* Stop using ansible.netcommon.next_nth_usable.
It requires netaddr to be installed before running the playbook, which isn't the case in CI.
This undoes cf61474c3b.
* Use own filter to avoid needing netaddr installed.
* Adjust deprecation versions.
* Remove redirects that are already made in ansible/ansible's ansible_builtin_runtime.yml
* Remove modules that were moved to the google.cloud collection according to ansible/ansible's ansible_builtin_runtime.yml.
* The _info module is in google.cloud.
* The gcp doc_fragment is a copy of the one in google.cloud and is only used by one lookup. Mark as deprecated/internal.
* Remove entries of modules that no longer exist.
* Update ignore.txt.
* Try to fix test.
* Remove debug output.
Setting the exit code in a trap overrides the exit code that caused the
trap to be called. This means if the test failed and called the trap,
the test will exit with 0 rather than the failure exit code.
* Don't die when docker CLI doesn't want to work with docker daemon.
* Allow more 'docker volume ls' failures
* More.
* Fix more failures.
* Forgot ignore_errors.
* Another one.
* More places.
* Try to install newer docker on RHEL8.
* Also restrict docker-ce-cli.
* Add Fedora 32 to CI; drop Fedora 30 which is soon EOL.
* Docker only supports Fedora up to 31; fortunately the 31 package seems to work for 32 too.
* Also skip for Fedora 32, as it apparently has a similar problem to Fedora 31.
ci_complete
