astrom is a powerful package which allows you to add or improve sky coordinate information to your images, which in turn helps when you are, for example, registering (multi-wavelength) data. However, astrom is quite hard to use: autoastrom tries to make it painlessly automatic.

This bundle includes a version of astrom which generates FITS WCS headers. That astrom version is bundled within the autoastrom download below, but if you want, you can also get it separately.

This is a beta release of autoastrom, version 0.5-12. It builds on Linux, Alpha and Solaris.
Download export_source tarball
autoastrom-0.5-12.tar.gz. Also astrom-3.7.tar.gz is available as a separate download.
SUN/242 for autoastrom (html/ps), and SUN/5 (html/ps) for astrom.
  1. There is no important difference between version 0.5-11 and 0.5-12: the new release is necessary only to resolve a peculiar packaging bug which affected the Starlink distribution.
  2. autoastrom uses ESO catlib internally, and this (or at least version 3.7.2, though this is not the more recent version) appears not to build with gcc 3.3. It does build with 3.2, though the compiler's obviously unhappy about it.
  3. autoastrom is about to be completely repackaged -- this has been on the list for a while, but you'll be pleased to know that it'll become urgent very soon.


Starlink's astrom application provides powerful astrometry facilities, for analysing astronomical images; it is, however, rather cumbersome to use. As described in SUN/5, astrom can:

autoastrom provides a shell around astrom so that, as well as the core astrometric facilities of astrom, autoastrom will:


autoastrom requires at least CCDPACK version 4.0-1 and ATOOLS version 1.5. These may already be installed on your system, but if not should be obtainable from the Starlink software store. If these components are installed in non-standard places, then you need to point to them with the environment variables CCDPACK_DIR and ATOOLS_DIR respectively. Contact Norman if you need advice. autoastrom also relies on KAPPA and EXTRACTOR, but though these should be recent-ish versions, autoastrom isn't particularly sensitive to them.

The script can also use the matching program match. Version 0.7 of this is now distributed with autoastrom, but if you build a separate version (newer or older), and install it in the directory $AUTOASTROM_DIR, then autoastrom will use it.

To build, give the sequence of commands

./mk build
./mk install


Although everything works on the development system, as checked out of CVS, several of these tests fail in the distribution bundle, for silly but hard to fix reasons. Thus these regression tests aren't actually terribly useful for anything in the way of checking that the build worked. Pity about that, really.

There's a pre-installation regression test (if you haven't installed match, a couple of the regression tests will fail in a harmless but obvious way):

$ mkdir /tmp/autoastrom
$ setenv AUTOASTROM_DIR /tmp/autoastrom # or whatever
$ ./mk test-install
$ ./mk pretest

A couple of regression tests fail on DUX, apparently due to rounding errors somewhere.

One of the ASTROM regression tests fails due to inadequate build-time configuration (which is really hard to fix in the distribution bundles because of the ... curious makefile conventions we have).

The regression test t/t9 (which consists of two runs of autoastrom right after each other) seems to hang on rlspc5 at RAL (though it works on my box in Glasgow), with an error which suggests that the Perl/ADAM system has got itself into a terminal fankle. I suspect that this is caused by something like the monoliths from the first run not being fully shut down before the second starts, but I don't have any more specific explanation at present.

2004 June 18