Merge ~peppepetra/+git/makefile_spec:makefile_spec into ~peppepetra/+git/makefile_spec:master

I've reviewed this change partially. I'll finish later today.

Added comments per our discussion with Jeremy this morning.

Few comments on the exclusions, maybe using .builds rather than 'build', though if it's a regex it'll probably cover the exclusion as is.

I'd be keen to see the clean remove any pyc files from the entire tree, as these often find their way into builds when we're using cp to make them and that'll cause all manner of weird failures in production.

Reply inline

minor nit on build step cp command.

There's some small diff comments for stuff that actually should be changed...

And then there's some larger comments to reflect recent discussion re: a few issues. I tried my best to represent the different points of view expressed.

Discussion desired, but be careful; this could easily devolve into bikeshedding. I'd like for opinions to be expressed, but ultimately we may just need a decision made for the sake of moving forward.

Be aware Openstack is not the only team we run infrastructure for anymore. K8s charms are increasingly common and they do not follow the Openstack naming conventions they use tox + pytest I believe with other test libraries. It's possible there is a convention from the Charmcraft team we could adopt?

I wouldn't put a lot of time/effort into trying to match every other team as long as we pick something and stick to it. There are several comments/discussions here that are kind of nitpicks and we should be mindful to not spend significant effort rearranging files with little benefit.

I don't want anyone to take that to mean 'leave it as it is' simply let's be mindful the amount of time we put into things based on the value they will bring.

some inline comments.

Added a few comments re: the discussion.

Replied inline to comments

One comment inline

More suggestions, re: "make clean" specifically. It likely affects all the Makefiles, but I just commented once.

Found another bit which I think ought to be changed.

Overall, it lgtm. However, I've added a question about using "git clean" passing "-f" twice instead of once.

I think this needs to be fixed; I encountered this on a charm which was being ported from nosetests to unittest discover.