Patch from Mark Hammond:
bdist_wininst attempts to use the correct MSVC runtime for the current
version of Python. This doesn't work correctly when --target-version
is set. In that case, bdist_wininst still uses the *current*
sys.version (ie, 2.4) rather than the version specified as
--target-version. Thus, the msvc7 runtime based executable stub is
*always* used.
This patch "hard-codes" knowledge of earlier Python versions,
providing the correct result when Python 2.4 is used to build Python
2.3 and earlier distributions.
Remove the short variant (-v) of the --target-version command line
options, it conflicts with the --verbose/-v standard distutils switch.
This serves to made the context representation more useful (the names in
match the names out).
Removed the link to ReXX. That document did not shed add anything to
information already in the spec.
Further tighten the public API to improve usability:
* Emax and Emin are set through Context(). Their defaults are set in the
DefaultContext, so there is no need to expose DEFAULT_MAX_EXPONENT
and DEFAULT_MIN_EXPONENT.
* The string functions isnan() and isinfinity() were only used internal to
the module and were not among the test cases. External use cases are
served by the Decimal constructor. So, made these private.
and format the warning similarly to an exception for both that check and for
warnings raised in the subprocess.
M NEWS.txt
M Pyshell.py
M ScriptBinding.py
M run.py
* delete temporary globals immediately after use
* move a global into a class variable
* Rename BasicDefaultContext and ExtendedDefaultContext
to BasicContext and ExtendedContext.
Added error checking to log() to check that level is an integer, and raise a TypeError if not (as long as raiseExceptions is set).
Minor documentation corrections.
* Added test for pickling contexts
* Renamed ExceptionList to Signals (to match wording in the spec)
* Simplified Context constructor by allowing flags=None to automatically
generate a zeroed-out flags dictionary.
* inlined _convertString() which was used only once
* _rounding_decision is private, so excluded its contants from __all__.
* added an XXX comment with concerns about subclassing signals results in
a deviation from the spec (maybe important, maybe not).
* Taught the test_suite to determine its own directory (modeled after code
in regrtest.py). Enables it to be run when the current directory is not
the test directory.
* Added a clear_flags() method to the Context API to make it easier to do
a common operation with flags.
* Fixed the trap_enablers defaults in BasicDefaultContext to match the spec.
during interpreter shutdown instead of masking it with another traceback about
accessing a NoneType when trying to print the exception out in the first place.
Closes bug #754449 (using patch #954922).
- weakref.ref and weakref.ReferenceType will become aliases for each
other
- weakref.ref will be a modern, new-style class with proper __new__
and __init__ methods
- weakref.WeakValueDictionary will have a lighter memory footprint,
using a new weakref.ref subclass to associate the key with the
value, allowing us to have only a single object of overhead for each
dictionary entry (currently, there are 3 objects of overhead per
entry: a weakref to the value, a weakref to the dictionary, and a
function object used as a weakref callback; the weakref to the
dictionary could be avoided without this change)
- a new macro, PyWeakref_CheckRefExact(), will be added
- PyWeakref_CheckRef() will check for subclasses of weakref.ref
This closes SF patch #983019.
The builtin eval() function now accepts any mapping for the locals argument.
Time sensitive steps guarded by PyDict_CheckExact() to keep from slowing
down the normal case. My timings so no measurable impact.
[ 924301 ] A leak case with cmd.py & readline & exception
by ensuring that the readline completion function is always reset
even in the case of an exception being raised. As a bonus, this
makes the documentation for pre & postloop accurate again.
Fix memory leaks revealed by valgrind and ensuing code inspection.
In the existing test suite valgrind revealed two memory leaks (DB_get
and DBC_set_range). Code inspection revealed that there were many other
potential similar leaks (many on odd code error paths such as passing
something other than a DBTxn object for a txn= parameter or in the face
of an out of memory error). The most common case that would cause a
leak was when using recno or queue format databases with integer keys,
sometimes only with an exception exit.
opening the DB to implement legacy interface flag='n' support as
BerkeleyDB 4.2.52 no longer allows it in transaction protected
environments. Do the os.unlink ourselves.
regrtest.py after it ran test_frozen. This test was really only
designed to be run immediately after startup. Afterwards, other
modules could be loaded when had not been fixed-up by site.py
Took the chicken way out and only tested those modules known to
be imported by site.py.
* Normalized whitespace.
