Fix decoration spec #895
Fix decoration spec #895
Conversation
|
Looks like the integration specs are flaky, gonna check. |
|
Right, so they weren't flaky. Ruby 3 doesn't bundle the webrick gem anymore, so starting a server without mentioning it in the Gemfile wouldn't work. Also, newer versions of Rails need an activestorage configuration in order to work. Added those. |
tbuehlmann
force-pushed
the
fix-a-failing-spec
branch
3 times, most recently
from
072da06 to
f783d97
Compare
|
Righty, but I found a real flaky spec now. Could reproduce with |
Draper::CollectionDecorator#{is_a?,kind_of?} triggers
decorating the object. When stubbing a method call
on a CollectionDecorator, RSpec itself will call #is_a?
on the object. See:
https://github.com/rspec/rspec-mocks/blob/ea4d8ea4532480500462cc7a633d3408656a4dcd/lib/rspec/mocks/proxy.rb#L38
That's why the object will always be decorated when
stubbing a method on it.
We can circumvent this issue by directly checking
the definition of @decorated_collection.
Newer versions of rails need a config/storage.yml in order to work.
Ruby 3 no longer bundles the webrick gem so we need to add it to the Gemfile.
When decorator_generator_spec.rb:51 was performed before or
after another spec (not sure), it would fail. It has to do
with requiring some rails constants.
To reproduce, run:
bundle exec rspec spec/generators --seed=53929
Fixed by correctly requiring other files.
tbuehlmann
force-pushed
the
fix-a-failing-spec
branch
4 times, most recently
from
dc51c26 to
490f734
Compare
Simplecov doesn't support Ruby 2.4 in newer versions and the latest possible version generates reports which are not processable by the code climate reporter. So it seems we're stuck with simplecov 0.17.1 until Ruby 2.4 is dropped: codeclimate/test-reporter#413 (comment)
tbuehlmann
force-pushed
the
fix-a-failing-spec
branch
from
490f734 to
51e9e21
Compare
|
@mgerst: Finally fixed it to green, but had to fixate simplecov to an old version. |
|
@codebycliff I think this one can be merged then #897 Then we'll have a green CI ready to be tested against Ruby 3.0 \o/ |
|
Also it could be great to have appraisal to test against different versions of Rails. wdyt? |
|
I think that would be awesome if someone wants to whip it up. We should also test against Ruby 3, but didn't want to make that change as part of the CI change. |
I can do a PR for that. |
Draper::CollectionDecorator#{is_a?,kind_of?} triggers
decorating the object. When stubbing a method call
on a CollectionDecorator, RSpec itself will call #is_a?
on the object. See:
https://github.com/rspec/rspec-mocks/blob/ea4d8ea4532480500462cc7a633d3408656a4dcd/lib/rspec/mocks/proxy.rb#L38
That's why the object will always be decorated when
stubbing a method on it.
We can circumvent this issue by directly checking
the definition of @decorated_collection.
Edit:
This PR got bigger than intended, but it fixes more failing and flaky specs.