On Tue, Feb 21, 2017 at 02:15:08PM +0100, Jürgen Spitzmüller wrote: > Am Dienstag, den 21.02.2017, 12:57 +0100 schrieb Kornel Benko: > > Sure, because 'inverted' means "invert the test result". > > This means, the tests actually passed. This was previously not the > > case. > > I see. Thanks.
For exports that fail, we do not want to keep them as is because the output of running all the tests will be confusing (e.g. "wait, did that test fail before or is it a new failure?") and it will be more difficult to see what is a regression. If we just omit the failing export from running (which we can do, see ignoredTests file), then we won't know if the export goes from failing to passing. By "inverting" it, we solve this problem: the test technically passes (although the name indicates that the export actually fails) so it does not complicate the output, and if the inverted test "fails", we will see it because it shows up in our failing test results. Thus, once we see that we cannot easily fix the underlying cause of a failed export, we document why the test fails and invert it. Scott
signature.asc
Description: PGP signature