Re: summarizing arch status

[Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>]

Steve Langasek <vorlon@debian.org> writes:

> On Mon, Jun 05, 2006 at 11:17:02PM +0200, Goswin von Brederlow wrote:
>
>> Don't forget to also filter out non-free (or contrib sometimes)
>> packages that maintainers no longer build, packages that are not for
>> etch, package that have never been build and maybe even all packages
>> with FTBFS bugs and patches.
>
> non-free isn't in w-b, so is already filtered.

http://unstable.buildd.net/buildd/m68k-all.txt

non-free/text/figlet_2.2.1-4: Installed [optional:out-of-date]
  Previous state was Uploaded until 2004 Oct 14 09:25:22

Just to name the first non-free entry. I've seen other states too.

> never-been-built is filtered out of the graph showing how well the arch is
> keeping up.  It is (deliberately) not filtered out of the graph showing how
> much of the archive is built for the arch.  The latter was not the graph I
> was looking at when commenting.
>
> Packages with FTBFS bugs should be either NMUed by porters (if the failures
> are arch-specific), or kicked out of unstable (if not arch-specific and left
> unfixed).  Either way, when comparing with other architectures common build
> failures will clearly not show as counting against any particular arch.

Not enough people do take advantage of porter NMUs though. OK, lets
blame them for not utilizing all their power.

>> Wouldn't it be much more meaningfull to go through the update excuses
>> and count the number of packages kept out of testing due to each arch
>> (or potential number for non-blocking archs)?
>
> We do this too.
>
> m68k doesn't look so hot by this metric either.
>
> Out of dates holding up testing:
>      14 i386
>      24 amd64
>      64 s390
>      68 sparc
>      77 powerpc
>      86 mipsel
>      98 mips
>     109 hppa
>     111 ia64
>     115 alpha
>     124 arm
>     191 m68k

Can you put those numbers somewhere on a daily basis and maybe make a
graph?

Those numbers nicely show that m68k does have problems currently.
Hopefully a lot of those will go away now that gcc is gcc-4.1 as there
have been a lot of failures with gcc-4.0 on m68k.

MfG
        Goswin