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

Re: BREAKING NEWS: Debian developers aren't trusted

Anthony Towns a écrit :
> On Thu, Feb 15, 2007 at 06:34:16PM +1100, Hamish Moffatt wrote:
>> On Thu, Feb 15, 2007 at 01:13:36PM +1000, Anthony Towns wrote:
>>> -vote dropped
>>> On Wed, Feb 14, 2007 at 03:06:01PM +0100, Martin Zobel-Helas wrote:
>>>> i think someone running more than one autobuilder for more than _two_
>>>> years now (okay, not for the officical archive, but i see that as
>>>> nonrelevant here) demonstrats very good that he mets your mentioned
>>>> technical constraints.
>>> AIUI, Aurelian doesn't have the capability to run a non-emulated arm
>>> buildd. While http://blog.aurel32.net/?p=25 is a good demonstration
>>> of some things, I don't think it's the level of buildd we want for our
>>> release architectures.
>> Wow, was there a point to your post or was pure insult?
> What's insulting in that (apart from the misspelling of Aurelien)?
> From http://blog.aurel32.net/?p=33:
>     Yes I agree that real machines would be better, but I dont have a
>     stack of fast ARM machines at home.

First of all I haven't asked to become arm build daemon maintainer. What
I asked for a responsive arm buildd maintainer who handle mails sent to

The goal of the emulated arm buildd farm was to be able to handle the
requests by myself, in fully automated way (building and uploading
package by hand takes a lot of time).

I also asked some news about the arm machine offered by Bill Gatliff.
This machine is really fast compared to the current build daemons and
replace a few of the current one. As described by Steve Langasek on
-vote, this can reduce some problems and the load of the build daemon

> so, afaict, he doesn't have the capability to run a non-emulated arm
> buildd. And hacking together a buildd quickly and easily is impressive
> and useful for new ports, but it's more important for buildds for release
> architectures to be well-connected and reliable over a long period. We
> probably could change that expectation and have buildds be put together
> by DDs from whatever they have lying around and hosting them at home;
> I don't think it'd be a good idea, but others mileage may vary.

I fully agree that knowing how to install wanna-build + buildd + sbuild
don't make you a buildd maintainer.

FYI, I am running a wanna-build database for hurd-i386, kfreebsd-i386
and kfreebsd-amd64 on my home server, and running three build daemons,
two for kfreebsd-i386 (yes, contrary to some official architectures we
have buildd redundancy), and one for kfreebsd-amd64. And that for almost
2 years.

I have learned a lot from that, experienced hardware problems, chroot
breakages due too buggy maintainer scripts, and even toolchain problems.
The kfreebsd-amd64 build daemon has been added very early in the
development of this architecture (ie two or three weeks after the
toolchain has been ported), and I think I have learned more from that
than if it has been an official and fully mature architecture.

  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Reply to: