Changed numericprog regexpr to make it faster to check.
Removed now unnecessary checks for os.curdir, os.pardir.

Add __members__ attribute.

though the checkin message failed to note that).

2. Fix two bugs in complex():

   - Memory leak when using complex(classinstance) -- r was never
   DECREF'ed.

   - Conversion of the second argument, if not complex, was done using
   the type vector of the 1st.

Added test that ensures that int() and long() truncate float numbers
towards zero.

Removed unused import tempfile.

Added some docstrings.

GDBM module.

attribute, this method silently ignores this fact.  The default (bsddb's dbhash) does.

passed to walk() visitor.

unmarshalling code is actually rather naive and can easily be
caused to crash by feeding it invalid data.  This should be fixed in
the marshal module, but I don't have the time to fix it now :-(

Here's a "keyword" module which, in the spirit of "token.py", updates
the list of keywords automatically from a source file (in this case,
"graminit.c" seemed like a reasonable choice, easier than "Grammar/Grammar").

You get "kwlist", a sorted list of keywords; "kwdict", a dictionary
mapping each keyword to 1; and "iskeyword", a function which tells
you if a given string happens to be a keyword.

This makes it possible to write 1L<<1000000, memory permitting.

was some locking code in this file that needed it...

last variable to which a floating point expression is assigned.  The
macro passes its address to a dummy function so that the optimizer
can't delay calculating its value until after the macro.

(it should return a list of tuples, not a list of lists).