> You're welcome, and thanks for following this through. I still have
> machines around that have PATHEXT=.pyc;.py;... and will now remove .pyc
> from all of them. It would probably be nice to trace this back to the
> origin, find whether there was a good rationale for it being that way in
> the first place, and either update a FAQ somewhere or get the problem
> fixed once and for all.
>
> I don't think the standard install messes with PATHEXT, so my bet is on
> ActiveState right now. Can anyone confirm? Trent, if you're reading
> this and it is from ActiveState's distribution, do you know where that
> PATHEXT=.pyc;.py choice came from and whether it might have been an
> ill-advised decision that should be undone?
Yes, I can confirm as well. This is a bug in ActivePython. Definitely
not an intended decision. That is: having .pyc (and/or .pyo) before .py
(and/or .pyw) on PATHEXT is definitely a bug. The ActivePython MSI just
added the .pyc, .pyo, .py and .pyw extensions to PATHEXT *backwards*. My
apologies for not having caught this.
http://bugs.activestate.com/ActivePython/show_bug.cgi?id=33311
As to whether to have .pyc (ditto .pyo) on PATHEXT at all: ActivePython
has been doing so since time immemorial. I think that ActivePython
should stop doing that (as implied by Duncan) unless someone can suggest
a reason why to keep it. The only use case would be for calling a Python
script on Windows without specifying the extension *and* for which there
is ONLY a .pyc or .pyo. That seems to unlikely (or uncommon) a scenario
to justify it.
I'll get a new ActivePython for Windows out soon to take .pyo and .pyc
off of PATHEXT.
Trent
--
Trent Mick
[EMAIL PROTECTED]
--
http://mail.python.org/mailman/listinfo/python-list
