Some users have been reporting some problems with 1.6.8 parsing, particularly over-zealous security checks against certain shell/command strings, in particular.
The development branch now contains fixes for nearly all of these issues we’re aware of, in addition to integration tests that cover and check for all of these items.
See test/integration/roles/test_{good,bad}_parsing/tasks/main.yml for details.
Feel free to submit pull requests to show cases that don’t work and add them to the tests - though this should be pretty comprehensive.
We’re planning on a dot release to close out the remainders soon, but want to hold off just a bit to make sure we’ve legitimately got things.
So if you’ve been having problems with parsing, check on devel, if you can.
There are some legitimate scenarios that we have to keep filtering, but these should be pretty minimal. They usually involve Jinja2 examples in Ansible playbooks in some ways that are not typically idiomatic, but obviously we’d like most of this to remain possible, even when it is discouraged for stylistic/readability reasons 