* Change handling of error cases to global post_run_hook
* handle license errors correctly again
* Fix some issues with line ordering from the custom logger thing
* Remove debug log statement
* Use PermissionDenied for license errors
* More elegant handling of line initialization
Update tests to new exception type
Catch all save errors, fix timing offset bug
Fix license error handling inside import command
* proot now enabled at task-level
since tasks are no longer calling
awx-manage (which would set up its own proot)
* dropping proot env var since it's not
relevant to the test
* noting that the inv update task only uses the
inventory update management command to
save the inv to the database
(it doesn't do the work of fetching hosts / groups)
- in the past, inv. update jobs called `awx-manage inventory_update`
which took care of setting up process isolation
- at this point, though, inv. update jobs call runner / ansible-inventory
directly, so we need another way to put process isolation in place
- thankfully, there was already support for providing process isolation
for other types of jobs (namely JT Jobs, Project Updates and Ad Hoc
commands)
- so, we do what those other jobs do and override the stub for should_use_proot
(which by default returns false) so that it keys off of the
`AWX_PROOT_ENABLED` setting
* perform_update can be called from either awx-manage
or the RunInventoryUpdate task
* need to make sure that the inventory updates
that happen with perform_update are atomic
This commit makes the needed changes to inventory update
post_save_hook logic so that the historic log lines that
inventory updates write will be written to stdout,
but this hack bypasses the ansible-runner verbose event
logic and dispatches verbose events directly.
Fix the venv application with the ansible-inventory system
(note: much of this is undone in a later commit)
Deal with some minor test updates for
the ansible-inventory interface changes
