Re: /usr -> .

[Manuel Menal <mmenal@hurdfr.org>]

Thomas Schwinge wrote:
> [ Moved to the Debian GNU/Hurd mailing list. ]

[snip]

> AFAIK this is not the first time for something like this to happen.

It sure isn't.

> I strongly suggest to remove the '/usr -> .' symlink from the Debian 
> GNU/Hurd system. Having this symlink (and having no /usr anymore, 
> eventually) is suited for the GNU system, but not for Debian.

This has been discussed at lengths. The compromise is to have the
/usr -> . symlink as an *option* with the default being a separate /usr.
Is there really something wrong with that? I see no reason to forbid
those of us that want this symlink to have it.

> And I doubt that anyone will volunteer to fix all the postinst, 
> configure, etc. scripts that are currently "broken".

There are two issues, really. The one I pointed out (with man-db using
'pager' and gzip which are in '/usr/bin' and '/bin') makes clear that
every Debian GNU/Hurd package should be built on a machine with a
separate /usr. But that's only for builders, not users. I've never seen
any cases where properly-built packages caused problems with '/usr' -> .
symlink, except what happened with 'nano' a few months ago, when it
shipped both '/bin/nano' and '/usr/bin/nano' (the second being a symlink
to the first), which resulted into a recursive symlink. But this is rare
and easy enough to fix. Did you see many other cases ?

--
Manuel Menal