Re: [gentoo-alt] prefix bootstrap on RHEL7

[Michael Schubert]

Thank you for your answers (and sorry for the delayed reply).

I managed a bootstrap with prefix now (rap failed for other reasons), but I
had to do what I mentioned above and the following:

* add ebuild for python 3.5.2 and adjust patches
* disable python 3.4
* try..except in setting user and group in portage (this fails on my system)
* openssh-7.1_p1.ebuild --> openssh-7.1_p2.ebuild and fix patch levels

cheers
Michael

On Sat, 16 Jul 2016 at 17:46 Fabian Groffen <grobian@g.o> wrote:

> On 16-07-2016 21:58:41 +0900, Benda Xu wrote:
> > Hi Michael,
> >
> > Have you tried RAP at
> >
> >   https://wiki.gentoo.org/wiki/Prefix/libc
> >
> > It passed the test on CentOS 7.
> >
> > Michael Schubert <mschu.dev@×××××.com> writes:
> >
> > > I'm currently trying to bootstrap prefix on RH7.
> > >
> > > I'm running into a couple of issues, but would rather try and sort
> > > them out than submitting a bug report.
> > >
> > > * ${ROOT}/bin/bash [defunct link] collides with bash in stage3
> > > line #1448, disable collision protect works as a workaround:
> > > USE=-collision-protect emerge_pkgs --nodeps "${pkgs[@]}" || return 1
> >
> > This one look unfamilar to me.  How could it become a [defunct link]?
>
> bootstrap_portage:
>     [[ -x ${ROOT}/tmp/bin/bash ]] || [[ ! -x ${ROOT}/tmp/usr/bin/bash ]]
> || ln -s ../usr/bin/bash "${ROOT}"/tmp/bin/bash || return 1
>
> so all that needs to happen is ${ROOT}/tmp/ to be wiped, which we do in
> stage3, because we assume that emerge -e system has taken care of this.
> I guess the hash -r needs to move after the rm, which was forgotten in
>
> https://gitweb.gentoo.org/repo/proj/prefix.git/commit/?id=f3c51fb01b54321c026cfb2857ba3ee65cecaf98
>
> > > * =sys-libs/libseccomp-2.3.1 needs keywords ** in $
> > > {ROOT}/etc/portage/package.accept_keywords, otherwise fail in emerge -
> > > u system #1544
> >
> > That's because prefix overlay is a bit behind the main tree.
>
> If it's keyworded in gx86 and not in prefix-overlay, then that's just
> stupid IMO.
>
> > > * python 3.5 fails building _crypt and nis
> > > WARNING: renaming "_crypt" since importing it failed:
> > > build/lib.linux-x86_64-3.5/_crypt.cpython-35-x86_64-linux-gnu.so:
> > > undefined symbol: crypt
> > > WARNING: renaming "nis" since importing it failed:
> > > build/lib.linux-x86_64-3.5/nis.cpython-35-x86_64-linux-gnu.so:
> > > undefined symbol: yp_master
> >
> > It is https://bugs.gentoo.org/show_bug.cgi?id=587664
> >
> > Probably a glibc compatibility issue, could be solved by a new glibc
> > with RAP.
>
> Yes, could be, but there's likely another problem here.
>
> Fabian
>
> --
> Fabian Groffen
> Gentoo on a different level
>