Github user ottobackwards commented on the issue:
https://github.com/apache/metron/pull/814
@jjmeyer0 @cestella : Ok.
So, I added some code, where it is possible context wise to guard against
the case where there is a single variable as the check, and no default. This
resolves the test issues, save for one, where there is no default and we are
using == true. This is more difficult.
I feel as if all of these things are wrong, because the validation by
executing with null variables is itself logically wrong. We have discussed
this before, but this is really evident here. At least I feel it is.
I think that we should use compilation instead of validation. We can
decide to make it a separate option when running tests, or to make it the way
we do all tests etc. But if compilation does what I think it does, then I
think it is more correct.
I would like to do that, and remove the work around I have introduced here
( basically detecting that we are validating and have a single var that is null
because of validation ) since changing validation would be the real complete
answer.
---
