Ruby on Rails

Most of the tests in validations_tests take an existing model, give it a new validation and then do some asserts to make sure that the validations work. There is code in setup to blank out the validations for Topic (the model used in most of the tests), however, other models are used and they are not being revertted.

The problem is that any test that is run after validations_test may encounter problems where models can't be saved because of the validations added by validations_test.

Also, the current method of simply zero-ing the validations for Topic before each test isn't the best case for this because Topic does have some default validations (has_many :replies adds a default validate_associated_records_for_replies), so the state of the Topic model going into these validations_tests is wrong. It's also wrong after the last validation_test as either anything that happens inside that test will not be reverted or the default Topic validations are not reset.

The attached patch provides a repair_helper for activerecord tests that currently knows how to:

a) repair validations: will record the validation state of a model class, execute the block and then reset the validations afterward.

b) repair default error messages: will record the default error messages inside ActiveRecord::Errors, execute the block and then reset the default error messages.

These can be called with a block inside a test, or without a block in the context of a test case to supply default setup / teardown callbacks to do the same.