| 1 |
On Sat, Sep 17, 2016 at 9:54 AM, Bertram Scharpf |
| 2 |
<lists@×××××××××××××××.de> wrote: |
| 3 |
> |
| 4 |
> On Saturday, 17. Sep 2016, 10:31:17 +0200, Kai Krakow wrote: |
| 5 |
> > Am Tue, 13 Sep 2016 19:49:04 +0200 |
| 6 |
> > schrieb Bertram Scharpf <lists@×××××××××××××××.de>: |
| 7 |
> > |
| 8 |
> > > The rfkill make suite looks for a version number and in case |
| 9 |
> > > the work directory is a Git repo it tries to query the |
| 10 |
> > > version number there. Then, it finds the Git repo in |
| 11 |
> > > /usr/portage and fails. Not very friendly! |
| 12 |
> > |
| 13 |
> > You could |
| 14 |
> > |
| 15 |
> > # mount -o bind /usr/portage/tmp /var/tmp/portage |
| 16 |
> > |
| 17 |
> > and then point the portage tmp dir to that directory. From there, |
| 18 |
> > ebuilds cannot see the .git of /usr/portage. |
| 19 |
> |
| 20 |
> As long as I do not set GIT_DISCOVERY_ACROSS_FILESYSTEM... |
| 21 |
|
| 22 |
I don't think that if you navigate upwards after following a bind |
| 23 |
mount that you end up in the tree that was mounted. Ie, |
| 24 |
/var/tmp/portage/.. = /var/tmp, not /usr/portage. With a symbolic |
| 25 |
link it would resolve to /usr/portage unless the shell does something |
| 26 |
clever. |
| 27 |
|
| 28 |
You can mount bind mounts into containers, and I'm pretty confident |
| 29 |
the container can't navigate out into the rest of the filesystem that |
| 30 |
way. |
| 31 |
|
| 32 |
> |
| 33 |
> I decided to add the following line to /etc/portage/make.conf: |
| 34 |
> |
| 35 |
> GIT_CEILING_DIRECTORIES="$GIT_CEILING_DIRECTORIES:$PORTAGE_TMPDIR" |
| 36 |
> |
| 37 |
> This should be part of the documentation. Where do I best |
| 38 |
> report it? |
| 39 |
|
| 40 |
Well, for anything on the wiki I'd suggest just adding it. For the |
| 41 |
manpages I'd submit a bug on bugs.gentoo.org against portage (the |
| 42 |
software). |
| 43 |
|
| 44 |
|
| 45 |
-- |
| 46 |
Rich |
Archives