Ensure that 'once' objects are declared using the macro.  This is either the third or fourth time I've screwed this up when down-integrating, because our internal code does not require the macro (it's not portable) and on Linux a pthread_once_t that is zero-initialized just happens to work.  So, I only discover the problem when I test on Mac, then kick myself for making the same mistake yet again.  No more!  This time, I have renamed GoogleOnceType to ProtobufOnceType, thus making the type name differ from our internal code.  As a result, if you don't update the decls to use the macros, they won't compile.  Hah!  Take that, future self!

be properly set. writeTo() may be invoked without a call to
getSerializedSize(), so the generated serialization methods would
write a length of 0 for non-empty packed fields. Just call
getSerializedSize() at the beginning of writeTo(): although this
means that we may compute the byte size needlessly when there
are no packed fields, in practice, getSerializedSize() will
already have been called - all of the writeTo() wrappers in
AbstractMessageLite invoke it.

Tested: new unittest case in WireFormatTest.java now passes

Explicitly call superclass constructor in generated code to avoid pedantic warning.  Based on patch from Oleg Smolsky.

Fix cross-compiling when building in a dedicated build directory.  Fix pointed out by Pierre Kestener.

Make sure to quality calls to std::swap.  Otherwise, if a google::swap() exists (e.g. because the user is using our own dense_hash_map implementation) it will be chosen instead, leading to a compile error.

Don't use RTTI in down_cast if GOOGLE_PROTOBUF_NO_RTTI is defined.  Patch from Chris Masone (of Google).

Don't build unittest_enormous_decsriptor.proto in C++ because it's very slow on older GCCs.  This file was meant for testing Java anyway.