On Sat, Nov 21, 2009 at 12:35:09AM +0300, Alexey Tourbin wrote:
> On Sat, Nov 21, 2009 at 12:10:38AM +0300, Dmitry V. Levin wrote:
> > On Fri, Nov 20, 2009 at 10:08:09PM +0300, Alexey Tourbin wrote:
> > > On Fri, Nov 20, 2009 at 09:43:16PM +0300, Dmitry V. Levin wrote:
> > > > On Fri, Nov 20, 2009 at 06:33:48PM +0300, Alexey Tourbin wrote:
> > > > > On Fri, Nov 20, 2009 at 06:16:54PM +0300, Dmitry V. Levin wrote:
> > > > > > At least builds of two packages (util-linux and git) were broken due to
> > > > > > this bug.  So I'd better be late than sorry.
> > > > > 
> > > > > Can you please provides some more details?  Did those two packages faild
> > > > > to build exactly because of malformed cache/chroot.cpio?
> > > > 
> > > > I know exactly that e.g. git build failure was due to truncated locale files.
> > > 
> > > [at@people ~]$ hsh-run -- find /usr/share/locale -type f -links +1
> > 
> > I've said nothing about translations, I mean locale files:
> > 
> > $ hsh-run -- find /usr/lib64/locale -type f -links +1 |wc -l
> > 3156
> > $ hsh-run --root -- sh -c 'find /* -type f -links +1 |wc -l'
> > 4036
> 
> Okay, I assume you're reluctant to actually provide any details.

Why so you assume that such tests as t3900-i18n-commit.sh and
t9129-git-svn-i18n-commitencoding.sh may work properly when locale
definitions are severely broken?

No, I'm not going to explain so obvious things here.  If you are so
curious, you can create a build chroot, truncate all hardlinks there,
and you'll see the breakage.  Note that /etc/skel contains hardlinks
but /usr/src (instantiated from there) doesn't.


-- 
ldv