Gentoo Archives: gentoo-project

From: Matt Turner <mattst88@g.o>
To: gentoo-project@l.g.o
Subject: Re: [gentoo-project] Council discuss: overlapping council terms of two years
Date: Sat, 06 Aug 2011 15:54:03
In Reply to: Re: [gentoo-project] Council discuss: overlapping council terms of two years by Markos Chandras
On Sat, Aug 6, 2011 at 11:42 AM, Markos Chandras <hwoarang@g.o> wrote:
> On 08/06/2011 04:33 PM, Matt Turner wrote: >> On Sat, Aug 6, 2011 at 6:00 AM, Markos Chandras <hwoarang@g.o> >> wrote: >>> I never said to completely drop these arches. When did I say that? >>> I just want a more realistic approach on how well an arch is >>> supported. Why you people are afraid to admit that we have >>> problems? Having an arch with constantly >200 stabilization bugs >>> open clearly proves that the manpower cannot handle the situation. >> >> I think it's important to put some numbers on this. >> >> x86              80      2      13 amd64            40      1      7 > > This is just hilarious :) The numbers of developers are not even close > to reality
Same situation with the rest of the architectures, really.
>> The only architecture that is seriously backlogged in ppc, which is >> probably due to the fact that we used to have lots of users. Just a >> couple of weeks ago, ppc64 was in the same situation, until >> xarthisius > > What if xarthisius, armin76, me and jer take 3 months off? What an evil > scenario :). The problem is when an architecture relies on a *single* > (or max 2) developers. You can't possibly claim that this architecture > is supported. You have a single/double point of failure. They can easily > retire someday or even lose their motivation. And then what? It would be > far too late to act
We're in pretty bad shape if that happens. We do have Maintainers could help arch teams greatly by giving their package a quick test build on the development boxes. Not sure that's really an acceptable solution though. I wonder if we can't set up an automated system where a package maintainer goes to a webpage and enters the package they'd like to submit for a keyword request. The page would display the requirements for each architecture, and the maintainer could then start a test build that would run on the development boxes in a testing chroot, and then give the results back to the maintainer. That would certainly make the situation simpler for everyone. Maybe that's deserving of a separation thread. Matt