Report utility process JS memory in task manager.

content/public/browser/child_process_data.h

Index: content/public/browser/child_process_data.h
diff --git a/content/public/browser/child_process_data.h b/content/public/browser/child_process_data.h
index 395c4876b8d736e4de46b11a71970bac95a2e818..c796895fb831a4d01dd44271a0d042f5f5ef6a1a 100644
--- a/content/public/browser/child_process_data.h
+++ b/content/public/browser/child_process_data.h
@@ -5,14 +5,16 @@
 #ifndef CONTENT_PUBLIC_BROWSER_CHILD_PROCESS_DATA_H_
 #define CONTENT_PUBLIC_BROWSER_CHILD_PROCESS_DATA_H_
 
+#include "base/memory/ref_counted.h"
 #include "base/process/process.h"
 #include "base/strings/string16.h"
 #include "content/common/content_export.h"
+#include "content/public/browser/process_resource_usage.h"
 
 namespace content {
 
 // Holds information about a child process.
-struct ChildProcessData {
+struct CONTENT_EXPORT ChildProcessData {
   // The type of the process.
   int process_type;
 
@@ -30,9 +32,12 @@ struct ChildProcessData {
   // current process.
   base::ProcessHandle handle;
 
-  explicit ChildProcessData(int process_type)
-      : process_type(process_type), id(0), handle(base::kNullProcessHandle) {
-  }
+  // Resource usage reporter to obtain certain resource stats such as JS memory
+  // usage. May be null.
+  scoped_refptr<ProcessResourceUsage> process_resource_usage;

[ncarter (slow), 2015/05/01 21:16:52]

As implemented, this modification does plumb the process_resource_usage
everywhere it needs to go, and effortlessly. But is it too much of a shortcut?
Does this member belong in this class?

This type, ChildProcessData, up until now has functioned as the signature/id of
the process: it's passed to every method of BrowserChildProcessObserver (process
creations and deletions alike); historically (ages ago) I think it was intended
for use as a map key. It's a struct, copyable, and those are usually reserved
for passing around plain data.

But here you're adding a scoped pointer to a stateful object that has a Refresh
method and whose values could change synchronously at any moment. That doesn't
feel like something that belongs in a copyable-identifier-struct-like to me.

Is there a way to implement this without adding a stateful object to
ChildProcessData? How much grosser is that?

[Anand Mistry (off Chromium), 2015/05/06 02:44:18]

Done. It requires more plumbing, but isn't as bad as I thought it would be. 

+
+  explicit ChildProcessData(int process_type);
+  ~ChildProcessData();
 };
 
 }  // namespace content