On Mon, Mar 21, 2016 at 9:20 AM, Peter Robinson <pbrobinson(a)gmail.com> wrote:
>> > Dne 17.3.2016 v 15:12 Dennis Gilmore napsal(a):
>> > > On Thursday, March 17, 2016 2:24:30 PM CDT Vít Ondruch wrote:
>> > >> Dne 17.3.2016 v 14:10 Dennis Gilmore napsal(a):
>> > >>> On Thursday, March 17, 2016 10:55:47 AM Miroslav Suchý wrote:
>> > >>>> I just discovered - by coincidence - that Koji is already
>> > >>>> using DNF
>> > >>
>> > >> for F24
>> > >>
>> > >>>> builds.
>> > >>>>
>> > >>>> So unless somebody object I will change mock default
config
>> > >>>> for F24
>> > >>
>> > >> to use
>> > >>
>> > >>>> DNF too.
>> > >>>
>> > >>> It was announced correctly[1] The change was made in rawhide 5
>> > >>> months
>> > >>
>> > >> ago,
>> > >>
>> > >>> there is no way we would change f24 back to yum after having
>> > >>> build
>> > >>
>> > >> rawhide
>> > >>
>> > >>> using dnf
>> > >>
>> > >> Please make sure that Koji as well as Mock are (not)installing
>> > >> weak dependencies.
>> > >
>> > > Koji is installing them.
>> >
>> > Yes, I know and I also know the mock does not install them by
>> > default. That was the point I wanted to made.
>> >
>> > > I have asked FPC to decide how they should be
>> > > handled.
>> >
>> > Any ticket number by a chance?
>> >
>> > > do note that yourchanges around rubypick have caused issues
>> > > because for soem reason dnf is aacting different on different
>> > > arches about
>> >
>> > it. so if
>> >
>> > > we enforce weak dependencies not being installed you will have a
>> > > lot
>> >
>> > of work
>> >
>> > > to fix up all the ruby packages
>> >
>> > I am fine with that work, since it is step in right direction ...
>> > and it was workarounded by Peter anyway.
>> >
>> > TBH, since it seems secondary arches are currently doing just fine
>> > without weak dependencies installed (due to some weird DNF bug),
>> > I'd say that it should be fine to not install them on primary
>> > arches.
>> >
>> We have taken steps to ensure weak dependencies get installed on
>> secondary arches
>
> the workaround for explicitly BR: ruby was applied by Peter to a subset
> of packages only and unfortunately the problem with weak deps still
> exists on s390(x). But it went away on aarch64.
In my experience you need to add deps for rubypick and rubygems to get
it to build. TBH I'm not sure whether it's a dnf issue or weird stuff
the ruby packager is doing.