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

Bug#167409: glibc 2.3.1: breaks XEmacs builds; system breaks on revert to 2.2.5



>>>>> "Mark" == Mark Brown <broonie@sirena.org.uk> writes:

    Mark> You need to downgrade everything that depends on the new
    Mark> libc before downgrading libc itself.  dpkg should have told
    Mark> you this when you were doing the downgrade (IIRC it should
    Mark> have required some explicit cooercion to do the downgrade).

Right.  That's the point.  As far as I can recall (that system has
since crashed, and what bash history I can recover doesn't contain the
upgrade command) I did

dpkg -i glibc$version_stuff.deb glibc-dev$same_version_stuff.deb

with no --force-depends, expecting to get a bunch of errors from dpkg,
and by looking at what else needed to be reverted, I could decide what
to do.  No such luck; I got a half-downgraded system, with no
warnings.

BTW, that's what I normally do, it's the most straightforward way to
get the needed transitive dependency information.  I guess I'm going
to have to find less dangerous way to get the same information.

    Mark> I don't think any library can reasonably guarantee full
    Mark> backwards as well as forwards compatibility without using a
    Mark> new soname for every single ABI change.  Doing this for libc
    Mark> itself would be unreasonable - it'd be the glibc transition
    Mark> over and over again.

The library can't, but Debian can.  ld --static for essential
utilities, at least in `sid'.  For heaven's sake, `sid' users revert
packages all the time (at least I have to).

-- 
Institute of Policy and Planning Sciences     http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
        Economics of Information Communication and Computation Systems
          Experimental Economics, Microeconomic Theory, Game Theory



Reply to: