Modernize the acceptance tests

[ekohl]

This takes the lessons from rubocop/rubocop-rspec#1231 and applies them here.

[ekohl]

This here is to combine things with .and since you can't use .not_to().and().

[ekohl]

Updated to top the Debian backport acceptance test. See that commit for details.

[ekohl]

I had one copy-paste error where .not_to was replaced by .to. Fixed that.

[pirj]

This line confuses me. As mentioned in the comment above - isn't not_to ...and disallowed in RSpec in some way?

[pirj]

Alright, here it is:

        `expect(...).not_to matcher.and matcher` is not supported, since it creates a bit of an ambiguity. Instead, define negated versions of whatever matchers you wish to negate with `RSpec::Matchers.define_negated_matcher` and use `expect(...).to matcher.and matcher`.

I'd suggest either matching against a negated predicate if it exists (disabled? and stopped?) with a regular .to runner:

specify { expect(service(redis_name)).to be_disabled.and be_stopped }

If those predicates are not available on what service returns, define negated matchers for be_running and be_enabled.

Or just plain split this expectation in two:

specify do
  expect(service(redis_name)).not_to be_enabled
  expect(service(redis_name)).not_to be_running
end

Whatever you like the most.

[ghoneycutt]

Two different tests is much easier to read and maintain.

[pirj]

I suppose you meant two spec examples.
Sure, I agree, unless expensive setup is involved, which I've heard, is sometimes the case with serverspec.

[ekohl]

You're absolutely right that this line is vague and confusing.

It is very common to always want it both enabled and running or both disabled and stopped. I've been wondering if it makes sense to have some short hand for it, but I struggled to find a good name for it.

Looking at https://serverspec.org/resource_types.html#service there is no be_disabled nor be_stopped. I'll think a bit more about this.

[ekohl]

For now I went with the recommendation to split it in two statements. That was the easiest.

For example, not_to be_running is clear. But is to be_stopped the same? to be_stopped could imply it should be present but stopped while not_to be_running doesn't care whether the service is present or not. Splitting it in two avoids these issues.

[ekohl]

I think this should be good to merge. I'm leaning to dropping Archlinux from the metadata for now since the tests obviously fail. Also #439 suggests it's entirely broken at the moment.

[bastelfreak]

I'm fine with dropping Arch Linux from the metadata.json for now, or merging this as is.

[ekohl]

That would be #441. I still think the changed in this PR that relate to Arch are good to keep in. Those are in separate commits so easy to distinguish.

@pirj do you think this is now all good to merge?

[vox-pupuli-tasks]

Dear @ekohl, thanks for the PR!

This is Vox Pupuli Tasks, your friendly Vox Pupuli GitHub Bot. I noticed that your pull request contains merge conflict. Can you please rebase?

You can find my sourcecode at voxpupuli/vox-pupuli-tasks

[zilchms]

2 year ping @ekohl :) do you want to rebase this to the current master?

[ekohl]

Rebased to resolve merge conflicts. I didn't run it locally yet, so it may have broken something in the rebase.

[zilchms]

I dont know if we want to fix the legacy facts in this PR or do all legacy stuff in a second one. Additionally we should look at why the Puppet 7 on Centos 8 test times out.
Otherwise this looks good :)

[ekohl]

Additionally we should look at why the Puppet 7 on Centos 8 test times out.

It's common for the CentOS 8 tests to time out. I suspect pulling from quay.io may cause problems.

[ekohl]

Updated to drop legacy facts in the test suite.