Re: autopkgtest needs-recommends
On Thu, Jun 14, 2018 at 08:48:22AM +0200, Paul Gevers wrote:
> On 13-06-18 19:46, Niko Tyni wrote:
> > Couldn't you install the test dependencies just like now, then look at
> > the installed package(s) that '@' brought in and explicitly install its
> > recommendations in a separate apt run?
> > This only works for those needs-recommends tests that have '@' in their
> > dependencies, but I'd expect that to be the vast majority of cases.
> > It's certainly the case for the autopkgtest-pkg-perl ones.
> This is a variation of the 3.b./4.b. solution in my summary e-mail
> (copied below). I like the idea of limiting it even further than "all
> test dependencies". If we limit the scope of packages, I don't think it
> should be limited to '@' (albeit it being easier and less hackish to do
> just that, because expansion of '@' is already available after we merge
> your MR), but to all binary packages from the same source as the test. I
> forgot to mention in my summary e-mail that this would also solve issue
> 5, but I hope that was trivial to conclude.
Well, '@' is equivalent to 'all binary packages from the same source as
the test', right? Except that '@' guarantees that the actual packages
will get installed rather than some others that Provide the same name,
which we would actually need to happen here so we can look at their
Re limiting: I always thought (somewhat naively I guess) that
'needs-recommends' only means 'the recommendations of the package(s)
to be tested' rather than 'the recommendations of all the test
dependencies'. But I see the documentation doesn't really back that,
and neither does the traditional implementation.
In practice I'd be somewhat surprised if anything relied on the difference.
That's something that could be tested against the archive I guess. Just
needs somebody to do the work :)