- split it into 3 files
- use ssl instead of socket.ssl
- make it possible to check both regular sockets and gevent sockets

- test__subprocess.py: ignore test_nonblock_remove on win32
- test__example_portforwarder.py
- test__examples.py: ignore geventsendfile.py and psycopg2 on Windows
- testrunner.py: set timeout to 3 min
  some systems have really slow DNS and test__socket_dns.py fails
- test___example_servers.py: use 127.0.0.1 instead of localhost
  helps on Windows server where DNS calls are slow for some reason

always calculate it from environ and file name

it failed with LoopError because thread pool was accidentally used inside thread pool

set_version.py: add --dry-run option
set_version.py: update regexes to match only start of the line
set_version.py: recognize 'dev' as a version
makedist.py: use --version option instead of argument
makedist.py: show git branch

see http://bugs.python.org/issue7735

- remove waitall() from gevent/event.py
- reimplement joinall() with wait()
- remove joinall from __all__
- Hub.join() no longer accepts event= argument

- add result attribute to subprocess.Popen which is an AsyncResult that
  receives return code once the process is finished
- wait() on Windows now has timeout argument and only spawns one thread

- add patch_sys() which patches stdin,stdout,stderr
- add sys=False argument to patch_all()
- remove the module docstring (it's not up to date)

make __enter__ and __exit__ truly noop (previously they called acquire/release)

- even more relaxed comparison of IPv6 addresses
- test__socket_dns.py: replace confusing VERBOSE with a more straightforward DEBUG
- test__socket_dns.py: uncomment a test previously commented out

- testrunner.py: implement --full and --discover
- setup.py now recognizes GEVENTSETUP_EV_VERIFY env var which is passed as EV_VERIFY to make
- relevant env vars are now logged long test names
- add support for expected failures
- allow passing tests on the command line in test___monkey_patching.py
- show longest running tests before final status
- make exit status report number of failing tests
- greentest/util.py: add 'setenv' and 'name' arguments to run()
- add tests_that_dont_use_resolver.txt
- add expected_failures.txt
- fix test___monkey_patching.py to do clean up via atexit

in 5 more tests; this ensures more checks, esp. in debug mode

read() in small chunks, as reading in big chunks from non-blocking fd
fails on Mac OS X.

Do os._exit() in "finally" section, to avoid raising an exception and thus
executing unrelated code of the main process.

Original patch by Vitaly Kruglikov.