Update third_party/protobuf to protobuf-v3.0.0-beta-3

third_party/protobuf/README.chromium

Index: third_party/protobuf/README.chromium
diff --git a/third_party/protobuf/README.chromium b/third_party/protobuf/README.chromium
index 0af6f79293d752e9d1bba1ce4224e80313803892..787afed7d0c7d79ee67ba09aa8afe3dedb4ca801 100644
--- a/third_party/protobuf/README.chromium
+++ b/third_party/protobuf/README.chromium
@@ -3,14 +3,13 @@ Short Name: protobuf
 URL: https://github.com/google/protobuf
 License: BSD
 License File: LICENSE
-Version: 3.0.0-beta-2
-Revision: d5fb408ddc281ffcadeb08699e65bb694656d0bd
+Version: 3.0.0-beta-3
+Revision: 3470b6895aa659b7559ed678e029a5338e535f14
 Security Critical: yes
 
 Steps used to create the current version:
 1. Pull the release from https://github.com/google/protobuf/releases
-2. Use the cleanup_release.sh script to remove unnecessary files.
-3. Add build files (BUILD.gn, proto_library.gni, protobuf.gyp,
+2. Add build files (BUILD.gn, proto_library.gni, protobuf.gyp,
    protobuf_lite.gypi, protobuf_nacl.gyp).
 
    Be sure to update the list of source files, as additional .cc files and
@@ -19,9 +18,9 @@ Steps used to create the current version:
 
    Other things to care about are defines required by protobuf on various
    platforms, warnings generated by compilers, and new dependencies introduced.
-4. Get open-source library six.py from https://pypi.python.org/pypi/six/ and add
+3. Get open-source library six.py from https://pypi.python.org/pypi/six/ and add
    it to protobuf/third_party/six/six.py.
-5. Apply patches in patches/ (see the description below):
+4. Apply patches in patches/ (see the description below):
 
    $ for patch in patches/*; do patch -s -p1 < $patch; done
 
@@ -53,6 +52,12 @@ Description of the patches:
   I plan to remove those occurences, and then this patch will no longer be
   necessary.
 
+- 0002-revert-patch-internal-change-111557819.patch
+
+  This reverts the commit 76195058e25d19fc918996d55d3ad69ee55cb77e, which bloats
+  the Chrome binary by 200 KB. Its purpose was to reduce the memory usage, but
+  Chrome workloads don't seem to be benefitting from this.

[Peter Kasting, 2016/05/24 23:45:34]

Did you determine the actual memory savings from this?

[xyzzyz, 2016/05/25 00:15:28]

I tried, but the benchmarking results are really fuzzy.

http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-05...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-05...

One is with that patch kept, the other with reverted. Try guessing which is
which (be sure to also click on "memory" at the top to show all cases). Note
that ToT itself has bigger variance.

[Peter Kasting, 2016/05/25 00:19:50]

To me the second link (05-20) looks fairly consistently better than the first
link (05-19).

Which is which?

[xyzzyz, 2016/05/25 00:25:47]

The 05-20 is the optimize-memory-with-larger-binary-size one. Why would you say
it's better, though? 

[Peter Kasting, 2016/05/25 01:17:56]

Every relevant browser memory stat is improved compared to the other run:
* Commit charge is 5 MB lower
* Peak RSS is 1.25 MB lower
* Final VM size is 8 MB lower
* Peak VM size is 6 MB lower
* Dirty private VM size is 1.5 MB lower
* Final VM RSS size is 1.5 MB lower

Most of these also show improvements from one run to the other if viewed as
"delta from TOT" instead of as absolute values.  (It's a little confusing to
read the chart because the "delta" column shows how much better or worse TOT is
compared to the patch instead of the reverse, so if that column gets "worse",
that actually means the patch did better at improving memory use.)  The only one
that doesn't is the overall commit charge, and I'm suspicious of the TOT number
there because it dropped by an incredible amount compared to the 5-19 run even
though the other numbers didn't really move at all.

So to me this seems like a clear win for the "decrease memory use" patch and I'd
leave it in the codebase.  I think it's worth the 200K binary size increase.

+
 - 0003-remove-static-initializers.patch
 
   This patch removes all static initializers from Chromium. The change in Status
@@ -68,17 +73,12 @@ Description of the patches:
   protobuf symbols in Linux .so libraries, so that protobuf can be built as a
   component (see http://crrev.com/179806).
 
-- 0005-fix-no-threadlocal-on-mac.patch
-
-  Apple's Clang doesn't support thread-local variables on Mac.
+- 0005-fix-include-js-generator.protobuf
 
-- 0006-fix-clang-type-errors.patch
+  During merge with internal branch, the <> in one of the #includes were
+  accidentally replaced with "", which results in a failure in checkdeps.
 
-  New versions of Clang have type checking semantics that check the conformance
-  with the C++ standard more thorougly, which makes them a bit more
-  restrictive. This cherry-picks upstream commits
-  3cc35adb6dca9057a790d253b9f5e2b9b000a106 and
-  58f07644eab25f68973ecab4bdc43fc318c26131.
+  Fixed in https://github.com/google/protobuf/pull/1547.
 
 - 0007-uninline_googleonce.patch
 - 0008-uninline_get_empty_string.patch
@@ -87,12 +87,11 @@ Description of the patches:
 
   These patches uninline some functions, resulting in a significant reduction
   (somewhere between 500 KB and 1 MB) of binary size.
-- 0011-use-offsetof-for-clang.patch
 
-  This patch replaces a few invalid reinterpret_casts with __builtin_offsetof,
-  if compiled with Clang. This is a temporary patch, it should be deletd, when
-  the official github repo is updates with this changes already submitted to
-  the internal Google repo.
+- 0011-libprotobuf_export.patch
+
+  During merge with internal branch, the dll export attribute was accidentally
+  removed in C++11 mode.
+
+  Fixed in https://github.com/google/protobuf/pull/1549
 
-  See also: https://crbug.com/607751 and
-  https://github.com/google/protobuf/issues/1450