Different fields from different messages can map to the same JSON name
and the original global lookup table is only capable of mapping one of
such fields. This change converts the global table to per-type tables
so fields from different messages won't conflict.

update_failure_list.py: fixed Python "raise" statement.

Ruby: wrap calls to memcpy so that gem is compatible with pre-2.14 glibc

Avoid redundant type casts for oneof bytes fields.

Undef major/minor if they are defined as macro.

Add annotations for deprecated messages in Java

This commit adds a __wrap_memcpy function and a linker flag to use that
in place of memcpy for our Ruby gem C extension. This allows us to
always use the 2.2.5 version of memcpy, making it possible to use the
gem on distributions with pre-2.14 versions of glibc.

Before this change:
$ objdump -T protobuf_c.so | grep memcpy
0000000000000000      DF *UND*  0000000000000000  GLIBC_2.3.4 __memcpy_chk
0000000000000000      DF *UND*  0000000000000000  GLIBC_2.14  memcpy

After:
$ objdump -T protobuf_c.so | grep memcpy
0000000000000000      DF *UND*  0000000000000000  GLIBC_2.2.5 memcpy
0000000000000000      DF *UND*  0000000000000000  GLIBC_2.3.4 __memcpy_chk
0000000000042450 g    DF .text  0000000000000005  Base __wrap_memcpy

This is based on gRPC's solution to a similar problem:
https://github.com/grpc/grpc/blob/5098508d2d41a116113f7e333c516cd9ef34a943/src/core/lib/support/wrap_memcpy.c

This fixes issue #2783.

Method Builder#clone() has been changed to bypass Java 1.6 compiler issue.

This better shows the semantic of the API. For already setted fields,
mergeFromString do replacement for singular fields and appending for
repeated fields.

Removed a stray return statement, causing compilation error.

 optimization and quick workaround to memory leak

Merge pull request #2788 from anandolee/master

Added a header only cc_library target for the protobuf library.

Return uint32 from Log2FloorNonZero64

Add Ruby compatibilty test against 3.0.0.

Handing threading race resolving methods.

- Don't prune the extension registry as that can lead to failures when two
  threads are racing.
- If adding the method fails, check and see if it already is bound to decide
  the return result. Deals with threading races binding the methods.

A uint32 is big enough to hold any return value from that function, and
doing it this way prevents compiler warnings in coded_stream.h about
narrowing a uint64 to a uint32.

Use closurebuilder.py in favor of calcdeps.py for compiling JavaScript

Disable static analyzer for message semaphore creation

* Fix gcc 4.1.2 compilation of map_field_inl.h

Fixes "error: object missing in reference to '...'" errors from #1035

* Disable 64-bit map keys on gcc <= 4.1

* Add missing case statements

This is both more correct, and the build fails on AIX without it

This method merges the contents of the specified message into the
current message. Singular fields that are set in the specified message
overwrite the corresponding fields in the current message.  Repeated
fields are appended. Map fields key-value pairs are overritten.
Singular/Oneof sub-messages are recursively merged. All overritten
sub-messages are deep-copied.

There are two motivations for this:
1) calcdeps.py is deprecated and replaced by closurebuilder.py.
2) As part of this I was able to tweak things so that the Closure
compiler does not attempt to examine every .js file in the tree under
js/. This makes it possible to put compatibility tests and related files
in a subdirectory without them getting mixed up with the main .js files
we care about.

Removed log statement from writer_test.js

This tweak makes the test output a bit easier to read because it cuts
out a lot of unnecessary logging.

Remove the use of C++11 features.

Fix links to the license in meta-data

There might be duplicated enum values when allow_alias is true. Add PreferredAlias into OriginalNameAttribute to remove the duplication (#2727)

The first is the newer name, which is also more telling.

The "New BSD License" is the "3-Clause BSD License", but the links were
pointing to the "2-Clause BSD License".