on the action's default, reverting it back to previous behavior.  Conversion
is only done on string defaults.

Add a test for this and another test that ensures such type conversions are
only called once.

Cleanup 'k' when the creation of PySTEntryObject fails. ad3824a90261 used to decref 'k' in too many error cases.

Fix out of bounds read in long_new() for empty bytes with an explicit base. int(b'', somebase) calls PyLong_FromString() with char* of length 1 but the function accesses the first argument at offset 1. CID 715359

  when action='append', type='str' and default=[].

Fix null pointer dereferencing in structmember.c PyMember_SetOne() for T_CHAR. _PyUnicode_AsStringAndSize() can return NULL without touching the len argument. Also remove unnecessary PyUnicode_Check(), _PyUnicode_AsStringAndSize() performance the test again. CID 486815

Fixed memory leak in error branch of object_repr which may leak a reference to mod when type_name returns NULL. CID 715371

Fixed memory leak in error branch of parsestrplus. CID 715374 Variable s going out of scope leaks the storage it points to.

Also adds tests to verify the documented behavior (which is probably a bug, as
indicated in the added comments).

Patch by Chris Jerdonek.

	changeset:   77827:c23b442b5d5e
	user:        Antoine Pitrou <solipsis@pitrou.net>
	date:        Thu Jun 28 01:20:26 2012 +0200
	summary:     Avoid using scrdir, it's broken.

	changeset:   77826:f0e58e778215
	user:        Neil Schemenauer <nas@arctrix.com>
	date:        Wed Jun 27 15:58:37 2012 -0600
	summary:     Fix bug in test_tools that prevented building is separate directory.

Py_TYPE() has already dereferenced self before the NULL check. Moved Py_TYPE() after the check for self == NULL

they are created in the install locations and with the proper timestamp.
(Solution suggested by MvL)

#15865: add "*" in the signature to document keyword-only args in the docs.  Patch by Chris Jerdonek.

Patch by Chris Jerdonek.

This fixes a regression introduced by the fix for issue #13922.  Although args
is not documented as being allowed to be a tuple, previously this worked and
so naturally there are programs in the field that depend on it.

Patch by Zbyszek Jędrzejewski-Szmek.

This was a regression caused by the hash randomization patch.

BSD's make doesn't support some of the features.

sporadic crashes in multi-thread programs when several long deallocator
chains ran concurrently and involved subclasses of built-in container
types.

Because of this change, a couple extension modules compiled for 3.2.4
(those which use the trashcan mechanism, despite it being undocumented)
will not be loadable by 3.2.3 and earlier. However, extension modules
compiled for 3.2.3 and earlier will be loadable by 3.2.4.