Mock, Rawhide and DNF

Clark Williams williams at
Mon Feb 16 20:57:45 UTC 2015

On Mon, 16 Feb 2015 16:05:35 +0100
Kevin Kofler <kevin.kofler at> wrote:

> Miroslav Suchý wrote:
> > On 02/14/2015 04:33 AM, Kevin Kofler wrote:
> >> IMHO, it is the job of the EPEL package of mock to disable the
> >> config_opts['package_manager'] = 'dnf' option in the shipped version of
> >> the Rawhide configs.
> > 
> > But then you will not be warned that your results may differ from Koji.
> Surely that's a minor issue compared to the builds failing entirely! We need 
> to ship configs that work out of the box. If we cannot provide DNF for EPEL, 
> the EPEL-hosted configs must not require it.

To make sure I understand the issues here, let me restate (or "sum up"
as Inigo Montoya would say):

1. DNF is the preferred package management tool for F22+. 

2. The EPEL mock configs for RHEL6 and RHEL7 hosts will *never* have dnf
available and must use yum to build packages. 

3. Anyone building an F22+ package on a RHEL{6,7} system will have
different depsolving and possibly different results than someone
building the same package on a Fedora system.

If the above are correct and the only issue here is that we have
different depsolving with yum versus dnf, let's detect that we're
building a Fedora package on EPEL and print a warning to that effect. I
don't think it does us any good to ship a configuration that won't work
(i.e. setting config_opts['package_manager'] = 'dnf' on an EPEL host).
It's either that or fail the build immediately and I'm more in favor of
printing a warning and continuing. 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <>

More information about the devel mailing list