Filled https://pagure.io/fedora-infrastructure/issue/6111.

That sounds like a bug in autoconf (or we need special patch for Fedora)....

config.status which is generated using autoconf contains:
...
S["target_alias"]=""
S["host_alias"]="ppc64le-redhat-linux-gnu"
S["build_alias"]="ppc64le-redhat-linux-gnu"
...
host_alias='ppc64le-redhat-linux-gnu'
host='powerpc64le-redhat-linux-gnu'
host_os='linux'
build_alias='ppc64le-redhat-linux-gnu'
build='powerpc64le-redhat-linux-gnu'
build_os='linux-gnu'
...
S["host_os"]="linux"
S["host_vendor"]="redhat"
S["host_cpu"]="powerpc64le"
S["host"]="powerpc64le-redhat-linux-gnu"
S["build_os"]="linux-gnu"
S["build_vendor"]="redhat"
S["build_cpu"]="powerpc64le"
S["build"]="powerpc64le-redhat-linux-gnu"


So either RPM should somehow replace powerpc with ppc or autoconf should put ppc64le instead... I think autoconf does its job right, but RPM should do replacement.

I fail to understand what's the problem.  Is this because fedora rawhide has
outdated config.guess in redhat-rpm-config?

(In reply to Pavel Raiskup from comment #3)
> I fail to understand what's the problem.  Is this because fedora rawhide has
> outdated config.guess in redhat-rpm-config?

Problem is that rpm on ppc64le thinks that _host_cpu is powerpc64le rather than ppc64le and when you have ExclusiveArch: ppc64, COPR fails to build package because architecture is not included.

I'm not sure why it is happening, just noted that macros is generated out of config.status and that one has powerpc64le rather than ppc64le.

Btw. the _host_cpu is baked into rpm during build of rpm, and that info is
taken from config.status, which comes from 'gnuconfig' upstream.
I checked some quite old fc22 build of rpm and the %_host_cpu was also set
to powerpc64le.

The %_target_cpu is however by default set in mock configuration, from
config['target_arch'] by default.

I meant s/config.status/config.guess/ of course.

autoconf is calling it all powerpc*-* so that ppc64* is done by rpm.
So I believe it is up to rpm to fix that.

Hms, rather than doing some substitutions on rpm side I guess we should invent
something like %power64le macro to respect gnuconfig opinion about the naming;
if that was the issue in the end.

Anyways, after discussing this with Igor, turns out this is related to
bootstrapping chroot (new feature of mock), where Igor has overwriten rpm
package.  So sounds like NOTABUG at this point.

By default, mock respects the 'target_arch' config option, so that the rpmbuild
is run like 'rpmbuild --target ppc64le' which in the end overrides the default
%_target_cpu macro contents (rpm compares ExclusiveArch contents against
%_target_cpu).

This bug appears to have been reported against 'rawhide' during the Fedora 27 development cycle.
Changing version to '27'.

FTR, this is eventually being handled here:
https://github.com/rpm-software-management/mock/pull/1195

Hmm, this rings a bell... https://github.com/rpm-software-management/rpm/issues/791