Notes for automated testing:

   morf upg -a ppc64le el7toel8 @noop -- --hostrequire '<key_value key="CPUMODEL" op="like" value="%POWER8E%"/>'

should get upgrade on suitable machine, although the require node here is wild guess (based on years old rhel6>rhel7 issue where we had to avoid powervm) and there's probably better way.

Alternatively,

   morf upg -a ppc64le el7toel8 @noop -- --machine suitable.machine.example.com

Adding --brew-task should work as well, updating the machine first with eg. Javier's build:

   morf upg -a ppc64le el7toel8 @noop -- --hostrequire '<key_value key="CPUMODEL" op="like" value="%POWER8E%"/>' --brew-task=23736760

I'm going to try it right now and also add test case to morf so that we can verify it later more easily.

Oh wait I did not realize --brew-task won't work because it will try to bolt the package on rhel7 but this fix is on rhel8 side.  I'll come up with something better.

Currently the status is:

 *  rhel-alt ppc64le - fixed  (old fails, new is OK)
 *  rhel     ppc64le - fixed  (old fails, new is OK)

For regression test, I also tried to run it with x86_64 but it looks like it's broken for another reasons which I still need clarified by Petr.

 *  rhel     x86_64  - ??? (blocked w/ another poblem that kills the test much earlier)

We can take care of verifying this in context of upgrade.  However, I presume RTT also have their regression testing process, so just let's make sure we're not avoiding that.

OK, so we've found out:  the repo provided by Javier was invalid; did not contain all files from the build (missing noarch).  Javier, did you use some Brew feature to create the repo?  If so, there could be bug to report.

Anyway, I'm going to find or create a repo manually and use that to re-run the suite.

(In reply to Alois Mahdal from comment #22)
> OK, so we've found out:  the repo provided by Javier was invalid; did not
> contain all files from the build (missing noarch).  Javier, did you use some
> Brew feature to create the repo?  If so, there could be bug to report.
> 
> Anyway, I'm going to find or create a repo manually and use that to re-run
> the suite.

The repo is valid. But may need to include both for x86_64 and i686 arches depending on which set of grub2 packages you have installed.

No, it is not valid Javier. There is not grub2-pc-modules-2.02-78.el8.noarch.rpm inside the repo for x86_64 that should be there.

Adding that one rpm extra into the transaction resolved whole problem.

Looking at compose, what rhel-8 normally includes in repo is x86_64 and noarch. (I can't see any i686.)  The repo is indeed missing at least this one particular .noacrh file (while at the same time other .noarch files were present).  It almost looks as if there was if there was some failure while copying the files.

Anyway, I've composed own repo (x86_64 + noarch) and I'm now running tests against it:

   https://liver3.brq.redhat.com/jenkins/view/OAMG/job/oamg-dispatch-adhoc/241/console

(Note that once we get the package in a compose or stage, we can---and eventually will---re-run tests again.)

So apart from some minor waivable issues (leftover packages etc.)  the above tests passed.  Ie. status:

 *  rhel-alt ppc64le - fixed  (old fails, new is OK)
 *  rhel     ppc64le - fixed  (old fails, new is OK)
 *  rhel     x86_64  - unaffected (old OK, new OK)

Thanks for the fix, Javier!


Rest is up to RTT, I guess.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2019:3587