By not relying on the foreman class parameters, it becomes easier to use
this class without Foreman. This can be useful when deploying it using a
git checkout rather than packages.

This adds support for keycloak when using the
keycloak-httpd-client-install command. A more native integration would
rely on the oidc support in puppetlabs-apache, but that's unreleased.

It drops the dependency chaining for the passenger package because it
was leading to dependency cycles and I don't see a reason for the
explicit chaining.

While Puppet is case insensitive on data types, Kafo does care. This
matches the type to its actual definition.

This sets the config file mode to 0640 by default because they may
contain secrets such as credentials. To keep it readable for Foreman,
the group is modified to $foreman::group.

Previously this was duplicating the plugin prefix for packages that
didn't follow the convention. By using regsubst, this can be avoided.
This removes the need for separate params classes.

0e86214 added this, but was a left over
from an earlier iteration.

Various changes have gone in that make this a new major version.

When using systemd socket activation, it's important that the
ListenSocket matches what Puma binds on. Otherwise it may fail. This
happens when is configured on [::]:3000 (dual stack) and Puma on
0.0.0.0:3000. Puma will then attempt to bind and fail because the port
is already in use.

The service bind is now made explicit because systemd's
ListenSocket=3000 binds on :: where Puma by default binds on
0.0.0.0:3000. This is IPv4-only, but is what was done prior to this as
well. Apache is configured dual stack and the recommended deployment.

281f196 introduced the systemd socket
override, but used Service instead of Socket. This causes the service to
bind incorrectly.

281f196 introduced the override with
the incorrect section. e642db1
corrected the section, but not the headers.

8bc1ba3 made the default deployment use
the Puma service. That means the default requires systemd and a soft
dependency no longer makes sense.

It also updates the documentation in README to match.

These config settings are never actually read by anything.

Parameters with a single value aren't really useful.

Foreman 2.0 dropped this and needs to be removed.

Since 8bc1ba3 he default deployment is
now a reverse proxy so this test is redundant.

All testing is now happening on a reverse proxy so these are no longer
needed.

This is the default deployment we support and test with.

Note that it disables EPEL which we no longer need.

This avoids generating certificates for every example and makes the
examples easier.

This removes the repo parameters from the main class, in favor of a
standalone class that can be included. It uses an anchor because that
can be collected in the main class to keep the correct dependency
chaining while using composition.

875c837 started to use modern facts.
However, for these branches there is no coverage and facterdb can't find
these modern facts. This causes issues in puppet-puppet which does call
these branches. Ignoring lint is the quickest fix now.

Longer term these classes will be split out to its own module as
theforeman/puppetserver_foreman so there can be test coverage.

This fact is a core fact that's only set on AIO installs. Setting this
custom fact is also easier than the structured fact in our test suite
which increases reliability of our tests.

This allows to make use of redis as backend, but still defaults to file.

This removes duplication from the examples and allows it to only show
what's really needed. It currently relies on my git fork of beaker-hiera
since that's archived.

It also updates to PostgreSQL 12 on EL7 since Rails 6 now needs that.

Previously foreman_config_entry could be executed before the database
was available. This ensures the database is ready and the plugin
installed before attempting to change a config setting.

This prevents a collision with the ipa fact from the ipa module.

6.x was breaking since it made repository management modules soft
dependencies. Since we don't use these, it's not breaking here.

6.x was breaking since it made repository management modules soft
dependencies. Since we don't use these, it's not breaking here.