Build third-party code with -Wall, build chromium_code with -Wextra on Linux

base/allocator/allocator.gyp

Index: base/allocator/allocator.gyp
diff --git a/base/allocator/allocator.gyp b/base/allocator/allocator.gyp
index 33508ab43fabef3a8a26106c58bafa61af13066a..45a95bbdc2669a0dafdd9fa9caaf0f050d7b921b 100644
--- a/base/allocator/allocator.gyp
+++ b/base/allocator/allocator.gyp
@@ -284,6 +284,20 @@
             '<(tcmalloc_dir)/src/debugallocation.cc',
             '<(tcmalloc_dir)/src/tcmalloc.cc',
           ],
+          'variables': {
+            'clang_warning_flags': [
+              # tcmalloc initializes some fields in the wrong order.
+              '-Wno-reorder',
+              # tcmalloc contains some unused local template specializations.
+              '-Wno-unused-function',
+              # tcmalloc uses COMPILE_ASSERT without static_assert but with
+              # typedefs.

[dcheng, 2015/12/30 23:30:19]

Isn't our copy of tcmalloc a fork? Can we fix these, or is it not worth it?

[Nico, 2015/12/30 23:53:51]

We could, but I try to not make the diff larger than necessary.

[Lei Zhang, 2015/12/31 00:07:37]

It looks like https://github.com/gperftools/gperftools/tree/master/src has
recent-ish commits. Perhaps file a bug to bring in a more recent tcmalloc and
eventually get rid of these?

[Nico, 2015/12/31 02:44:17]

From what I understand, that tcmalloc is completely separate from
google-internal tcmalloc. I'm not sure if we want  to depend on that?

+              '-Wno-unused-local-typedefs',
+              # for magic2_ in debugallocation.cc (only built in Debug builds)
+              # typedefs.
+              '-Wno-unused-private-field',
+            ],
+          },
           'conditions': [
             ['OS=="linux" or OS=="freebsd" or OS=="solaris" or OS=="android"', {
               'sources!': [