[csw-maintainers] stable release?

[James Lee]

On 16/10/09, 18:06:07, Dagobert Michelsen <dam at opencsw.org> wrote regarding
Re: [csw-maintainers] stable release?:

> I can't remember that we even have defined this is,

Yes we defined it.


> besides
> "Q/A", but what specifically? I know James has an infrastructure
> with cleanroom-install-depend-checks, ldd-inspection of libs and
> stuff. It would be good to formalize as much as feasible what a
> package must actually conform to to be good for stable.
> Currently at least I don't know how many, let alone what packages
> are really broken. If James already has all the tests we
> should apply them routinely on the farm for all packages.
> And: no, "somebody must have used the package" is not the
> only criterium for stable.

It's not far off.  The criteria are:

  * It installs

  * It works

  * It can be removed

No surprises there and nothing too taxing I trust.  "It works" is
difficult to judge and you are trying to apply a automatic test which
isn't possible.  What is possible is to test for the reverse state
and ensure it doesn't exist.  For this there are some automated
checks.

Then the are inferences from these such as "can be upgraded from
previous", (means previous installed and not the previous version),
which should work according the above if the previous can be removed.
Doesn't clash with another packages.  Blah, blah, I think it's been
said before.

Note the QA is of the packaging and not of the underlying software.

Note it's possible to have 2 perfect packages that don't work
together.




James.