Re: Clinica package
Le Fri, Jan 13, 2012 at 08:54:22AM +0100, Andreas Tille a écrit :
>
> > Perhaps you did not update the pristine-tar branch before trying to build the
> > package ? The commande ‘gbp-pull’ will do this conveniently for you.
>
> I simply assumed that a `git pull` would simply do the right thing and
> I guess this assumption is valid:
>
> $ gbp-pull
> gbp:info: Branch 'master' is already up to date.
> gbp:info: Branch 'upstream' is already up to date.
> gbp:info: Branch 'pristine-tar' is already up to date.
‘git pull’ will only update the currently checked out branch. I wish there
were an option to update all of them, but it does not seem to exist; I hope it
will be added some day. In the meantime, there is gbp-pull. But why didn't it
work in your hands, I am not sure. If you do not have anything committed, it
is sometimes simpler to just delete the clone and restart from scratch.
> IMHO, *.tar.bz which is the case for clinica does not have the same
> problem as gz and files are identical. Or am I wrong here?
When I checked this morning, the tarballs were not identical, and it is at
least related to the fact that when I use get-orig-source, the tar file
recorded my username and group (charles/charles), whereas they are obviously
not the same in the pristine-tar branch. There may be a workaround… probably
the --group and --name parameters of tar ?
> > About just keeping the debian directory; I would not like it. Having the
> > upstream source checked in is an addictive convenience for bug hunting, and
> > also for copyright checking and other inspections done during upgrades. This
> > is one of my main motivations for using Git. And within that workflow, the
> > pristine-tar branch helps me to work off-line.
>
> I could accept this if it would work as expected but as I had it always
> caused trouble to me (I do not say that it's not me who missunderstood
> the process - but it should work solid enough that it is error prone).
Seriously, I think that you had bad luck. Or taken differently, in some cases
where you worked on git-managed packages, it was to help others, and if others
need help, it is already a sign that the situation is not as easy as usual :)
> $ git-buildpackage -k "Andreas Tille <tille@debian.org>"
> dh clean
> dh_testdir
> dh_auto_clean
> dh_clean
> ./waf --nocache distclean
> 'distclean' finished successfully (0.000s)
> rm -f gschemas.compiled
> if test -d "waflib"; then find waflib -name "*.pyc" -delete; fi
> gbp:info: Orig tarball 'clinica_0.2.1~dfsg.orig.tar.bz2' found at '../tarballs/'
> gbp:info: Exporting 'HEAD' to '/home/tillea/debian-maintain/alioth/debian-med_git/build-area/clinica-tmp'
> gbp:info: Moving '/home/tillea/debian-maintain/alioth/debian-med_git/build-area/clinica-tmp' to '/home/tillea/debian-maintain/alioth/debian-med_git/build-area/clinica-0.2.1~dfsg'
> /bin/sh: 1: Syntax error: end of file unexpected
> gbp:error: ~/bin/git-pbuilder returned 2
> gbp:error: Couldn't run '~/bin/git-pbuilder -k Andreas Tille <tille@debian.org>'
You must have a git-buildpackage configuration file somewhere that has some
non-standard options enabled. On my system, it does not run pbuilder by
default. My workflow is the following:
debcheckout -a --git-track='*' $package
cd $package
git-buildpackage -us -uc # (I do not want to sign them before testing)
# sometimes it fails and recommends to add --git-ignore-new
# git-buildpackage -us -uc --git-ignore-new
cd ..
# Inspect, test, install, …
sbuild -sAd unstable $package.dsc
debsign …
Cheers,
--
Charles
Reply to: