iraf.net

>From massey@lowell.edu Tue Oct 24 12:26:08 2000
From: Phil Massey <massey@lowell.edu>
To: mfitzpatrick@noao.edu
Subject: "install"Hi, Mike---I've been running "install" as root
on each workstation being fed IRAF from our server, but our system
manager keeps questioning the necessity for doing so. Could you give
me a simple answer about what this actually accomplishes? So far, the
only thing I've been able to figure out is that it provides a link
on /usr/include/iraf.h. In particular: if "install" was run once a long
time ago, is there any need to rerun it?If it were just up to me I would of course just rerun it rather than question
the necessity, as it's just a few minutes to do, but you know how system
managers are...thanks,
phil>From fitz Tue Oct 24 12:39:58 2000
To: massey@lowell.edu
Subject: Re: "install"Hi Phil,
The biggest thing that install does is to make links for such
commands as 'cl' and 'mkiraf', as well as hardcopy graphics translators,
compilation tools such as 'xc' and 'mkpkg', these are normally put in
public bin directories such as /usr/local/bin which require root per-
missions. It also creates a /usr/include/iraf.h link which is used by
iraf networking to get the location of iraf on a system, and occassionally
by users who don't define a $iraf in their environment (whether they know
it or not). It also creates /dev/imt1i and /dev/imt1o which are the
fifo pipes needed for image display, although only SAOimage really needs
these at the moment. Lastly, the script changes the ownership/mode for
the alloc.e binary which we learned about yesterday 8-) Internal to IRAF
the install script also edits the iraf root path into various files which
need it (e.g. the 'cl' command).
Root permission is required because we write to root-owned
directories. The script needs to be run on each machine in the local iraf
network to establish the <iraf.h> link and /dev fifos, even if the bin
directory is shared via NFS. Bits of all this could certainly be done by
hand but you're on your own to get all the pieces right.
Generally you need to rerun the install script anytime the iraf
root path changes, or there is an OS upgrade which might have wiped out
/usr/include or /dev. It doesn't hurt to run it at each IRAF upgrade so
you're sure to pick up changes like new hardcopy translators, or revisions
to the install script itself which may be required for a new version. For the truly paranoid sysadmin there are user-level workarounds for
all of this but that requires tweaking each and every user environment to
work, it's simplest to just run the install script.Cheers,
-Mike>From massey@lowell.edu Tue Oct 24 12:49:27 2000
To: fitz@noao.edu
Subject: Re: "install"See, we have one funny situation that I didn't both to describe: we all share
a common /usr/local. (I know, it's insane...but....): it's aliased to
/opt/local, which is the shared disk that is serving iraf. So, running install
once on the server actually sets all the stuff on /usr/local. But I think
your point about /dev/imt1i and /dev/imt1o for SAOimage will be convincing.THANKS!
---phil>From fitz Tue Oct 24 12:53:28 2000
To: massey@lowell.edu
Subject: Re: "install"We use a shared /opt/local here as well, it's not that uncommon. The more
convincing argument is the need for a /usr/include/iraf.h so that you
(or future summer students) can use iraf networking for remote tape
access, image/display access, etc. If IRAF is shared via NFS then you'll
also want to visit each machine and perhaps create a /iraf symlink to
bypass the NFS mount point so that on each machine something like
/iraf/iraf is a valid path (otherwise things only work on some machines).
The install script won't create this for you, but as long as you're going
around to the machines anyway ....