gen_py target directory for makepy

[Alan Meyer]

I had an unusual problem tonight running makepy to install some
Microsoft COM interfaces in a Python 2.5 Windows XP installation
created using the ActiveState installer.

In earlier versions of Python, the files were generated to:

\PythonXX\Lib\s ite-packages\win32c om\gen_py

But in my 2.5 installation they were generated to my temp
directory, in my case this was:

c:\temp\gen_py\ 2.5

I hadn't paid attention to the messages and when I cleared my
temp directory, which I occasionally do, the COM interfaces
stopped working.

Tracing through the makepy code, I found the following
(reformatted for email) in

\Python25\Lib\s ite-packages\win32c om\__init__.py

if not __gen_path__:
try:
import win32com.gen_py
__gen_path__ = sys.modules["win32com.gen_p y"].__path__[0]
except ImportError:
# If a win32com\gen_py directory already exists, then we use it
# (gencache doesn't insist it have an __init__, but our
# __import__ above does!
__gen_path__ = \
os.path.abspath (os.path.join(_ _path__[0], "gen_py"))
if not os.path.isdir(_ _gen_path__):
# We used to dynamically create a directory under win32com -
# but this sucks. If the dir doesn't already exist, we we
# create a version specific directory under the user temp
# directory.
__gen_path__ = os.path.join(
win32api.GetTem pPath(), "gen_py",
"%d.%d" % (sys.version_in fo[0],
sys.version_inf o[1]))

I was able to make everything work right by creating the gen_py
directory in the expected place and re-running makepy. But I'm
curious to know:

1. Why does it suck to create gen_py dynamically?

2. Why didn't I have a gen_py directory already? Did I somehow
destory it or is it not created by default?

3. Why is the temp directory the alternate choice for where to
put these?

Thanks.

Alan