Re: Bug#610749: fakeroot fails ownership with kfreebsd systems
reassign 610749 eglibc
severity 610749 serious
thanks
On Mon, Jan 24, 2011 at 12:15:25AM +0100, Aurelien Jarno wrote:
> On Sat, Jan 22, 2011 at 01:15:51AM +0100, Mats Erik Andersson wrote:
> > Package: fakeroot
> > Version: 1.14.5-1
> > Severity: important
> > X-Debbugs-CC: debian-bsd@lists.debian.org
> >
> > I have observed a very strange failure when using fakeroot
> > with dpkg-buildpackage under kfreebsd-amd64.
> >
> > The issue at stake is that links are not assigned the ownership
> > of the superuser. Sometimes even group setting fails, but that is
> > irregular enough that I have problems reproducing also the
> > group-failure in a small package.
> >
> > I include in this bug report a minimal package "wrong-issue",
> > which consistently fails in setting ownership "root" on the
> > symlink "still-wrong-issue" when built on kfreebsd-amd64.
> > The command was
> >
> > dpkg-buildpackage -rfakeroot -B
> >
> > In contrast, with no surprise, the build succeeds with
> >
> > dpkg-buildpackage -rsudo -B
> >
> > and there is no problem with fakeroot in GNU/Linux.
> > I have no good explanation at this moment, but I have
> > rebooted the kfreebsd-system and still the same failure
> > persists.
> >
>
> The problems is triggered by the new version of tar that is present only
> in sid, squeeze is not affected. However given the packages are built in
> a sid chroot, we might be in trouble, especially given the package has
> been uploaded almost 50 days ago...
>
Ok, I have finally understood the problem:
1) in eglibc we implement at* functions, but mark them as stub as the
emulation for FreeBSD 7.x kernels is not perfect. FreeBSD 8.x have
this syscalls natively, and thus don't need emulation.
2) newer versions of tar start to use the fstatat syscall.
3) fakeroot doesn't emulate this syscall because it is marked as stub
in eglibc.
Even if testing is not affected, it's a serious problem as packages are
in unstable before moving to testing. This is way we have roughly 300
broken binary packages in the archive.
Therefore I think it should be fixed as soon as possible. IMHO the best
way to fix that is to remove the stub for at* functions as we don't want
to support 7.x kernel anymore.
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurelien@aurel32.net http://www.aurel32.net
Reply to: