompiler::VisitBlocks() {
   }
 
   if (is_optimizing()) {
-    LogBlock lb(Isolate::Current());
+    LogBlock lb(Thread::Current());
     intervals.Add(IntervalStruct(prev_offset, prev_inlining_id));
     inlined_code_intervals_ =
         Array::New(intervals.length() * Code::kInlIntNumEntries, Heap::kOld);