Re: Debian needs more buildds. It has offers. They aren't being accepted.
Ingo Juergensmann <ij@2004.bluespice.org> writes:
> On Fri, Feb 13, 2004 at 11:32:55AM +0100, Michael Banck wrote:
>
> > > You say that this issue should be brought up to the ctte, but
> > > that we (the three persons you're mentioning) basically can't do
> > > that because we're not a developer and therefore that would be
> > > inappropriate.
> > Look. I really think your kind-of-renegade attitude towards becoming a
>
> There's no renegade. At least only as much as there's a cabal... ;)
And he is not the only one in on this so his not wanting to become a
DD shouldn't be made an issue. Thats his personal decision and he has
reasons for it.
> > DD ("They can't communicate! I don't want to be part of a project where
>
> I didn't said that all DDs can't communicate. Some are excellent in that.
> Some are not.
> But I do think that communication is part of the work in some role positions
> where other peoples work depend on the work of those people who are in these
> positions. And when these people can't communicate because it's always been
> said that they have too much work to communicate, then I think it's best to
> load off some work from those people or find another way to work around this
> communication problem. See my mail to tbm.
>
> > they can't communicate! But nevertheless I donate my time to the
> > project") is kind of cool (not as cool as Overfiend used to before he
> > became a politician, but still), but you're wasting everybody's time
> > here.
>
> So it's better to have other people waste *my* time then? (Because I waste
> my time as others do because those are unresponsive or don't do their work
> subjectively?)
And my time and Wouters time and Martin Loschwitzes time and the time
of the maintainers of the other 79 packages being stuck in sarge and
the time of countless users that wonder why their favourite package is
still stuck in sid. Woah, thats a lot of time all of a sudden.
If we can come to an agreement that just prevents one of those
occasions from hapening in the future I thing we saved enough time to
counterbalance this flame.
> > If Nathanael and you don't want to become DDs, that's fine, but both
> > sides have to live with the consequences. One of the consequences is
> > that you don't get to demand stuff from respected DDs, just because you
> > happen to think otherwise.
>
> Oh, it's not the case I don't want to become a DD... I really would like to,
> but there are reasons why I don't apply as NM.
>
> > > > For buildds that's James and Ryan; and they've been doing that
> > > > very well for years now.
> > > Apparently that isn't anymore the case.
> > Apparently *you believe* this is not anymore the case. Apart from calc,
>
> See the mails on debian-mips the last months. I'm not alone with my believe.
>
> > I don't see a lot of DDs argue about this. And aj, elmo and neuro seem
>
> See the archive for m68k-build for "please built my package!"-mails.
Or the many claims about doorstoper archs and threats to drop them or
release without them.
> > to be either not much concerned or working on it, so this is perfectly
>
> "not much concerned" is maybe the right word... ;-)
>
> > alright with me. And for the record, I'm working on the Hurd port, and
> > we did not have an autobuilder *for over half a year*.
> >
> > IMHO, buildds are important, but keeping them current is not nearly as
> > important as you want to believe. It's not like "OMG, a buildd went
> > down, stop the presses!!!1!". And it's not a sport either ("If we've had
> > three more buildd for arch foo, we'd be up from 97.3% to at least 98.5%,
> > but *you* block the future of mankind")
>
> Wrong. See ajs Mail about that port become real thingie. The ports must
> fulfill a certain percentage of built packages to get released (as one
> requirement). Dropping from 97% to under 95% or 90% can happen fast, whereas
> getting back to 97% can take *weeks*. When this happens at release time,
> everything is getting worse because of frequent uploads of packages.
> So, blocking suitable buildds from getting w-b access *is* actually blocking
> at least the future of the port.
> As said in another mail, we wouldn't have that small backlog when Goswin,
> Wouter and Stephen would invest their time to workaround this issue. I guess
> we would have more likely a backlog of about 250-300 packages.
Build daemon statistics from 20040205-1951 to 20040212-1950 (7.00 days):
#packages % of taken pkgs/day
-------------------------------------------
builds : 215 0.00% 30.72
uploaded : 111 0.00% 15.86
time % of total
----------------------------------
building: 159:35:13 95.00%
idle : 00:30:00 0.30%
Add those to the needs-builds list each week. The buildd is now
running fully configured by Wouter for weeks and I have been building
packages at the same rate before that with the buildd setup I did
myself.
The buildd has to be fed and managed manualy though which eats up much
more time than the simple "reply to mail" managment possible with
wanna-build access.
> When Goswins machines would have w-b access that would take extra and
> unnecessary workload from these people giving them time to work on other
> issues like d-i.
>
> > > > *None* of them have been necessary or productive, and if that's the
> > > > reward for the level of contributions James has offered the project,
> > > > then it's clear that the project doesn't want responsible committed
> > > > people to be a part of it.
> > > Maybe they're just not productive because some[TM] people block the whole
> > > process from being productive? Would be worth a though, eh? ;)
> > I don't see how the whole process is being blocked. Why don't you let
>
> When they would do their work, nobody would complain. But there are
> complaints. Still there was no reasoning of Ryan why qt-x11-free wasn't
> built for about a month, although it is his work as a buildd admin to ensure
> the packages to get built. When he can't deal with that, he has to accept
> unnice questions about it.
qt-x11-free is in weak_no_auto_build on Ryans autobuilder. It will
only be build when it runs idle otherwise. That simply hasn't been the
case all this time.
> > elmo and neuro do their work in the way they think is best, while *you*
> > do *your* work (as in the m68k team does the m68k porting) in the way
> > you think is best.
>
> Right. And sometimes we need to interact with each other and work together,
> which can't be done when - you guess it! - there's one party that is
> unreachable for communication.
> Sometime they need some information from me, sometimes I need information
> from them. And neither can do their work when the other side don't
> communicate.
>
> > If this includes running a seperate wanna-build
> > database, I think this would be an acceptable compromise for the time
> > being, and I don't get why you guys rather like running against the
> > 'James Troup Wall[TM]' then getting the work done.
>
> Running a seperate w-b produces extra work for others, such as ftp-masters.
> This is not the way to go. M68k had its own w-b back then and it was
> shutdown for that reason.
It would also still need access to the accepted/autobuild repository
for the normal efficient operation which is also controled by, guess
who, James.
Running ones own wanna-build is just a patchwork fix for the larger
problem, not a real solution.
MfG
Goswin
Reply to: