Type profiler by intercepting 'new' and 'delete' expressions.

build/common.gypi

Index: build/common.gypi
diff --git a/build/common.gypi b/build/common.gypi
index 1b73dcd310bea7fbf3d54405554f158c319ed162..db40045d85b0f58fc73288d920b1bd9d0137754a 100644
--- a/build/common.gypi
+++ b/build/common.gypi
@@ -238,6 +238,13 @@
       # See http://clang.llvm.org/docs/ThreadSanitizer.html
       'tsan%': 0,
 
+      # Use a modified version of Clang to intercept allocated types and sizes
+      # for allocated objects. -fintercept-allocated-type only works with the
+      # modified clang, but clang_use_allocated_type=1 implies clang=1
+      # TODO(dmikurube): Add a link to a description page.
+      # TODO(dmikurube): Support mac.

[Alexander Potapenko, 2012/08/02 14:57:12]

Note that tcmalloc does not work on Mac within Chromium.

[Dai Mikurube (NOT FULLTIME), 2012/08/02 16:14:35]

Ah, that's true.  Thanks.
I have to find another allocated-type-map storage in Mac...

+      'clang_use_allocated_type%': 0,
+
       # Set to true to instrument the code with function call logger.
       # See src/third_party/cygprofile/cyg-profile.cc for details.
       'order_profiling%': 0,
@@ -1181,6 +1188,12 @@
         'clang%': 1,
       }],
 
+      ['clang_use_allocated_type==1', {
+        'clang%': 1,
+        'clang_use_chrome_plugins%': 0,
+        'make_clang_dir%': 'third_party/llvm-allocated-type/Linux_x64',
+      }],
+
       # On valgrind bots, override the optimizer settings so we don't inline too
       # much and make the stacks harder to figure out.
       #
@@ -1586,6 +1599,28 @@
       }],
     ],  # conditions for 'target_defaults'
     'target_conditions': [
+      ['_type=="executable" or _type=="loadable_module"', {
+        'target_conditions': [
+          ['_target_name=="mksnapshot" or _target_name=="protoc" or _target_name=="ppGoogleNaClPluginChrome"', {

[Dai Mikurube (NOT FULLTIME), 2012/08/02 07:43:55]

Some binaries doesn't use tcmalloc.  We use no-op __op_{new|delete|_intercept__
for these binaries since the allocated type mapping is recorded in a
tcmalloc-dependent data structure.

[Alexander Potapenko, 2012/08/02 14:57:12]

Please write a comment about this.

[Dai Mikurube (NOT FULLTIME), 2012/08/02 16:14:35]

Yes, I'll do that.  Thank you.

[Dai Mikurube (NOT FULLTIME), 2012/08/03 10:01:50]

Done.

+            # Ignore allocated types.
+            'conditions': [
+              ['clang==1 and clang_use_allocated_type==1', {
+                'dependencies': [
+                  '<!(realpath <(DEPTH)/base/allocator/allocator.gyp)' + ':allocated_type_ignore#target',
+                ]},
+              ],
+            ]}, {
+            # Store allocated types in TCMalloc's allocated_type_map (extended).
+            'conditions': [
+              ['clang==1 and clang_use_allocated_type==1', {
+                'dependencies': [
+                  '<!(realpath <(DEPTH)/base/allocator/allocator.gyp)' + ':allocated_type_tcmalloc#target',

[Nico, 2012/08/02 15:35:03]

The realpath bit here shouldn't be needed

[Dai Mikurube (NOT FULLTIME), 2012/08/02 16:14:35]

It is actually needed because of http://crrev.com/146172.
(I was surprised when it happened.)

[Nico, 2012/08/02 16:18:49]

No other dependency in gyp uses realpath though, right? Why is this special?

It feels like we should open a bug for this being necessary and probably revert
146172 if this is the consequence :-/

[Dai Mikurube (NOT FULLTIME), 2012/08/02 16:54:39]

I guess it happens because it is a very hacky gyp dependency description.  This
'dependencies' is lazily evaluated after most paths are translated using
'script_dir' and other parameters.

To describe it completely, I have to reread gyp.  I remember that it is an
effect of 'target_conditions'.  Such a hacky dependency is required since we
need to add dependency depending on its target.  (gyp is complicated...)

+                ]},
+              ],
+            ]},
+          ]
+        ]},
+      ],
       ['enable_wexit_time_destructors==1', {
         'conditions': [
           [ 'clang==1', {
@@ -2329,6 +2364,22 @@
               '-fcolor-diagnostics',
             ],
           }],
+          ['clang==1 and clang_use_allocated_type==1', {

[Alexander Potapenko, 2012/08/03 09:57:07]

If clang_use_allocated_type=1 implies clang=1, you don't need to check both.

[Dai Mikurube (NOT FULLTIME), 2012/08/07 10:39:18]

Done.

+            'cflags_cc!': ['-fno-rtti'],
+            'cflags_cc+': ['-gline-tables-only'],
+            'target_conditions': [
+                ['_target_name!="allocator" and _target_name!="allocator_extension_thunks" and _target_name!="libcmt" and _target_name!="allocator_extension_thunks_win64" and _target_name!="allocated_type_log" and _target_name!="allocated_type_ignore" and _target_name!="allocated_type_tcmalloc"', {
+                    'cflags_cc+': [
+                        '-frtti',
+                        '-fintercept-allocation-functions',
+                        '-I../../base/allocator',
+                        '-include',
+                        'allocated_type.h',
+                    ],
+                }],
+            ],
+            'defines': ['USE_ALLOCATED_TYPE'],
+          }],
           ['asan==1', {
             'target_conditions': [
               ['_toolset=="target"', {