Re: inconsistent versions of M-A: same packages
Hi Holger,
Quoting Holger Levsen (2014-11-07 16:31:09)
> > I agree with Ralf, that this would best be done not by debcheck but by a
> > small script which compares if the Packages files for all distributions
> > ship M-A:same packages in the same version.
>
> I'd happily run this script on jenkins.debian.net... if someone "gives it" to
> me ;-)
is jenkins not triggered by pushes to git and thus sub-optimal for jobs that
should be run like a cron job?
I thought I'd run such a script on bootstrap.debian.net because that one has
all Packages files for all architectures already and it would be trivial to run
such a script in addition.
But first I'd like to make sure what the interesting information would be.
Would it be interesting to get the highest version of a binary package across
all architectures and then check if that version exists in all architectures?
Or would it be interesting to make sure that all versions exist across all
architectures? (binary packages can exist in more than one version in unstable)
botch currently ships the program botch-check-ma-same-versions which checks
whether two Packages packages files agree on the M-A:same versions. Running it
with amd64 and armhf unstable Packages files as input yields that all 25
version skews are due to binNMUs. I can extend the script to allow more than
two Packages files as input and make the output machine readable.
> > One interesting scenario for which to detect co-installability problems is
> > when it comes to satisfying crossbuild dependencies.
> >
> > The following page is regenerated daily:
> >
> > http://bootstrap.debian.net/cross.html
>
> cool, very nice!
>
> > I have new hardware now, so I plan to extend the set of source packages
> > that I check for crossbuild dependency satisfaction.
>
> not sure what ressources you need, but maybe jenkins.d.n can also help here?
> (or alternatively, jenkins can also be used to just notify about problems like
> the d-i_overview* jobs from https://jenkins.debian.net/view/d-i_misc/ do...)
Gladly, dose3 is very fast and even checking the whole archive doesn't take
more than 3 minutes :)
The problem before was, that I was running the script on a shared server with
*very* limited resource constraints.
But thanks for the offer!
cheers, josch
Reply to: