Final Neon Support Fix

[s0undt3ch]

What does this PR do?

See title.
Additionally:

• Drop support for Fedora < 30
• Drop support for Debian < 8
• Adds preliminary support for Fedora 32, although disabled since it defaults to Py3.8 which salt doesn't yet support
• Locks kitchen config for Arch Linux to python 2 since Arch is also on Py3.8

What issues does this PR fix or reference?

Closes #1395
Closes #1400
Closes #1423
Closes #1381

[myii]

@s0undt3ch As we discussed in #formulas a couple of days back, the fedora-31 image build that was failing then is now passing. To be completely sure, I'm rerunning all of the jobs again anyway:

• https://travis-ci.org/myii/salt-image-builder/builds/644860865

[s0undt3ch]

@myii Thank You! Please report back any problems.

[s0undt3ch]

@vutny see anything requiring changes?

[vutny]

Wow, this PR is huge!
I was unable to spot anything critical, and the Drone looks happy.
Well done, @s0undt3ch . Thanks.

[myii]

@s0undt3ch Just one note for you, which appears to be a regression:

• https://travis-ci.org/myii/template-formula/builds/644862005

These were all passing before this set of changes (not just this PR but the last batch of development). It no longer allows us install the latest master on top of an existing pre-salted image.

However, I wouldn't like to see this PR blocked, though -- the current bootstrap script needs to be fixed since it's broken. Hopefully this can be addressed in another PR.

[myii]

@s0undt3ch That last link was the wrong one. I've redone the comparison, before and after:

• https://travis-ci.org/myii/template-formula/builds/645908860 -- all failing
  • This is using the version of the bootstrap from this PR's branch
• https://travis-ci.org/myii/template-formula/builds/645913121 -- all passing except arch-base-latest
  • This is based on a commit before these recent developments
  • https://github.com/saltstack/salt-bootstrap/blob/7b192471e63174dc6cfc26369d8b17b4d600abd0/bootstrap-salt.sh

All of the above are failures to build or find pycrypto.

[myii]

@s0undt3ch Just had a bit more time to test and I've established that the pycrypto regression was introduced by this commit: 39ed3f8.

[s0undt3ch]

The upcoming Neon release, salt's master branch has now defaulted to setuptools.
With that change, and because we hit an issue when testing the package build process, we reverted back to using pycrypto, for which there isn't a wheel available.
As such, we must build it from source, which seems to work (nevermind the opensuse, that a bad github actions workflow issue).

This is the whole reason why git bootstraps post Neon(which include Salt's master branch) are now done without system packages and only using pypi packages.
When I get some more time, I'll look at your CI pipeline to find out why its failing.

[myii]

@s0undt3ch OK, thanks for the explanation. We'll just avoid using the bootstrap on top of the pre-salted images for the time being.

[s0undt3ch]

By pre-salted you mean salt has been installed before and it's not the master branch, correct?

[myii]

We're building pre-salted images to use with SaltStack Formulas, for numerous platforms and all supported releases of Salt + master. This is the repo:

• https://github.com/netmanagers/salt-image-builder

To ensure our master testing is the very latest, we use the bootstrap on top. We'll just avoid that going forward; we've got a Travis cron job that runs every week anyway, so it will be pretty recent in any case.

[s0undt3ch]

Ok. Thanks for all your help!

[myii]

You're welcome. Any idea when this PR will be merged?

[s0undt3ch]

As soon as drone passes, its going to be merged and a sable release will be done right after.

[myii]

Excellent, appreciate all of the work you're putting in, @s0undt3ch.