[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gentoo-dev
Subject:    [gentoo-dev] Lastrite: app-admin/eselect-boost & Boost plans
From:       Tiziano =?ISO-8859-1?Q?M=FCller?= <dev-zero () gentoo ! org>
Date:       2012-08-24 9:50:11
Message-ID: 1345801811.68322.26.camel () storm
[Download RAW message or body]

Some of you may have already noticed, that boost >=1.50.0-r1 does not
pull in eselect-boost anymore and does not install a profile for it
either. This is on purpose since app-admin/eselect-boost will be
removed.

Why: the purpose of eselect-boost was to make the introduction of
slotted-boost easier (since we wouldn't have to fix all packages at
once) and to help people using Gentoo as a development platform.
It was never the intention that Ebuilds should depend permanently on the
symlinks created by eselect-boost depending on a users selection.
Especially with EAPI-5 slot-operators the PM must be able to assume that
a package depending on boost links/uses the best version of boost
according to its dependency specification.

I apologize for not having made that clear earlier.

How to make sure your package uses latest boost (if it doesn't detect it
properly)? I'm using something like this in my ebuilds which require
boost:

    BOOST_PKG="$(best_version ">=dev-libs/boost-1.40.0")"
    BOOST_VER="$(get_version_component_range 1-2 "${BOOST_PKG/*boost-/}")"
    BOOST_VER="$(replace_all_version_separators _ "${BOOST_VER}")"
    BOOST_INC="/usr/include/boost-${BOOST_VER}"

If someone has a better idea (possibly introduce pkg-config files for
boost) and/or write/extend an eclass to handle such cases (probably
similar to the db-use.eclass) I'd appreciate a helping hand.

Further plans:

When unmasking glibc-2.16, most of the old boost versions will be broken
and old unstable boost-versions must be removed.
As soon as glibc-2.16 gets stable, boost-1.50.0 has to be stabilized and
old stable versions of boost must be removed.

Boost-1.51.0: mgorny (thanks a lot, btw) and I are working on splitted
ebuilds. The wishlist also includes switching to the cmake-based
modularized boost distribution which would make the ebuilds a lot
easier. It would also reduce the required disk-space when building and
testing.
Since 1.51 got released already, we may decide to introduce this in 1.52
and release 1.51 as usual.

At this point I'd also like to thank hwoarang, floppym and Arfrever for
their help in maintaining boost... and of course everyone participating
in testing, bug-fixing and stabilizing that beast.

Ideas? Hints? Comments?

Cheers,
Tiziano





[prev in list] [next in list] [prev in thread] [next in thread]