Re: m68k arch falling ever behind
On Thu, Mar 25, 2004 at 07:17:06PM +0100, Sebastian Ley wrote:
> > A more sophisticated approach to determine which package needs to get built
> > first in order to build more packages successfully in given time. :->
> Which probably is not an easy problem that would take some computing
> time. Time which could be used for something useful, e.g. compiling ;-)
Too bad that the computing algorithm for the queue order determination
doesn't run on a buildd itself but on a different machine. The client just
fetches a short list of packages that needs to be built.
So, until you or someone else is suggesting a better way to deal with the
current buildd problems that aren't related to m68k only, I won't comment on
specifics anymore.
You can invest some time in finding reasons why some archs has a high number
of building packages, f.e. s390.
Funny enough, m68k is the arch that has the lowest number of packages in
state building:
alpha_stats:Building : 156
arm_stats:Building : 162
hppa_stats:Building : 61
i386_stats:Building : 6
ia64_stats:Building : 85
m68k_stats:Building : 56
mips_stats:Building : 143
mipsel_stats:Building : 162
powerpc_stats:Building : 100
s390_stats:Building : 212
sparc_stats:Building : 79
Take s390 f.e.: http://www.buildd.net/buildd/s390_stats.png
Look at Needs-Build and Building.
Maybe you should re-read Chris Cheneys mail and my first reply to his mail before
"argueing" with non-arguments. No personal offense, but I see way too often
people comment on things they're not involved in and can't really comment on
therefore, too often that non-issues are "discussed", too much time wasted
and nothing happens to the real existing problems, which are too often
simply ignored or denied.
(Sorry, that you're the guy that has the bad luck of getting this kind of
mail now! There's an effort of making a better replacement to w-b/buildd, so
be assured that many problems will be addressed to get solved with that new
framework.)
--
Ciao... //
Ingo \X/
Reply to: