develop for Windows on GNU/Linux, using Python

[Brian Curtin]

On Tue, Sep 21, 2010 at 03:36, Lawrence D'Oliveiro
<ldo at geek-central.gen.new_zealand> wrote:

> In message <mailman.889.1284921376.29448.python-list at python.org>, Kev
> Dwyer
> wrote:
>
> > To be confident that your code is good you need to test it on a Windows
> > box (we all test, right?).
>
> Preferably more than one. Test with Seven as well as Vista (yes, there are
> still some Vista users out there). What about the difference between
> Starter
> versus Home Basic versus Home Premium versus Professional versus Ultimate
> versus whatever? To be 100% safe, you probably need to test with all those
> as well.
>
> Test on machines running the various major antivirus packages, because you
> have no idea how they might screw things up.
>
> In other words, with Windows the guideline is: test till it hurts. Then
> test
> some more.


Testing with every single Win7 SKU is not necessary. The Windows API is the
same on all of them, the file system is the same, they are the same
underlying operating system, but they are equipped with different types of
tools. For example, at one of those levels they enable Terminal Server,
something that should not affect your project (unless your project is a
Terminal Server related tool, of course).

For CPython, we test with XP (Win32 API version 5.1) and 7 (ver 6.1, same as
Server 2008 R2). I have previously done a lot of my development on Server
2003 (ver 5.2, same as XP x64), although there is no buildbot slave running
for that OS. If I need anything tested for Vista (ver 6.0, same as Server
2008), another OS without a buildbot slave, I know of a Vista user that can
usually lend a hand. There would be no benefit to us to test CPython on all
available releases of each OS -- it would be more of a burden, actually.


As for antivirus screwing things up, that may be of interest to some
projects, mainly those opening and then deleting files very quickly. The
CPython test suite occasionally runs into issues deleting temporary files
because the AV tools that some users have installed try to scan them in the
very small time frame the file is intended to exist. TortoiseSVN also
latches on long enough that handles to the file exist when a test tries to
clean up after itself, thus causing seemingly random test failures. There's
an open issue to fix that, though.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-list/attachments/20100921/347e5582/attachment-0001.html>