paul at nohats.ca
Sun Feb 3 02:29:50 UTC 2019
On Sat, 2 Feb 2019, D. Hugh Redelmeier wrote:
> | It’s a bug but an old one. It might be triggered now with some secrets file install change
> So: is this test known to fail?
For this one I am not entirely sure. For some others we do know that
they are failing because work needs to be done.
> If it is, it is a waste of time and a cognative burden for each person
We have gone through a few rounds of discussions on this. There are no
> - we need an announced category of "expected failures"
It won't help, because people will always ignore it and never fix it.
> - we need reference logs reflecting such bugs so that when something
> changes in the output we get a signal.
If we ship with known bugs, we can have the test system show them as
> - if the expected failure creates non-deterministic results, we need
> to declare that too
non-deterministic is something no one wants. It just happens due to a
number of known and unknown reasons.
> The alternative is to ignore any test that failed "before". That's
> negligent, but it's what most of us do.
Yes, there are no good options, other than getting more resources to fix
More information about the Swan-dev