Ben/wip/gha fedabipkgdiff

[woodard]

Try a second time to add some fedabipkgdiff tests.

I have this on my reading list to work out the problems.

[vsoch]

@woodard I can take a look / debug this to help tomorrow!

[vsoch]

@woodard I don't understand what you are trying to do - I wouldn't bring in that wrapper around essentially running the same thing that was already there. It adds complexity and only saves a little time of not installing the package. I would restore tests to how they were before.

[woodard]

So far what I have been doing is mostly diagnostic and I expect once, I really get a clear understanding of the problem then I will back out most of these "hammer" changes. If you look at 4c3bd58 463abb0 I tried to make sure that the dependencies are there. By putting in hard paths, I was able to tell that there is some sort of path problem. The python-unversioned-command rpm should provide a /usr/bin/python -> /usr/bin/python3 link.

So right now I can see that I have two problems:

1. the path isn't setup correctly, I suspect that I will also need to add a LD_LIBRARY_PATH as well. I am not sure but I think that this will need to be setup in some sort of env block in the yaml.
2. That got me to the place where something is missing that provides the koji config for fedora. I haven't figured out where that is coming from yet. My system seems to have it while the container does not. I need to figure out where that is coming from. There is the koji.read_config on line 51 of fedabipkgdiff, I need to track that back.

[vsoch]

If you rely on Koji it means you can't extend this testing to another OS, something to think about.

Here is someone else setting it up in GitHub actions if it helps! https://github.com/containerbuildsystem/koji-containerbuild/blob/master/test.sh

[woodard]

Relying on koji in this case is not a problem. We are simply using Fedora to provide a broad swath of code written in different styles and in a handful of languages as handy test suite of real world code to verify that libabigail is operating correctly. It is sample data.

Once libabigail operates reliably on fedora as a set of sample data, then I would like to introduce additional sample data. In particular, IIUC you have a build of E4S built with -g and thus with DWARF. This will include code from several compilers that are not represented in the fedora sample data and some of the HPC codes are extremely old or extremely cutting edge both of which can be challenging for libabigail.