Replaced deprecated v8 API calls by their non-deprecated counterparts.

net/proxy/proxy_resolver_v8.h

Index: net/proxy/proxy_resolver_v8.h
diff --git a/net/proxy/proxy_resolver_v8.h b/net/proxy/proxy_resolver_v8.h
index 3adc672a1ffebdb9db891e255f06fd25c704406a..3c3dadc0fc7528bd04538df14a40fde2e4ef2b85 100644
--- a/net/proxy/proxy_resolver_v8.h
+++ b/net/proxy/proxy_resolver_v8.h
@@ -11,6 +11,7 @@
 #include "net/proxy/proxy_resolver.h"
 
 namespace v8 {
+class HeapStatistics;
 class Isolate;
 }  // namespace v8
 
@@ -95,6 +96,10 @@ class NET_EXPORT_PRIVATE ProxyResolverV8 : public ProxyResolver {
   static void RememberDefaultIsolate();
   static v8::Isolate* GetDefaultIsolate();
 
+  // Get statistics for heap memory usage of all v8 instances used by the proxy
+  // resolver.
+  static void GetHeapStatistics(v8::HeapStatistics* heap_statistics);

[eroman, 2013/03/05 00:40:45]

Rather than filling a HeapStatistics, can you make it return a size_t for heap
size instead? (It looks like this is the only number the task manager actually
cares about).

[Sven Panne, 2013/03/05 07:22:39]

I would very much like to keep the API as it is unless there is a very strong
reason against it, because from a design point of view for the simple case at
hand (1 Isolate) there should be a 1:1 correspondence between the v8 API and the
API entry in net. What should be abstracted away is the fact that the proxy
resolver might use *any* number of Isolates internally, not anything else.
Perhaps the task manager might want to report more details in the future or we
might have other clients. We already have a good data type for heap statistics
(HeapStatistics :-), so I don't see why the net subsystem should be any
different and inventing yet another signature seems like useless creativity.

Another possibility would be that we add an API entry to return a collection of
Isolates used by the proxy resolver, but this seems to be too general and leads
to questions about locking. Hmmm, thinking about the latter: I think we should
use a v8::Locker around the call to v8::Isolate::GetHeapStatistics in the
implementation, unless it is somehow guaranteed that the calling thread is
already holding a lock, but I fail to see why this should be the case.

+
  private:
   static v8::Isolate* g_default_isolate_;