ci(travis): reduce matrix down to 5/6 instances (ref: #118)

[myii]

• The selected 5 in a "T"-shaped matrix:

  - 2019.2 (py3): debian-9   fedora-29   opensuse-leap-15
  - 2018.3 (py2):           ubuntu-1604
  - 2017.7 (py2):             centos6

• Covers each os, each Salt version and both Python versions

---

I was going to go for centos-7 and fedora-28 instead but centos-6 is a "special" platform, which always throws up interesting situations (including not being systemd-based).

[myii]

So initial results from my fork:

Instances	Running time	Total time
15	10 min 8 sec	31 min 36 sec
5	4 min 12 sec	13 min 10 sec

[aboe76]

good idea to have different init systems as well, good point.

[myii]

Thanks for the reviews @aboe76 and @n-rodriguez. What do you think @javierbertoli and @daks? Can we get a merge on this so that we can start rolling out this change to the (semantic-release) formulas out there?

[javierbertoli]

Wouldn't centos-7 (yum-based installer), cover more cases (centos, redhat, fedora) than fedora (which installs using dnf)? @aboe76 and I suggested using it a few days back.

[myii]

@javierbertoli BTW, you've pinged someone else there, not @aboe76! Anyway, like I said at the top, I was going to do it using centos-7 but I thought it would be good to have a non-systemd system, to cover more scenarios. I'm relaxed either way.

[javierbertoli]

@javierbertoli BTW, you've pinged someone else there, not @aboe76!

Ooops, fixed the typo 😋

Anyway, like I said at the top, I was going to do it using centos-7 but I thought it would be good to have a non-systemd system, to cover more scenarios. I'm relaxed either way.

Perhaps I was not clear, I meant replacing fedora-29 with centos-7, not dropping centos-6.

The final matrix we suggested was more on the line:

• 2019.2 (py3): debian-9 centos-7 opensuse-leap-15
• 2018.3 (py2): ubuntu-1604
• 2017.7 (py2): centos-6

centos-7 and centos-6 differ enough as to make it reasonable to test them both.

If the argument to add fedora is the same as to add opensuse (a RPM-based distro, different package manager), then I'd say add it to the matrix, ie, 2018.3 (py2) fedora-28?

Perhaps I'm just nitpicking, so I'll 👍 whatever you decide.

[myii]

@javierbertoli This sounds like the ideal solution, 6 in a tree instead of 5 in a T(!):

  - 2019.2 (py3): debian-9   centos7   opensuse-leap-15
  - 2018.3 (py2):     fedora-29   ubuntu-1604
  - 2017.7 (py2):            centos6

[myii]

So the updated result from my fork (continuing the table above):

Instances	Running time	Total time
15	10 min 8 sec	31 min 36 sec
5	4 min 12 sec	13 min 10 sec
6	6 min 8 sec	15 min 13 sec

[javierbertoli]

Thanks @myii for your hard work, and all the rest for your feedback!

[myii]

You're welcome, thank you all for your comments, reviews and feedback. Time to start reducing our Travis burdens in the other formulas!

[saltstack-formulas-travis]

🎉 This PR is included in version 2.1.12 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀