A deeper heap profile dumper in third_party/tcmalloc/chromium.

third_party/tcmalloc/chromium/src/heap-profile-table.h

 // Copyright (c) 2006, Google Inc.
 // All rights reserved.
 // 
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 // 
 //     * Redistributions of source code must retain the above copyright
 // notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
@@ skipping 164 matching lines @@
   // Release a previously taken snapshot.  snapshot must not
   // be used after this call.
   void ReleaseSnapshot(Snapshot* snapshot);
 
   // Return a snapshot of every non-live, non-ignored object in *this.
   // If "base" is non-NULL, skip any objects present in "base".
   // As a side-effect, clears the "live" bit on every live object in *this.
   // Caller must call ReleaseSnapshot() on result when no longer needed.
   Snapshot* NonLiveSnapshot(Snapshot* base);
 
   // Refresh the internal mmap information from MemoryRegionMap.  Results of

[jar (doing other things), 2012/01/26 20:19:02]

nit: "Refresh ..." --> "Refreshes ..."

[Dai Mikurube (NOT FULLTIME), 2012/01/30 12:54:53]

Thanks for your comments.  Hmm, it's sure that these functions are added by me,
but they're in the upstream tcmalloc/google-perftools.  The change is accepted
and hopefully landed in the next release:
http://code.google.com/p/gperftools/issues/detail?id=383.

I guess we should fix them in the upstream if required.

And, they follow other existing comments in heap-profile-table.h.  They use
verbs without "-s".

On 2012/01/26 20:19:02, jar wrote:

   // FillOrderedProfile and IterateOrderedAllocContexts will contain mmap'ed

[jar (doing other things), 2012/01/26 20:19:02]

nit: "...will contain ..." --> "Updates ... to contain ..."

[Dai Mikurube (NOT FULLTIME), 2012/01/30 12:54:53]

Ditto, but thanks for a good alternative.  I'll consider it in the upstream.

   // memory regions as at calling RefreshMMapData.

[jar (doing other things), 2012/01/26 20:19:02]

nit: typo?  "...as at calling ..."

[Dai Mikurube (NOT FULLTIME), 2012/01/30 12:54:53]

Ditto.

   void RefreshMMapData();
 
   // Clear the internal mmap information.  Results of FillOrderedProfile and

[jar (doing other things), 2012/01/26 20:19:02]

nit: "Clear ..." --> "Clears ..."

[Dai Mikurube (NOT FULLTIME), 2012/01/30 12:54:53]

Ditto.

   // IterateOrderedAllocContexts won't contain mmap'ed memory regions after
   // calling ClearMMapData.

[jar (doing other things), 2012/01/26 20:19:02]

nit suggestion: Change "Results of ... won't contain ..." to "Empties ..."

[Dai Mikurube (NOT FULLTIME), 2012/01/30 12:54:53]

Ditto.

   void ClearMMapData();
 
  private:
+  friend class DeepHeapProfile;
 
   // data types ----------------------------
 
   // Hash table bucket to hold (de)allocation stats
   // for a given allocation call stack trace.
   struct Bucket : public Stats {
     uintptr_t    hash;   // Hash value of the stack trace
     int          depth;  // Depth of stack trace
     const void** stack;  // Stack trace
     Bucket*      next;   // Next entry in hash-table
@@ skipping 202 matching lines @@
   // Helpers for sorting and generating leak reports
   struct Entry;
   struct ReportState;
   static void ReportCallback(const void* ptr, AllocValue* v, ReportState*);
   static void ReportObject(const void* ptr, AllocValue* v, char*);
 
   DISALLOW_COPY_AND_ASSIGN(Snapshot);
 };
 
 #endif  // BASE_HEAP_PROFILE_TABLE_H_