patch to support hurd-i386

[mouring at etoh.eviladmin.org]

On Wed, 27 Dec 2000, Christian Kurz wrote:

> Hi,
> 
> here's a patch so that ssh also supports hurd-i386. Thanks for
> incorporating. The patch comes from Robert Bihlmeyer <robbe at orcus.priv.at>.
> 
> > openssh 2.2.0p1-1.1 does not build on the Hurd. The appended patch
> > fixes that. Changes in detail:
> 
> > * PAM is not (yet?) supported, so the PAM dependencies are only put into
> > the control file on architectures != hurd-i386.
> 
> > * Hurd has no random device, therefore /etc/ssh/ssh_prng_conf is needed.
> > Added to conffiles.
> 
> > * Do not try to generate a hostkey while building the package. This is
> > useless, and it will fail on the Hurd if ssh is not yet installed.
> 
"make install" should alway check if it should generate hostkeys.  The
package manager should not be calling 'make install' unless it's doing a
test run (via -n).  Also in the most recent CVS tree all ssh-keygen are
relative to the make directory.  

> > * Hurd has no MAXHOSTNAME. Two instances replaced by fixed limit, where
> > only the first N bytes were used, anyway. Two more instances replaced
> > by a grow-until-fits loop.
> 
HURD does not have MAXHOSTNAME?!  They must have removed it from the MACH
header files or NeXTStep added it.  I would want to know Markus' view on
that aspect of the patch.  I find it overly complex (besides the fact
we have xmalloc() and xrealloc() to handle the null returns that should
fail using fatal()). I see one or two nice things suggestions in the
patch.  I think they may want to consider either defining MAXHOSTNAME
(because it's used by a lot of software) or we can test and set a limit if
there is none.

- Ben