[Python-Dev] Buildbots for trunk are all red

[Brett Cannon]

On Tue, Feb 26, 2008 at 9:46 PM, Gregory P. Smith <greg at krypto.org> wrote:
>
>
>
> On 2/26/08, Brett Cannon <brett at python.org> wrote:
> > On Tue, Feb 26, 2008 at 1:52 PM, Facundo Batista
> > <facundobatista at gmail.com> wrote:
> > > 2008/2/26, "Martin v. Löwis" <martin at v.loewis.de>:
> > >
> > >
> > >  > They check out bsddb from subversion, see Tools/buildbot/external.
> > >  >  If you don't trust that they did so correctly, edit the script to
> > >  >  remove bsddb, check that in, wait for them to delete it, then revert
> > >  >  the script, check in again, and see how they build it.
> > >
> > >  No, I wasn't aware of this mechanisms at all. I don't even know how to
> > >  build Python in a Windows...
> > >
> > >  Anyway, I don't think it's a bad checkout or something, as the same
> > >  error happens in two different machines.
> > >
> > >  I don't know, :(
> >
> > Or we can get rid of bsddb and not have the problem anymore. =)
> >
> > -Brett
>
> -1
>
> Using that logic I prefer just getting rid of Windows to stop having these
> problems.  That'd solve the ssl applink issue and msi installer building
> issue as well. =P
>

=) Well, we can't have all our dreams come true.

> My opinion on bsddb as a standard library module is based mostly on "its
> always been there" and a vague memory of the last time this came up I
> thought people piped up saying they liked batteries being included,
> including bsddb and sqlite, but I don't have a handy reference to that email
> thread.

Well, in my opinion "batteries included" is great, but not when one of
the batteries consistently acts up and requires a good shake to get
working again. The bsddb module has consistent reliability issues when
it comes to testing (and I suspect it has more to do with Sleepycat
than the bindings). I know I am tired of getting buildbot errors
saying that the bsddb tests died more consistently than most tests
over their history. And I did some work on porting bsddb over to
Python 3.0 and it was painful.

Anyway, I am not pushing for this now. Any major removal plans will go
through the stdlib-sig first and require justification before I even
attempt to seriously suggest something beyond a joking line in an
email (in other words, no one needs to worry about anything right
now).

-Brett