broken packages in unstable

[Helmut Grohne <helmut@subdivi.de>]

On Thu, Nov 22, 2018 at 11:20:25AM +0000, Holger Levsen wrote:
> On Wed, Nov 21, 2018 at 08:37:28PM -0700, Sean Whitton wrote:
> > What harm are the packages doing sitting in unstable?  Distributing them
> > does not have much point, but neither does removing them.
> 
> the rather few people working on (fully and partly) automated QA have to
> spend brain and cpu cycles on it

I guess I'm one of the people you're referring to here. So let me share
my experience:

For cross building I used to consider all of unstable. That turned out
to be painful indeed for exactly the reasons given above. Now I consider
those packages in unstable that have some version in testing and I no
longer have that pain. That seems to be a great filter.

I also performed one feature archive rebuild. Given the number of
temporary failures, I ended up looking at each failure (not just feature
failures) and filing the relevant FTBFS bugs. I regret not having
ignored packages not in testing.

What I found more annoying is packages being removed from unstable that
recently were in testing. I sent patches and the corresponding bugs
where closed by the ftpmaster removal. Not ftpmaster's fault.

Let me draw two conclusions:
 * Maybe we should remove stuff from testing even more aggressively to
   reduce the pain for archive QA?
 * If archive QA gets painful due to broken packages in unstable:
   Ignore those that have no version in testing. It's an easy filter
   with little misclassification.

Helmut