Bug#704744: pbuilder: umounts /{dev, run}/shm of the *host* system

[Cyril Brulebois]

Thorsten Glaser <tg at mirbsd.de> (07/04/2013):
> reassign 704744 debootstrap
> found 704744 debootstrap/1.0.48
> retitle 704744 debootstrap: umounts /{dev,run}/shm of the *host* system
> thanks

Next time, can you please put the right people in the loop?!

Cc-ing:
 debian-bugs-dist at lists.debian.org

instead of:
  debian-boot at lists.debian.org

is just plain stupid. Maintainers of the package you're reassigning to
don't get your control mail. Way to communicate!

> Dixi quod…
> 
> >Okay, I just run “sudo env DIST=sid cowbuilder --create” and
> >it happened again. I micro-tested this and can point out where:
> 
> Nevermind, it’s debootstrap not pbuilder.
> 
> I changed to include “set -x”, run “mount | fgrep shm” using
> the DEBUG trap, and commented out the call to debootstrap as
> I had it already… and it kept my shm.
> 
> Turns out this looks like being the culprit:
> 
> tglase at tglase:~ $ fgrep -ri shm /usr/share/debootstrap/*
> /usr/share/debootstrap/functions:               umount_on_exit /dev/shm
> 
> The problem here is:
> 
> lrwxrwxrwx 1 root root 8 Apr  8 01:03 /var/cache/pbuilder/base.cow-sid/dev/shm -> /run/shm/
> 
> The symlink is then, of course, followed.
> 
> Reassigning this RC bug to debootstrap thusly.
> Sorry folks, but this does break unrelated software.
> 
> My idea to fix this is:
> 
> Move the “umount_on_exit /dev/shm” line away from
> /usr/share/debootstrap/functions and into the files under
> /usr/share/debootstrap/scripts/ and then change it to use
> /run/shm from wheezy onwards (this also involves breaking
> up the symlink from etch, etch-m68k, lenny, oldstable,
> squeeze and stable to sid, and then, for the wheezy release,
> reintroducing the symlink from stable to sid).
> 
> Unless you’ve got a better one, that is.

Well, wild guess, a consequence of the change introduced in
pbuilder/0.215 without actually checking what happens?

Surely, coordinating such a change with debootstrap people would have
been a better idea than implementing it blindly?

Not sure I agree that's an RC bug in debootstrap. I'd rather call it a
wishlist in debootstrap to support the new thing pbuilder imposes, and
an RC bug in pbuilder not to depend on a debootstrap version
implementing said improved behaviour.

Mraw,
KiBi.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/pbuilder-maint/attachments/20130408/d1d06522/attachment.pgp>