[Python-Dev] Using temp files and the Internet in regression tests

Brett Cannon drifty@alum.berkeley.edu
Sun, 13 Apr 2003 19:13:30 -0700 (PDT)

[Guido van Rossum]

> > I can write some basic instructions on how to use regrtest and
> > test_support; someone will just have to check them in.
> That would be great.  Do you have a SF userid yet?  Then we can give
> you commit privs!

bcannon is my username.  I was going to wait to ask for commit privs until
I had done more patches (specifically C stuff), but if you think I am
ready for it then it would be extremely cool to get commit privs (and not
have to wait for anonymous CVS updates when the servers get overloaded
or bug people to commit _strptime patches =).

> > > Hm, maybe these docs shouldn't be so hidden and there should be a
> > > standard library chapter on the test package and its submodules and
> > > the conventions for writing and running tests?
> >
> > That definitely wouldn't hurt.  It might also get people to write
> > tests more often and maybe help with improving our code if they knew
> > about regrtest and test_support.
> And I think regrtest and test_support are useful for testing 3rd party
> code as well.  Wanna make this a project?

I could.  Going to have to learn more LaTeX (and the special extensions).
So I can take this on, but I can't make any promises on when this will get
done (I would be personally horrified if I can't get this done before 2.3
final gets out the door, but you never know).

Should there be a testing SIG?  Could keep a list of tests that could
stand to be rewritten or added (I know I was surprised to discover
test_urllib was so lacking).  Could also start by hashing out these docs
and making sure regrtest and test_support stay updated and relevant.

Personally, I think writing regression tests is a good way to get new
people to help with Python.  They are simple to write and allows someone
to be able to get involved beyond just filing a bug.  I know it was a
thrill for me the first time I got code checked in and maybe making the
entry point easier by trying to get more people to write more regression
tests for the libraries will help give someone else that rush and thus
become more involved.

Or maybe I am just bonkers.  =)