Re: [gentoo-alt] Gentoo RAP bootstrap script

[Steven Trogdon]

On Sun, Sep 8, 2013 at 6:47 PM,  <heroxbd@×××××.com> wrote:
> Dear Steven,
>
> Steven Trogdon <strogdon@×××××.edu> writes:
>
>> I've tried bootstrap-rap.sh here. I get a failure during stage3 when
>> installing glibc:
>>
>>  ^[[33;01m*^[[0m QA Notice: the following files use invalid (possible
>> non-prefixed) shebangs:
>>  ^[[33;01m*^[[0m
>> storage/strogdon/gentoo-glibc/usr/bin/mtrace:/usr/bin/perl (script
>> mtrace installed in PATH but interpreter /usr/bin/perl not found)
>
> Hmmm, we don't have a perl inside $EPREFIX yet...
>
> I have updated the bootstrap script by symlinking an available perl to
> $EPREFIX/usr/bin:
>
>   https://github.com/redlizard/prefix-libc/commit/daceddc9a
>
> You can try out the new bootstrap script at
>
>   http://dev.gentoo.org/~heroxbd/bootstrap-rap.sh
>
> Thanks,
>
> Benda
>

Yes, this did the trick until emerging portage-2.2.01.22124-r1. A
failure happens when, from the script

*** emerge --oneshot --verbose sys-apps/portage

is executed. During the configure of portage I get:

checking for user id of strogdon...
/storage/strogdon/gentoo-glibc/usr/bin/id: strogdon: no such user
configure: error: error finding the user id of strogdon

!!! Please attach the following file when seeking support:
!!! /storage/strogdon/gentoo-glibc/var/tmp/portage/sys-apps/portage-2.2.01.22124-r1/work/prefix-portage-2.2.01.22124/config.log
 ^[[31;01m*^[[0m ERROR: sys-apps/portage-2.2.01.22124-r1 failed
(configure phase):
 ^[[31;01m*^[[0m   econf failed
 ^[[31;01m*^[[0m
 ^[[31;01m*^[[0m Call stack:
 ^[[31;01m*^[[0m          ebuild.sh, line   93:  Called src_configure
 ^[[31;01m*^[[0m        environment, line 2377:  Called econf
'--with-portage-user=strogdon' '--with-portage-group=math'
'--with-root-user=strogdon'
'--with-offset-prefix=/storage/strogdon/gentoo-glibc'
'--with-extra-path=/usr/bin:/bin'
 ^[[31;01m*^[[0m   phase-helpers.sh, line  521:  Called die
 ^[[31;01m*^[[0m The specific snippet of code:
 ^[[31;01m*^[[0m                        die "econf failed"

I've seen this sort of thing on other tries to build glibc in prefix.
Here the host OS is debian and authentication is via a radius server
over nfs. All user specific info is not stored in /etc/passwd but in a
user passwd database. On gentoo, glibc sets the location to search for
this database as under /var/db and I believe this needs to be
$EPREFIX/var/db. Here on the host the database is stored under
/var/lib/misc. One cannot simply copy over the debian-provided
database to $EPREFIX/var/db and get things to work, at least I don't
thing so. It is possible to get things to work with a little work,
once the search path is $EPREFIX/var/db. One has to, from the output
of "/usr/bin/getent passwd" on the host, generate a database for
gentoo. And then some minor changes have to be made to
$EPREFIX/etc/nsswitch.conf so that things are database aware. To see
the problem, from within $EPREFIX

$ id
uid=105600 gid=925(math) groups=925(math)

and using the host id

$ /usr/bin/id
uid=105600(strogdon) gid=925(math) groups=925(math)

I don't see an easy way to deal with this without manual (user)
intervention, once the search path to the location of the passwd
database is correct. If one symlinks the $EPREFIX id to the host id
then portage will build but will not install due to failures when
python checks for uid info.

Steve