…implecov's dir name Otherwise, the test runners overwrite each other's coverage data, producing false negatives. TEST_ENV_NUMBER is the variable used in parallel_tests for database names, etc.
…helper Instead of playing whack-a-mole with new warnings every time someone ships new code, this simply removes anything originating in gems or the stdlib.
For a few tests, we shell out to `ruby -e "require 'fakeweb'; ..."` to get a clean environment in order to check what warnings appear when you combine FakeWeb with other libs. Recently these processes starting printing coverage reports, which messes with our assertions about the warnings output. This sets the SimpleCov formatter to `MultiFormatter` for those processes, which keeps them collecting/logging coverage data while suppressing any output. p.s. this also sneaks in a more robust technique for checking whether the full suite is being run (in #running_all_tests?).
This reverts part of commit 3b78bb4.
… the whole suite
* myronmarston/nil_response_body: Return nil when the body is stubbed as '' or nil. Conflicts: test/test_fake_web.rb
This includes a quick extract-method refactor (well, extract-task) so you can run the preflight checks without running all the tests.
Also, move that method out of the test file.
Some tools like RVM default to `rake test`, so it's easier if that task does the right thing. Also, it's nice to be able to refer to the main task by name so you can use it as an argument on the command line, etc.