Using the new MemoryPressureListener instead of the LowMemoryObserver when the enhanced memory mana…

content/public/browser/browser_thread.h

 // Copyright (c) 2012 The Chromium 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 CONTENT_PUBLIC_BROWSER_BROWSER_THREAD_H_
 #define CONTENT_PUBLIC_BROWSER_BROWSER_THREAD_H_
 
 #include <string>
 
 #include "base/basictypes.h"
 #include "base/callback.h"
 #include "base/location.h"
 #include "base/logging.h"
 #include "base/message_loop/message_loop_proxy.h"
 #include "base/task_runner_util.h"
 #include "base/time/time.h"
 #include "content/common/content_export.h"
 
 namespace base {
 class MessageLoop;
 class SequencedWorkerPool;
 class Thread;
+#if defined(OS_CHROMEOS)
+class MemoryPressureObserverChromeOS;
+#endif
 }
 
 namespace content {
 
 class BrowserThreadDelegate;
 class BrowserThreadImpl;
 
 // Use DCHECK_CURRENTLY_ON(BrowserThread::ID) to assert that a function can only
 // be called on the named BrowserThread.
 #define DCHECK_CURRENTLY_ON(thread_identifier)                      \
@@ skipping 112 matching lines @@
   }
 
   template <class T>
   static bool ReleaseSoon(ID identifier,
                           const tracked_objects::Location& from_here,
                           const T* object) {
     return GetMessageLoopProxyForThread(identifier)->ReleaseSoon(
         from_here, object);
   }
 
+#if defined(OS_CHROMEOS)
+  static base::MemoryPressureObserverChromeOS* GetMemoryPressureObserver();
+#endif

[piman, 2014/12/20 02:53:15]

This doesn't really look like it belongs to BrowserThread.

Should it be a top-level free function?


Also, it would be nice if this was not a ChromeOS-only interface.

[Mr4D (OOO till 08-26), 2014/12/20 03:01:05]

Sure - we could also do it that way. Do you prefer top level free?

After long consideration I thought that BrowserThread was a good fit.

For the interface:
That was what I initially intended. That was the reason for me to put it into a
public (base) place. But since this could not be done for Android I was told to
only do this for ChromeOS. Should we reconsider and do it for all OS'es?

[piman, 2014/12/20 03:05:17]

I think I would prefer top-level free.
This doesn't seem to fit with other members of this class.
Also, it's not looking up a member of BrowserThreadImpl. So it really feels
unrelated.

See e.g. content::GetContextFactory
I would prefer if we didn't have #ifdef warts in public APIs. In the future it's
not inconceivable that we can support the memory pressure logic, or at least a
subset, on other platforms.

[Mr4D (OOO till 08-26), 2015/01/06 22:22:06]

Done that way.

However - I still believe that this object should be a global base:: singleton
accessible from everywhere - this object is a rather obvious thing which could
be used from everywhere (even from base).

Also - yes, I was trying to make this object OS unaware, but that change  was
rejected. If we ever want to make this uniform for all OS'es I am all for it.

+
   // Simplified wrappers for posting to the blocking thread pool. Use this
   // for doing things like blocking I/O.
   //
   // The first variant will run the task in the pool with no sequencing
   // semantics, so may get run in parallel with other posted tasks. The second
   // variant will all post a task with no sequencing semantics, and will post a
   // reply task to the origin TaskRunner upon completion.  The third variant
   // provides sequencing between tasks with the same sequence token name.
   //
   // These tasks are guaranteed to run before shutdown.
@@ skipping 122 matching lines @@
  private:
   friend class BrowserThreadImpl;
 
   BrowserThread() {}
   DISALLOW_COPY_AND_ASSIGN(BrowserThread);
 };
 
 }  // namespace content
 
 #endif  // CONTENT_PUBLIC_BROWSER_BROWSER_THREAD_H_