A service for testing Python code on multiple platforms and versions

[David Moss]

Hopefully a service like this already exists and I just haven't found
it yet. If not it could be an idea for some kind soul(s) to pick up
and run with ;-)

As someone who writes and releases Python modules for the community, I
find it difficult to have a decent level of confidence in the efficacy
of my code on platforms and Python versions other than those that I
own or use regularly. My documentation states that I support Python
2.3 or higher. This ends up being more of a statement of good
intentions than a one of fact.

A case in point. A bug (in Python), that I believed to have been
killed off after Python 2.2, resurfaced in a 2.4.x release of Python
on PowerPC recently. As I don't own any PowerPC kit, it was very
difficult to a) investigate the bug and b) create an effective fix for
it in a timely fashion. Fortunately I'd come across it before so the
fix was easy but it might not have been.

While I realise one's code can never be perfect, you can cover for
these sorts of eventualities fairly easily by running your software
and unit tests under different environments. You'd also like to be
able to do this on a continual basis rather than just once or twice.
If this was done with some kind of automated testing and reporting so
much the better. Bigger projects that take code quality seriously
probably already have this sort of thing in place for their own
purposes, but for smaller ones it just isn't possible.

Wouldn't it be great to have a service/setup out there available for
Python developers to access that covered a fairly broad base of
possible Python installations for the purpose of improve overall code
quality? Am I the only one that would find something like this useful?

Dave M.

[Stef Mientki]

David Moss wrote:

Hopefully a service like this already exists and I just haven't found
it yet. If not it could be an idea for some kind soul(s) to pick up
and run with ;-)

As someone who writes and releases Python modules for the community, I
find it difficult to have a decent level of confidence in the efficacy
of my code on platforms and Python versions other than those that I
own or use regularly. My documentation states that I support Python
2.3 or higher. This ends up being more of a statement of good
intentions than a one of fact.

A case in point. A bug (in Python), that I believed to have been
killed off after Python 2.2, resurfaced in a 2.4.x release of Python
on PowerPC recently. As I don't own any PowerPC kit, it was very
difficult to a) investigate the bug and b) create an effective fix for
it in a timely fashion. Fortunately I'd come across it before so the
fix was easy but it might not have been.

While I realise one's code can never be perfect, you can cover for
these sorts of eventualities fairly easily by running your software
and unit tests under different environments. You'd also like to be
able to do this on a continual basis rather than just once or twice.
If this was done with some kind of automated testing and reporting so
much the better. Bigger projects that take code quality seriously
probably already have this sort of thing in place for their own
purposes, but for smaller ones it just isn't possible.

Wouldn't it be great to have a service/setup out there available for
Python developers to access that covered a fairly broad base of
possible Python installations for the purpose of improve overall code
quality? Am I the only one that would find something like this useful?

I would love to have such a tool, but have no idea how to create
something like that.
Something like a multi-virtual machine on a web server, that you launch
in the night,
and in the morning you would get a full report of all problems on the
specific OSs
(probably someone is going to say that this is impossible, but
fortunately I'm not hindered by any knowledge ;-)

cheers,
Stef

Dave M.
--
http://mail.python.org/mailman/listinfo/python-list

[skip]

DavidAs someone who writes and releases Python modules for the community, I
Davidfind it difficult to have a decent level of confidence in the efficacy
Davidof my code on platforms and Python versions other than those that I
Davidown or use regularly. My documentation states that I support Python
David2.3 or higher. This ends up being more of a statement of good
Davidintentions than a one of fact.

How about Buildbot? Here are some Python examples:

http://www.python.org/dev/buildbot/

And the Buildbot home page:

http://buildbot.net/trac

Skip

[skip]

StefSomething like a multi-virtual machine on a web server, that you
Steflaunch in the night, and in the morning you would get a full
Stefreport of all problems on the specific OSs (probably someone is
Stefgoing to say that this is impossible, but fortunately I'm not
Stefhindered by any knowledge ;-)

You could run Buildbot on a bunch of virtual machines on your computer
(assuming your computer is x86 and its OS is supported by something like
Sun's VirtualBox or VMWare's offerings).

Skip