This is a backport of r65005.

- Issue #2632: Prevent socket.read(bignumber) from over allocating memory
  in the common case when the data is returned from the underlying socket
  in increments much smaller than bignumber.

When a thread touches such an object for the first time, a new thread-local __dict__ is created,
and the __init__ method is run.
But a thread switch can occur here; if the other thread touches the same object, it installs another
__dict__; when the first thread resumes, it updates the dictionary of the second...

This is the deep cause of the failures in test_multiprocessing involving "managers" objects.

Backport of r64601.

whose write() method installs another sys.stdout.

Backport of r64633

seen after a "import multiprocessing.reduction"

An instance of a weakref subclass can have attributes.
If such a weakref holds the only strong reference to the object,
deleting the weakref will delete the object. In this case,
the callback must not be called, because the ref object is being deleted!

Backport of r34309

  conditions in the PyOS_vsnprintf C API function.

This is a backport of r63728 and r63734 from trunk.

now closes its stdout and stderr fds as soon as it is finished with them.

constructors raised an exception.  backportes from trunk c63403.

when open() returns EINVAL.  See issue2158.

reference to themselves.

Backport from trunk.

derive the same default base class.

Backport of r62463.

It seems to work, but will fail at the first flush.

This causes IDLE to silently crash when too many warnings are printed.

Backport of r62448.

Now the code is similar to the one in trunk/.

The behavior was funny:
   >>> print (), repr(())
   (), ()
   >>> print (), repr(())
   (), (...)

  having a self-referential tuple, which is possible from C code.  Nor did
  object's tp_str consider that a type's tp_str could do something that could
  lead to an inifinite recursion.  Py_ReprEnter() and Py_EnterRecursiveCall(),
  respectively, fixed the issues.  (Backport of r58288 from trunk to 2.5.)

remove the now obsoleted assertion.

Prevent PyString_FromStringAndSize() from passing negative sizes on to lower
level memory allocation functions.  Raise a SystemError and return NULL
instead.

Fix zlib crash from zlib.decompressobj().flush(val) when val was not positive.
It tried to allocate negative or zero memory.  That fails.

I have the email trail to prove that George Neville-Neil approved this.
(Backport of r62097.)

Test is run with "regrtest.py -R:: test_compile"

Backport of r62015

"" "" becomes """", which is invalid code.

Backport of r61979.