Bug#983233: New autopkgtest shouldn’t trigger a regression

[David Prévot <taffit@debian.org>]

Le 21/02/2021 à 16:02, Paul Gevers a écrit :
> Control: tags -1 moreinfo
>
> Hi David,
>
> On 21-02-2021 12:53, David Prévot wrote:
> > I recently added an autopkgtest to a package, and the autopkgtest failed
> > on all suites. I’m surprised to see that failure considered as a
> > regression (#983211)
[…]
> We realize that `regression` doesn't quite cover how we use it
> […] we consider
> failing tests in testing RC [1]. Hence, allowing a package with a
> failing test to migrate to testing would immediately make it RC buggy,
> hence we block it. `regression` doesn't fully cover it, we read it as
> "there wasn't an RC bug in testing, and now there will be, hence
> regression" but it's a bit stretched.
>
> Do you have a suggestion to use there instead (a word, not a sentence)?

Maybe “failing test” since you used it twice to describe the actual 
issue (but that’s two words) or even simply (test) “failure”?

The current policy makes it almost worth to remove autopkgtest in order 
to avoid what you call “regression”. Yet, such removal would check all 
boxes of what is called “regression” according to some some online 
[dict]ionary: “a return to a previous and less advanced or worse state, 
condition, or way of behaving”.

  dict: https://dictionary.cambridge.org/dictionary/english/regression

Regards

David