[Pkg-ia32-libs-maintainers] Bug#612398: Bug#612398: Bug#612398: ia32-libs: Please add libacl.so symlink

[Goswin von Brederlow]

Philipp Kern <pkern at debian.org> writes:

> Thijs,
>
> am Wed, Feb 09, 2011 at 10:33:49AM +0100 hast du folgendes geschrieben:
>> > IBM's proprietary Tivoli Storage Manager backup client has ACL support
>> > when
>> > running on Linux.  However it's only available if libacl.so can be
>> > dlopen()ed.
>> > As it's still a 32bit product it runs with ia32-libs.  Could you add a
>> > libacl.so -> libacl.so.1.1.0 symlink, please?  That would allow us to drop
>> > the symlinking in postinst, which could go wrong upon purge when another
>> > package needs to resort to the same kludge.
>> Yes, we run TSM with ia32-libs aswell and it works well for us. However I
>> don't know what 'ACL support' means in this context: does it mean the
>> backup will include filesystem acl metadata which can later be restored?
>> Or is it something different?
>
> it's exactly that.
>
>> As for the symlink, this may be possible I'm not sure how much effort
>> there will be put for wheezy for ia32-libs, given multiarch and all.
>
> It might even be helpful for squeeze, albeit it's a tad late for that, sadly.
>
> Kind regards
> Philipp Kern

That software is saddly buggy then. It should be opening
/lib32/libacl.so.1. But I believe that is a verry common mistake.
This means the i386 package would "Depends: libacl1-dev".

The libacl.so -> libacl.so.1.1.0 symlink logically belongs in
ia32-libs-dev and adding libacl1-dev to the packages.list will put it
there. So far only *-dev packages that people asked about have been
added, which is mostly stuff wine needs to compile.

Could we add libacl1-dev for R1 or is that too much change?

MfG
        Goswin