* **Fix bug.** If a cloudformation stack is updated with a new
parameter, that parameter should be honored. Several unit tests
had bugs where they were not providing parameters required by the template.
* **Fix bug.** Do not update stack parameters until after deleting removed
resources, so that any references to removed parameters can be resolved.
* **Fix bug.** Per the API, creation of a change set should not modify
a stack. The `diff` method, called in the creation of a
FakeChangeSet, was mutating the resource map which was problematic
This fixes#4141, and pave ways for future changes around changeset.
We had subclassed FakeChangeSet from FakeStack, not from BaseModel. This made
us easier to send the response for describe_change_set calls, but when we are
handling the details of change set, the old approach won't work at all.
For example, when we were creating a changeset, we were actually creating a
stack without registering it (self.stacks), and future update onto this stack
is not really possible.
Signed-off-by: Kai Xia <kaix+github@fastmail.com>
* Added explicit exception raise when no stack found.
Currently, any operation that uses 'get_stack' method from 'CloudFormationBackend' class
will fail with AttributeError or jinja2 exception if ran against non-existing stack(created/deleted)
To fix the issue I explicitly raised a 'ValidationError' exception.
Added tests for boto and boto3 responses.
* Moved non-existing stack tests to 'test_stack_events'
When using 'update_stack' to test raising an exception when the stack doesn't exist
test coverage dropped by 0.5%. I am using stack_events instead.
* Removed some unreachable paths
After adding the exception couple of paths in the code are unreachable as 'get_stack' doesn't return 'None' anymore.
This is the reason why coverall was reporting decreased coverage.
* Removed an unreachable path I missed
* Added couple of tests in cloudformation/models
* Added more assertions around raised exception
* Formatted document using black to fix issue with travis.
