On Mon, Dec 26, 2011 at 05:57:58AM +0100, Thomas Krennwallner wrote: > I see now, but then we have a problem. The semantics of libboost-*-dev > have always been, at least this was my understanding, to depend on the > latest stable version of boost.
Yes, that's the basic purpose of the defaults. In the previous email, I should have made it clear that reverting to the older boost is a temporary measure. Once the build failures are taken care of -- or are at least at a manageable level -- the default will be moved forward again. > The question is now, how shall we proceed? > > 1. If you think that libboost-*-dev should depend on /the currently > most stable version of boost (wrt. the debian package archive)/, > then there is nothing to do and we can close this bug. > > 2. But if you feel that libboost-*-dev should depend on /the latest > stable version of boost/, then I have the impression that we should > force people to fix their package to depend on libboost-*N.MM-dev, > as this is what they then depend on in reality. Because changing > boost-defaults later in time (weeks, months?) to depend on the > new boost release will break those packages eventually. > > I would opt for (2), as packages will only FTBFS with this change, it > won't break compiled packages, as they depend on the linked versioned > shared libraries. But that's just my 2 cents. I think option 2 is the right viewpoint. It's mainly a question of timing and coordination. The Release Team objects when a "package transition" such as this makes a bunch of unrelated packages fail to build. At their request [1], I temporarily backed up the default. The next steps are to test build the prominant reverse-deps, file appropriate bugs, and coordinate the boost defaults change. I could sure use some help with this task!! [1] http://lists.debian.org/debian-release/2011/12/msg00263.html Thanks for your interest in Boost, -Steve
signature.asc
Description: Digital signature
