[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Is an MBF and unblock for packages introducing new files in /bin or /sbin or /lib in Bookworm acceptable at this stage?



On Mon, 22 May 2023 at 20:34, Sam Hartman <hartmans@debian.org> wrote:
>
> >>>>> "Luca" == Luca Boccassi <bluca@debian.org> writes:
>
>     Luca> On Mon, 22 May 2023 at 20:22, Sam Hartman <hartmans@debian.org> wrote:
>     >>
>     >> >>>>> "Luca" == Luca Boccassi <bluca@debian.org> writes:
>     >>
>     Luca> Hello Release Team, If we were to do a MBF against packages
>     Luca> that in _Bookworm_ have introduced new files in /bin, /sbin or
>     Luca> /lib*, would you accept the consequent mass unblock request?
>     Luca> I am asking beforehand as there's no point in going through
>     Luca> the effort if you don't, the advantage is only if we can sort
>     Luca> it before Bookworm ships, and the bugs would become invalid
>     Luca> and be closed as soon as it does as per moratorium otherwise.
>     >>
>     >> This sounds like a really bad idea.  While technically this is
>     >> consistent with the TC's advice, what you are proposing to do
>     >> increases the chance that you're going to trigger the dpkg
>     >> disappearing file bug.
>     >>
>     >> Consider:
>     >>
>     >> * User installs version from testing with file in /bin *
>     >> Maintainer quickly moves the file to /usr/bin per your MBF *
>     >> Bookworm releases; user does not upgrade at this point * Package
>     >> reorganization; file moves between packages * User upgrades; file
>     >> disappears
>
>     Luca> What "package reorganization" would that be? Are you aware of
>     Luca> any such thing happening in the next couple of weeks before
>     Luca> release?
>
> Who said anything about next couple of weeks.  This affects testing and
> unstable users *after the release*.  It is my experience of Debian that
> outside of freezes package reorganizations happen regularly.

So what you are worried is the combination of a testing installation
from~one year ago, that is otherwise never touched for say another
year, and also that has one of those 23 packages installed in the old
version, and also that same package of those 23 gets rearranged? That
seems vanishingly unlikely, and I am not sure how supported it is to
upgrade from "testing as bookworm" to "testing as trixie" without
passing from bookworm stable, given we don't support skipping releases
I'd have imagined that would not be supported either, but what do I
know.
Anyway, this is Helmut's request, so I'll leave the decision to him.

> I think your plan is safe for stable users if we actually find a
> solution to canonicalize paths during the next cycle.  I do not think
> your plan is safe for people who track testing, and for me there's not
> enough benefit to justify breaking testing.

Please stop saying "your plan", as I have already mentioned it was
someone else's request:

Kind regards,
Luca Boccassi


Reply to: