Bug#284469: "cp -a" fails on some filesystems, causing postin st failure

[Hilmar Preusse <hille42@web.de>]

On 06.12.04 Frank Küster (frank@kuesterei.ch) wrote:
> Chad Miller <Chad.Miller@veritas.com> wrote:
>> From: Frank Küster [mailto:frank@debian.org]

Hi *,

> >> This is not a bug in tetex-bin. Every decent filesystem should be able
> >> to set timestamps. This is a bug in the kernel - most 
> >> probably you have
> >> installed from somebody's (Eduard Bloch's?) inofficial install images
> >> with xfs support, and these images used a kernel version with 
> >> buggy ACL
> >> implementation in XFS.
> >
> > Indeed, I am not using a Debian kernel.  I am stuck within a
> > "virtual server" on a huge physical server, over which I've added
> > Debian's userspace. So your position is that having a filesystem
> > in which I can set timestamps should be a prerequisite for
> > installing (not "using", note) tetex?
> 
> Honestly, maintainer scripts cannot cater for every thinkable buggy
> filesystem or kernel.  The "feature" of setting timestamps is part
> of the POSIX specification:
> 
> http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap01.html#tag_01_07_01
> 
> and cp -p is, too:
> 
> http://www.opengroup.org/onlinepubs/009695399/utilities/cp.html
> 
> If I cannot rely on POSIX being followed, I cannot write maintainer
> scripts at all.
> 
After reading this: How about simply closing these bugs with a little
note, that the submitter used a non-official kernel with a broken XFS
implementation. Well I'm using too home made kernels, but I don't
submit bugs to the DBTS if some things don't work, which would work,
if I'd use the official kernel.

Regards,
  Hilmar
-- 
sigmentation fault