[IPython-dev] Test failures and errors in trunk

[Fernando Perez]

On Wed, Mar 18, 2009 at 4:35 PM, Brian Granger <ellisonbg.net at gmail.com> wrote:

> Yep, I definitely agree with this.  Sorry about my test failure grumpiness :)

No worries :)  Let's not beat ourselves though; instead we can learn
useful lessons to get better:

- big merges==bad.  That was totally my fault on this particular
occasion, I left that branch of mine grow too big for too long.  Big
apology.

- When reviewing non-trivial branches from someone else: we must
always branch it locally, switch into it and run it before approving.
I did manage to do that with Jorgen's py2exe-fix one, and I got errors
he wasn't seeing.  I made some fixes and republished it for him to
pick up, and eventually we were able to clean things up outside of
trunk before it got merged.  That even lets you test the merge into
trunk locally to see how it would go (in a throw-away copy of trunk
you only use locally), before 'the public' sees it.

  For example, if Jorgen had gotten my branch (even without merging,
just running it as-is), he would have seen a ton of those errors that
would have alerted me to my mistake of forgetting about dependencies.

  For this kind of work, one thing that may be useful, if a particular
branch turns out to require collaborative work, is to change it from
~person to ~ipython-dev.  That way more than one person can directly
commit into it, svn-style, before it goes into the real trunk.  We've
done that with nipy a few times to speed up the collaboration between
two or three people in shaping up a branch before merging it into
trunk.



So the good thing is that we're refining our process, learning, and
these problems do look fixable (though a few are indeed subtle).

I'll continue to fix things as much as I can over the next few days, promised.


Cheers,

f