[katello-devel] Bundle install problems should be gone now
Lukas Zapletal
lzap at redhat.com
Fri Aug 5 12:55:11 UTC 2011
On 08/05/2011 02:41 PM, Jeff Weiss wrote:
> Are you assuming that the machine that we're building katello on is the
> same machine we're installing katello on? Because that is rarely going
> to be the case.
>
> If the point of running "bundle install" is just to test something, what
> exactly is it testing? If it's testing that certain gems are there,
> that is helpful, but there is still plenty of other problems - like the
> rpm spec not pulling in a gem. Maybe the gem is already on the build
> box but it won't be there on the machine where katello gets installed.
>
> I know that generally, unit tests require the same libraries to be
> present as the actual application (since obviously they run bits and
> pieces of the app). So perhaps what we are doing wrong is running unit
> tests from source first, THEN building the rpm? Maybe we should build
> the RPM first, then run the unit tests either from source or from a
> 'debug' rpm? I am not sure how this process works with other build
> tools and packages.
Well bundle install --local will just try to create Gemfile.lock and
fail if deps cant be filled.
Let's drop this discussion because it turns out guys from DeltaCloud
already know how to completely get rid of bundler for production setup.
We don't need it when running from RPM and wont need bundle install
anymore (even in the SPEC now).
After discussion with Shannon today we would like to change it and get
rid of bundler for RPM deployments completely. If this is not feasible,
we would go for the solution of running bundle install in the
post-install script. This won't need any build requires so I will remove
them. See this thread:
http://lists.fedoraproject.org/pipermail/ruby-sig/2011-August/thread.html#597
Let's wait until next week for Mike. For now we have this "bundle
install in SPEC" approach, we will either
1) remove bundler
or
2) bundle install in post-install script
--
Later,
Lukas Zapletal | E32E400A
RHN Satellite Engineering
Red Hat Czech s.r.o. Brno
More information about the katello-devel
mailing list