I'm not debating whether your test case is correct or not, or whether the
numbers are correct or not. The point is that 3 out of 4 people from mm list
who have looked at this patch have no idea what it's trying to do, why it's
important, or why these numbers *should* necessarily be wrong.
It's good that you have provided a way to demonstrate the effects of your
changes, but one can't solve cognitive gaps with testing. If one could, then
things like TDD would be effective on complex projects -- but they aren't[0].
We're really getting off in the weeds here, though. I just want to make it
clear to you that if you think "more testing" is a solution to kernel ailments
like this, you're going to be disappointed.
0: http://people.brunel.ac.uk/~csstmms/FucciEtAl_ESEM2016.pdf
