An unsubscribed reader sent me this gem, which I hope he or she won't mind my forwarding to the list. It offers a new way to attack the problem.
---------- Forwarded Message ---------- Subject: Re: [gentoo-user] Two identical systems - different! Date: Monday 03 Mar 2014, 02:33:08 Ah, the extra USE flags checks are coming from an eclass! On 03.03.2014 01:33, Peter Humphrey wrote: > !!! Multiple package instances within a single package slot have been > pulled > !!! into the dependency graph, resulting in a slot conflict: > > app-arch/bzip2:0 > > (app-arch/bzip2-1.0.6-r3::gentoo, installed) pulled in by > app-arch/bzip2[abi_x86_32(-)] required by (dev- > libs/elfutils-0.158::gentoo, binary scheduled for merge) > > (app-arch/bzip2-1.0.6-r3::gentoo, binary scheduled for merge) pulled > in by > (no parents that aren't satisfied by other packages in this slot) > > virtual/libiconv:0 > > (virtual/libiconv-0::gentoo, installed) pulled in by > virtual/libiconv[abi_x86_32(-)] required by (dev-libs/glib-2.36.4- > r1::gentoo, binary scheduled for merge) > > (virtual/libiconv-0::gentoo, binary scheduled for merge) pulled in by > (no parents that aren't satisfied by other packages in this slot) The dependencies in dev-libs/elfutils and dev-libs/glib are app-arch/bzip2[${MULTILIB_USEDEP}] and virtual/libiconv[${MULTILIB_USEDEP}] , where the actual value is calculated by multilib.eclass. It probably should end up empty in your case, but maybe having the chroot triggers some multilib functionality? I may be reading the use dep wrong, but it seems to me that those packages assume that the virtual/libiconv and app-arch/bzip2 packages found are 64-bit since they don't have the USE=abi_x86_32 set. That could at least explain why they get disqualified when glib and elfutils are in the set but work otherwise. I would probably try rebuilding the dev-libs/elfutils and dev-libs/glib packages next and checking that they don't end up still having those abi-requirements. qtbz2 -xO /usr/portage/packages/dev-libs/glib-2.36.4-r1.tbz2 |qxpak -x - -O RDEPEND should tell you what runtime dependencies the package has been built with. ----------------------------------------- I ran that command on both machines and got this output in each case: # qtbz2 -xO /usr/portage/packages/dev-libs/glib-2.36.4-r1.tbz2 |qxpak -x - -O DEPEND virtual/libiconv[abi_x86_32(-)] virtual/libffi[abi_x86_32(-)] sys- libs/zlib[abi_x86_32(-)] || ( >=dev-libs/elfutils-0.142 >=dev- libs/libelf-0.8.12 >=sys-freebsd/freebsd-lib-9.2_rc1 ) !<=app-emulation/emul- linux-x86-baselibs-20130224-r9 !app-emulation/emul-linux-x86-baselibs[- abi_x86_32(-)] app-text/docbook-xml-dtd:4.1.2 >=dev-libs/libxslt-1.0 >=sys- devel/gettext-0.11 >=dev-util/gtk-doc-am-1.15 !<dev-libs/gobject- introspection-1.36 !<dev-util/gtk-doc-1.15-r2 !<sys-devel/gettext-0.18.1.1-r3 || ( >=sys-devel/automake-1.13:1.13 >=sys-devel/automake-1.14:1.14 ) >=sys- devel/autoconf-2.68 sys-devel/libtool app-arch/xz-utils >=sys-apps/sed-4 >=sys-apps/coreutils-8.5 No mention of any kind of zip. Has a lamp flashed on in anyone's head? It would be good to see some light here. -- Regards Peter