Re: Aren't there any checks in place to prevent a package from becoming uninstallable?
On 2011-02-27, Paul Wise <email@example.com> wrote:
> On Sun, Feb 27, 2011 at 4:02 PM, Lars Wirzenius <firstname.lastname@example.org> wrote:
>> We don't care if something is temporarily uninstallable in unstable. The
>> only way to prevent that from happening would be to keep packages from
>> entering unstable unless all their dependencies are in unstable already,
>> and that would prevent bug fixes from coming into unstable faster. This
>> is important because a source package might produce several binary
>> packages, and some of them might both be fixing bugs and be
> Something that might work would be to keep the old source/binary
> packages around (as well as the new ones) until nothing depends on
> them. IIRC the release team have the ability to (temporarily) have
> multiple versions of a source package in testing, perhaps something
> like that could be added for sid.
What we do for some transitions is editing the source package a library package
comes from, so that both library revisions (libfoo1 and libfoo2) are considered
eligible to stay in testing at the same time. This eases the pain of some
transitions. For that to work in sid, as library removal is manual anyway, you
need to convince ftp-masters not to remove them when there are still quite a
bunch of reverse dependencies. Given that NBS removals should show the rdeps
too, it'd only be a policy decision.