Hello Pavel,
On Wed, Feb 14, 2018 at 10:18 AM, Pavel Raiskup <praiskup(a)redhat.com> wrote:
> On Tuesday, February 13, 2018 10:15:42 PM CET Michal Novotny wrote:
> > On Tue, Feb 13, 2018 at 9:51 PM, Michal Novotny <clime(a)redhat.com>
> wrote:
> >
> > > Hello,
> > >
> > > On Tue, Feb 13, 2018 at 12:54 PM, Michael Šimáček <msimacek(a)redhat.com
> >
> > > wrote:
> > >
> > >> On 2018-02-13 11:47, Pavel Raiskup wrote:
> > >>
> > >>> Sorry, I wanted to CC fedora devel before, forwarding.
> > >>>
> > >>> Pavel
> > >>>
> > >>> On Tuesday, February 13, 2018 10:54:55 AM CET Pavel Raiskup
wrote:
> > >>>
> > >>>> Because we are unable to find a consensus on implementation
details,
> > >>>> it's
> > >>>> likely we'll drop this feature from copr API and it will
be
> probably a
> > >>>> bit
> > >>>> more complicated to setup mock chroot for local tests in
future
> (you'll
> > >>>> need to have builder machine with copr-rpmbuild installed,
which
> brings
> > >>>> a
> > >>>> lot more runtime dependencies at least).
> > >>>>
> > >>>> From user perspective, do you mind if we dropped `copr
mock-config`
> > >>>> command?
> > >>>>
> > >>>>
> > >> I didn't know this command existed, but there were multiple times
in
> the
> > >> past where I wished something like this had been available (It
didn't
> exist
> > >> back then). It was usually situation like this: "Hi, I'm
trying to
> build
> > >> $package in $copr and it fails because of $build_tool that you
> maintain,
> > >> can you help me?". And since I had no idea how his copr was set
up,
> it took
> > >> me a lot of time before I was able to reproduce the problem. So, I
> would
> > >> find the feature useful, especially in instances outside Fedora,
which
> > >> usually have more complex configurations.
> > >> If it had to be dropped, I'd appreciate if copr could display the
> > >> configuration of given project for non-owners. That way it would be
> easier
> > >> to construct my own config, without trying to guess stuff based on
> the logs.
> > >>
> > >
> > > First, thanks for your input. This is very useful information for us.
> > > Next, I would like to ask if it was ok to put all the functionality
> about
> > > build-testing and building itself into just a single package:
> > > copr-rpmbuild. I think having things on just one place can help us
> focus on
> > > doing them really well and as the copr-rpmbuild tool is already
> responsible
> > > for building, I think it would be a perfect place to add additional
> > > build-debugging functionality like printing-out/dumping mock configs,
> > > enablement to run just a part of the build process, possibility to
> enter
> > > the build environment interactively etc. Would this be alright?
> > >
> >
> > I need to add that with this tool you really need to know _what_ you are
> > building to be on the safe side. It is similar to running rpmbuild
> locally
> > (unless you are really just dumping mock configs).
>
> I use 'copr mock-config' for working with buildroot, even if I don't
want
> to build anything (mock --shell). So except from mock, any other deps
> installed by copr-rpmbuild are useless and I don't want them on my box
> basically.
>
Alright, we can set some most prominent deps of copr-rpmbuild as weak deps.
It will be then possible to install the minimal package setup e.g. with:
# dnf -y install copr-rpmbuild --setopt=install_weak_deps=False
I opened
https://pagure.io/copr/copr/issue/222.
Thanks!
I wouldn't be able to install copr-rpmbuild on EPEL7 then, where I still can
install mock/copr-cli and I can experiment with copr bulidroot through `copr
mock-config` feature. IOW, enforcing user to install another client tool
for generating mock config doesn't make sense.
Pavel