PIPE_BUF is not defined on Windows, and probably has no meaning there.

Anyway the subprocess module uses another way to perform non-blocking reads (with a thread)

cleanup.  If this fix works, it means that Solaris is unique among
our platforms in what happens when shutil.rmtree is called on the
current working directory (ie: it doesn't work on Solaris, but
it does everywhere else).

if the test gives useful failure info on Solaris buildbot.

(I copied this from another import test, but currently this will fail if
TESTFN ends up in /tmp...see issue 2609).

and make sure we test resetting all three execute bits.

hopefully reveal the real problem.

make sure that out-of-range values consistently raise struct.error.

reorganize the test_struct module to remove duplicated code and tests.

on Windows.

added the shutdown_request() which can perform shutdown before calling close.  This is needed for the ForkingMixIn because different close semantics are required for child and parent process.  shutdown_request(), for TCP servers, calls socket.shutdown() and then calls close_request().  Therefore, this is not an backwards incompatible change, since subclasses that continue to override close_request() continue to work.

Without this change, test_unicode.UnicodeTest.test_codecs_utf7 crashes in
debug mode. What happens is the unicode string u'\U000abcde' with a length
of 1 encodes to the string '+2m/c3g-' of length 8. Since only 5 bytes is
reserved in the buffer, a buffer overrun occurs.

posix zap the execute bit in case it was set on the .py file, since the
compiled files are not directly executable on posix.  Patch by Marco N.

Fixed #6377: distutils compiler switch ignored (and added a deprecation warning if compiler is not used as supposed = a string option)

close_request() (which can send a socket.shutdown()) must be called by the child process in a forking server.  The parent must merely close the socket handle.

integer packing, and reenable some previously broken tests.

operator.sequenceIncludes.

Patch contributed by Jeff Balogh (and updated slightly by me).

names.

it does not fail when file descriptors are large.  Fixes issue3392.

Patch largely contributed by Frank Chu (fpmc) with some improvements by me.
See http://bugs.python.org/issue3392.

Candidate for backporting to release26-maint as it is a bug fix and changes no
public API.

Incorrect exception handling for xmlrpc client retry

Incorrect exception handling for xmlrp client retry

some platforms may raise ENOTCONN if the stack has disconnected the socket on behalf of the peer.