[Debian-med-packaging] Inconsistencies in pristine-tar of several packages

[rf at q-leap.de]

>>>>> "Andr" == Andreas Tille <andreas at an3as.eu> writes:

Hi Andreas,

sorry for the long delay with my answer. Other time-critical stuff got
in the way ...

    Andreas> I have no idea as well.  I'm just blindly following our
    Andreas> team policy document git hints and doing
    >>
    Andreas> git import-orig --pristine-tar <tarball>
    >>
    Andreas> expecting it to work nicely (and never observed a problem
    Andreas> in the described workflow).
    >>
    >> I just tested with "git import-orig --pristine-tar
    >> bambamc_0.0.50.tar.gz" after which the id in
    >> bambamc_0.0.50.orig.tar.gz.id is correct and corresponds to the
    >> commit of the upstream import. So git-buildpackage is doing the
    >> right thing but something fishy must have been going on with the
    >> repos after importing the upstream source which lead to the loss
    >> of some commits. Anybody done some history rewriting?

    Andr> I can not seriosly believe this.  The repository is quite new
    Andr> and from the log I'm the only commiter.  I would have probably
    Andr> noticed any other commits.  So I must most probably claim
    Andr> responsibility for this problem even if I have no idea at all
    Andr> in how far I did this - definitely not manually nor
    Andr> intentionally.
 
I don't think you're responsible for this, but some tool or step in the
laid out workflow. We observe this problem in a number of packages, a
lot of them from you, but the most recent example is from Alexander
(bowtie2 2.2.3 from today). So it would really be interesting to check
each step of your workflow and see at what stage the commit referenced
in <pkg>_<vers>*.id (branch pristine-tar) disappears from the repo or is
wrongly changed or whatever.

    >> >> - Shouldn't we fix bambamc_0.0.49.orig.tar.gz.id to contain
    >> >>   the correct
    >> >> commit $(git show-ref --tags -s upstream/0.0.49)
    >>
    Andreas> Well, I'm fine with anything that is fixing a potential
    Andreas> issue (even if I can not really see in what situation this
    Andreas> could cause a problem).
    >>
    >> OK, so if you don't mind, I will fix
    >> bambamc_0.0.49.orig.tar.gz.id (+ the other affected packages) and
    >> commit to alioth.

    Andr> Just go for it.

Meanwhile on our side I wrote a check script that detects, fixes and
pushes the id file to debian.org git with the correct commit id (you
should have seen some such commit mails today coming from Navid).

    >> On another note. Shall I commit an update to the new bambamc
    >> upstream version as well? When I do a backport, I usually like to
    >> do it for the most recent version, unless there are reasons
    >> speaking against it.

    Andr> Definitely!  Feel free to push any enhancement to any package
    Andr> you are noticing.  Please keep the target distribution
    Andr> "UNRELEASED".  Usually I will notice this since I'm reading
    Andr> the commit mailing list and will react as soon as possible.
    Andr> (Just for the record: I'll be on vac from 23.5. - 2.6.  but if
    Andr> needed you will hopefully find some other sponsor.)
 
OK.

    Andreas> I'm fine with any enhancement that might make things
    Andreas> cleaner.  However, we should also document in our policy
    Andreas> document what to do and how to prevent this issue in the
    Andreas> future.
    >>
    >> I'm thinking about how to add checks for such kind of
    >> things. Here at Q-Leap, we notice it automatically when using our
    >> package create scripts.

    Andr> Well, there are a lot of tools you could check.  For instance
    Andr> there is UDD dashboard:

    Andr>    http://udd.debian.org/dmd/?email1=debian-med-packaging%40lists.alioth.debian.org

    Andr> There is the QA team package overview

    Andr>    http://qa.debian.org/developer.php?login=debian-med-packaging@lists.alioth.debian.org&ordering=3

    Andr> You can also use the tasks pages - if something is colored
    Andr> yellow there is something to do:

    Andr>    http://blends.debian.org/med/tasks/bio-dev#libbambamc-dev

    Andr> and you can also run uscan in a cron job and let it send you
    Andr> an e-mail.  I also have code written that scans UDD for
    Andr> upgradable software of a blend and send an e-mail to the team
    Andr> which is in a "close to be finished" state for about two
    Andr> years. :-(

    Andr> Probably there are even more tools (Package Entropy Tracker =
    Andr> PET comes to mind) which could be effectively used.

    Andr> Hope this helps

Yup. An overwhelming amount of information :)

-- 
Roland

-------
http://www.q-leap.com / http://qlustar.com