Re: WARNING: libexpat1 1.95.2-2.1 breaks dependents
Junichi Uekawa (firstname.lastname@example.org) wrote:
> Ardo van Rangelrooij <email@example.com> immo vero scripsit
> > An NMU of libexpat1 1.95.2-2 done this weekend to make it work on
> > the ia64 platform introduced a severe incompatibility with previous
> > versions: it renamed libexpat.so.0.1.0 to libexpat.so.1.0.0. This
> > of course brings all dependent packages to a dead halt.
> Why was this done in this way ?
What do you mean? The person doing the NMU made a terrible mistake and
should never, never, never have done this (and it could easily have been
prevented). The NMU was done to make the package compile on the ia64
platform. This (apparently) required a changed in the source and hence
also a new source version of the package was uploaded. This was picked
up by the i386 build daemon (and probably also by build daemons for the
other platforms) resulting in a new binary package for the i386. And
the rest is history.
You make it sound the library number change was done deliberately. I
hope the NMU'er did do it unintentionally by simply goofing up.
(If you look at the policy for library numbering it did the _right_
thing, but it did it in the _wrong_ way.)
Ardo van Rangelrooij
home email: firstname.lastname@example.org
home page: http://people.debian.org/~ardo
GnuPG fp: 3B 1F 21 72 00 5C 3A 73 7F 72 DF D9 90 78 47 F9