Libraries getting out of phase with other distributions
Here's some mail from Red Hat regarding shared library compatibility.
It looks as if we are getting into no-interoperation situations. It would
be desirable to avoid that.
Thanks
Bruce
From: Erik Troan <ewt@redhat.com>
To: Peter Dalgaard BSA <p.dalgaard@biostat.ku.dk>
cc: Douglas Bates <bates@stat.wisc.edu>,
Kurt Hornik <Kurt.Hornik@ci.tuwien.ac.at>,
Martyn Plummer <plummer@iarc.fr>,
Friedrich Leisch <Friedrich.Leisch@ci.tuwien.ac.at>, bruce@pixar.com
Subject: Re: Binary linux packages for CRAN
In-Reply-To: <x2ra6duvbt.fsf@blueberry.kubism.ku.dk>
Message-ID: <Pine.LNX.3.96.980113070110.15910C-100000@lacrosse.redhat.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Status: O
On 13 Jan 1998, Peter Dalgaard BSA wrote:
> Douglas Bates <bates@stat.wisc.edu> writes:
>
> > There is a problem with automatically generating .rpm packages from
> > .deb packages for the most recent versions of those distributions.
> > The Debian libreadline.so.2.1 corresponds to the RedHat
> > libreadline.so.3.0. I have exchanged some e-mail with Guy Maor of the
> > Debian project about it and he suggests that it is safe for RedHat
> > users to put in a symbolic link. There appears to be a fundamental
> > difference in approach between the two maintainers of this library for
> > the two distributions. Guy goes with the 2.1 number because that's
> > the number the authors give it and it is upwardly compatible with the
> > 2.0 release. The RedHat maintainer claims it is a major revision
> > relative to 2.0 so should have a new major number.
>
> Well to be completely precise, Eric Troan (RedHat)'s point is that the
> *original* maintainer (Chet Ramey) has set the shared object number at
> 3.0, and that that should be respected. To wit:
>
> Apparently, Chet Ramey decided that it was more important to maintain
> the same .so version numbers on all platforms than to allow the new
> version to work as a drop-in replacement, even though it could, as the
> version number of the non-shared library also indicates.
>
> Eric: Are you *sure* you don't want to put in that silly little
> symlink for Debian compatibility? (Of course this instantly upgrades
> everything that was linked with libreadline-2.0...)
Yes, I'm sure. I really do think that Deiban is doing the wrong thing
here for no good reason. Their glibc library won't be drop-in compatible
with the libc 5.x one anyway, so why not use Chet Ramey's preferred
numbering?
Bruce, I'm copying you on this to see what you think. It would be a
good thing if Debian and Red Hat used the same soname for readline -- how
hard would it be to change the one in your current glibc development
release to the default specified by the readline package?
Erik
Sender: pd@blueberry.kubism.ku.dk
To: Erik Troan <ewt@redhat.com>
Cc: Douglas Bates <bates@stat.wisc.edu>,
Kurt Hornik <Kurt.Hornik@ci.tuwien.ac.at>,
Martyn Plummer <plummer@iarc.fr>,
Friedrich Leisch <Friedrich.Leisch@ci.tuwien.ac.at>, bruce@pixar.com
Subject: Re: Binary linux packages for CRAN
References: <Pine.LNX.3.96.980113070110.15910C-100000@lacrosse.redhat.com>
From: Peter Dalgaard BSA <p.dalgaard@biostat.ku.dk>
Date: 13 Jan 1998 13:51:14 +0100
Erik Troan <ewt@redhat.com> writes:
> > Eric: Are you *sure* you don't want to put in that silly little
> > symlink for Debian compatibility? (Of course this instantly upgrades
> > everything that was linked with libreadline-2.0...)
>
> Yes, I'm sure. I really do think that Deiban is doing the wrong thing
> here for no good reason. Their glibc library won't be drop-in compatible
> with the libc 5.x one anyway, so why not use Chet Ramey's preferred
> numbering?
>
> Bruce, I'm copying you on this to see what you think. It would be a
> good thing if Debian and Red Hat used the same soname for readline -- how
> hard would it be to change the one in your current glibc development
> release to the default specified by the readline package?
On closer examination, the cross-development turned out to be
problematic anyway, this time due to ncurses version skew (which I
suspect is a whole *other* nasty can of worms), so we're probably best
off making binaries on "virgin" releases - if we can find some
accessible machines with them.
(And as someone remarked, how hard is it for the user to type "make
install help" anyway?)
--
O__ ---- Peter Dalgaard Blegdamsvej 3
c/ /'_ --- Dept. of Biostatistics 2200 Cph. N
(*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk) FAX: (+45) 35327907
--
Balance the U.S. budget! - NO tax cuts until the deficit's gone.
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-devel-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: