[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: buildd administration

On Fri, Dec 09, 2005 at 05:48:13PM +0100, Josselin Mouette wrote:
> Le vendredi 09 décembre 2005 à 12:07 +1000, Anthony Towns a écrit :
> > That's non-sensical. Everything the buildds do is logged pretty much
> > immediately onto http://buildd.debian.org/, which also provides long
> > running statistics on how effective the buildds are, and even a schedule
> > of what the buildds will be working on next.
> There is absolutely zero documentation on how the buildd network works.
> You know, the things you have to be aware of if you want to give a hand.


The 3 html pages above contain about 3500 words of explanation about the
buildd network. Plus there is the source, and quite a number of people
with pretty good insight -- insight you too can become if you're
starting to read up about what it is, reading various sources, talk to
various people about it, etc. Ryan Murray didn't tell me much if
anything about the buildd network when I was trying to understand it,
because I had no reason to go ask him -- yes, documentation is certainly
not perfect, but that's a general issue of most of the things in the
free software world. Numerous people have shown that if you just try,
you'll find plenty of information.

What indeed really could be improved, and Frank Küster helpfully filed a
wishlist bug for that on www.d.o, is listing somewhere contact addresses
for the buildd admins in case there is a buildd-system related issue.
Isn't it more productive anyway that if there's percieved lack of
documentation to simply actively work on that, rather than complaining
loudly? Or that if you think the process itself has flaws or is
understaffed, to help productively, like Anthony Towns notes[1]?
Because, hey, Anthony is right there.

Let me tell you story of http://buildd.debian.org/~jeroen/status/ -- I
noticed that sometimes due to system crashes, network downtime, or
whatnot, a few packages might end up in state 'Building' for a prolonged
time, and that there was no automated mechanism to unstuck it -- it
needed someone to note that, and then the buildd admin requeued it.
Instead of complaining loudly about that on the lists, I asked myself
how this could be resolved. The first thing needed for that was
detection of the issue itself, so I wrote the above-mentioned page.
And I noted that the problem was much less than I thought. Anyway,
Steve Langasek picked up actually looking at it regularly, and feeding
back give-back suggestions to the buildd admins when needed. And after a
while he was granted full wanna-build access to do it himself, because
he has shown to understand how it works.

A similar issue I noted in the past is the big number of build failures
that don't get tagged 'Failed'. I tried working on classifying them, but
got bored so increadibly fast that I gave up, and decided for myself
this should be something the porters should rather look into. And thus I
mailed in June about that[2]. I don't believe anyone picked up that, but
I might be wrong, of course, my mail was hidden in a big thread about
various stuff, just like this very mail is... Someone interested in
porting (actually, I personally am myself not interested at all), should
maybe mail to all arch-specific lists some request similar to Vince
Sanders' request[3] regarding classifying arm failures.


[1] http://lists.debian.org/debian-devel/2005/12/msg00323.html
[2] http://lists.debian.org/debian-devel/2005/06/msg00674.html
[3] http://lists.debian.org/debian-devel-announce/2005/12/msg00002.html

Jeroen van Wolffelaar
Jeroen@wolffelaar.nl (also for Jabber & MSN; ICQ: 33944357)

Reply to: