* Notification backends now handle body of notifications differently
* .. depending on their type (webhook, email, and pagerduty) are
currently the only three notification types that use body
* email and pagerduty expect a string
* webhooks expects a dict in string format
* RBAC relaunch 403 updates
Addresses 2 things
1. If WFJ relaunch is attempted, and relaunch is denied
because the WFJ had encrypted survey answers,
a generic message was shown, this changes it to show
a specific error message
2. Org admins are banned from relaunching a job
if the job has encrypted survey answers
* update tests to raises access pattern
* catch PermissionDenied for user_capabilities
* Add support for credential_type
* Finish up credential_type parameter with tests
* make inputs mutually exclusive with other params
* Test credential type with dict input
* Instance Groups Instances slider renders properly, and that list wraps properly.
* Instance Groups responds properly
* assorted container groups ui fixes
updated responsiveness of instance groups and instances list
fix layout of container group form
update help text for container group form elements
update text for tech preview top bar
* update container group doclink
* list styling updates based on feedback
a status of error makes more sense, because failed generally points to
an issue with the playbook itself, while error is more generally used
for reporting issues internal to Tower
see: https://github.com/ansible/awx/issues/4909
runner_on_start events have zero-length strings for their stdout
fields. We don't want to display these in the ui so we omit them.
Although the stdout field is an empty string, it still has a recorded
line length of 1 that we must account for. Since we're not rendering
the blank line, we must also go back and set the event record's line
length to 0 in order to avoid deleting too many lines when we pop or
shift events off of the view while scrolling.
When jobs are still processing events, the UI uses numerical ranges
based on job_event.counter instead of page numbers. We can't apply
search filters in this state because then there would be no way to
distinguish between events that are missing due to being filtered out
by search and events that are missing because they're still being
processed.
The UI must be able to distinguish between the two types of missing
events because their absence is presented differently. Events that are
filtered out by a search query have no visual representation, while
events that are missing due to event processing or other causes are
displayed as clickable "..." segments.
When the page loads, we want to retrieve and initially display enough
content for the scrollbar to show. If the very last page doesn't
have enough content for the scrollbar to show, the user won't be able
to scroll up to see more job history. To avoid this scenario, we always
fetch the last _two_ pages when loading the view.
