init is started by host system on Prefix, so how about a new dependence like this? (mimics virtual/libc) --- init-0.ebuild.old 2012-05-12 13:31:53.000000000 +0900 +++ init-0.ebuild 2012-05-14 19:10:22.000000000 +0900 @@ -11,6 +11,6 @@ SLOT="0" KEYWORDS="alpha amd64 arm hppa ia64 m68k ~mips ppc ppc64 s390 sh sparc x86 ~amd64-fbsd ~sparc-fbsd ~x86-fbsd ~amd64-linux ~x86-linux" IUSE="" -RDEPEND="kernel_linux? ( || ( >=sys-apps/sysvinit-2.86-r6 sys-process/runit ) ) - kernel_FreeBSD? ( sys-freebsd/freebsd-sbin )" +RDEPEND="!prefix? ( kernel_linux? ( || ( >=sys-apps/sysvinit-2.86-r6 sys-process/runit ) ) + kernel_FreeBSD? ( sys-freebsd/freebsd-sbin ) )" DEPEND="" Reproducible: Always
do not post patches as comments as they get mangled. i cannot understand what your patch is doing here. attach them instead
Created attachment 311805 [details, diff] init.patch the DEP patch
Comment on attachment 311805 [details, diff] init.patch idea is fine, but you need to unwrap it: RDEPEND="!prefix? ( kernel_linux? ( ... ) kernel_FreeBSD? ( ... ) )"
Done. Thanks, vapier.
It isn't clear to me why virtual/init needs to be installed in a prefix environment at all since the way you set up the dependencies it doesn't bring in anything. If packages are depending on virtual/init, wouldn't it be better to have them do something like: !prefix? ( virtual/init ) Thanks, William
if every depend point is going to be changed to prefix?(virtual/init), it's simpler to update the destination so that random developers in the tree need not care about USE=prefix in their own ebuilds.
