Obviously, it fails.

Conflicts:
	src/lib/datasrc/memory/Makefile.am

Mostly by jinmei, small changes by vorner (assert->throw, as assert can
be turned off).

okayed on jabber

When merging trac2138,
  - Git id of #2138 on the ChangeLog entry will be updated.
  - Skipped tests on tests/system/bindctl/tests.sh will be uncommented out.

Conflicts:
	tests/system/bindctl/nsx1/b10-config.db.template.in

this fixes build failure reported at:
http://git.bind10.isc.org/~tester/builder//BIND10/20120825015000-Solaris10-sparc-Sunstudio/logs/build.out

this sometimes happens, and depending on the environment, what() isn't
always displayed.  so it's helpful to catch it and report the error.

not directly related, but would be useful for extending the benchmark
code and to check the server behavioer.

also clarified RdataEncodeSpec is a forward delcaration (maybe obvious,
but the sudden single declaration looked a bit awkward)

and, make class-vs-struct consistent for it; clang++ requires the consistency.

* Check for impossible return value and throw in that case.
* Call nextInternal directly, to avoid some indirection. This might
  cause a small performance benefit.

They share some bit of private stuff, so we get rid of the private
header. We still have a bit of the code included from outside, but this
is only for tests (and it's not a header).

using EXPECT_EQ(==) instead.

the experimental instrumental reader implementation is now removed.

- spacing
- const
- type clarification
- postion of ++
- indentation
- comment wording

Fixing conflicts in
	src/lib/datasrc/memory/rdata_encoder.h
	src/lib/datasrc/memory/rdata_reader.cc
	src/lib/datasrc/memory/rdata_reader.h
	src/lib/datasrc/memory/tests/rdata_serialization_unittest.cc

added a test case for it and updated the doc.
rbtree/domaintree was modified to avoid such operations.

As we want to make the impression the data is opaque, should not be
examined or pointed into the middle of it.

my compiler even checks assignment-only variables, apparently