or, "those python-dev people are awesome."

My experience with the Python bug tracker has been pretty sparse and largely limited to some of the eternaissues like "make HTMLParser deal with even more broken HTML" that never really get resolved because they're not very important and don't have a champion. So when I filed this minor bug report on test_distutils I was not expecting it to be on anyone's top 10 list.

I filed the bug report at 17:21 and dropped a note to python-dev.

Within an hour or so, several people reported that they could not duplicate it, and one person had reported that they could, on the mailing list.

At 19:58 Ned Deily reported through the bug tracker that he could dupe it.

By 20:36, Tarek had a suggestion for something to try (that wasn't the problem, but never mind).

By 21:58, Ned had tracked it down to a UNIX-y flavah difference (BSD vs SysV) in the way group ownership was set on files.

And at 22:30, Tarek had fixed the problem (by dropping the assertions).

So, umm, wow, that was quick! Just over 5 hours, across at least two continents, on a Sunday...

---

I was impressed by how many people chipped in to get a broad spectrum view of things, how quickly some hypotheses were generated & in the end how quickly the issue was resolved -- for a relatively minor problem in one corner of the test suite that didn't show up on any of the buildbots.

I'd bet that part of this speedy response is because Tarek Ziade has taken on stewardship of the distutils code. If so, this highlights how important it is to have people who feel responsible for various bips and bobs of the stdlib & just do whatever needs to be done.

Oh, and it also highlights the value of continuous integration across many machines; I saw the problem because I was running tests inside pony-build in a certain way on a Mac OS X 10.5 machine, and the error wasn't tripped in the 10.6 buildbot.

--titus

Legacy Comments

Posted by Tarek Ziad on 2009-11-30 at 07:17.

Hey Titus,    I am glad we could fix that very quickly. Thanks for the
feedback on this one.     What's great about python-dev is that as
soon as the problem is OS-related and doesn't need any specific
knowledge about the package it occurs in, people will help around. No
matter how many VMs a developer sets, he won't be able to get the same
feedback.    Now for the stewardship you mention, it's a very
important point: every module or package in the stdlib needs a
champion, otherwise it doesn't get the love it deserves. Besides some
specific corners, I think the stdlib has quite a good coverage right
now. And if a topic is not championed by someone, that's something we
should be concerned about.