[TIP] Best practice for failing tests

[Chris Jerdonek]

On Thu, Mar 22, 2018 at 12:34 PM, Bruno Oliveira <nicoddemus at gmail.com> wrote:
>> Another reason is flaky
>> tests that sometimes fail and sometimes succeed, so that xfail()
>> wouldn't work as a substitute.
>
>
> Indeed, but in James' case I think `xfail(strict=True)` is the way to go,
> because it will *fail* the test if it suddenly starts passing because the
> fake now works as expected, which is a good thing because then that
> opportunity can be used to remove the marker. A skip will stay in there
> forever until someone takes notice by some other means.

Yes, I was going back and forth in my mind in my first reply to James,
and xfail() does have that advantage. For example, if the developer
maintaining the fake adds the functionality, you'd want to be notified
when it occurs as opposed to having to remember to keep checking.

--Chris



>
> Cheers,
> Bruno.