View on GitHub

IRAF Community Distribution

IRAF maintained by the community

Home | Installation | Packages | X11IRAF | PyRAF | Forum

ERROR: floating point overflow: Bug within IRAF 2.16 and 2.16.1

ksmith wrote on May 10, 2014


I get the error below with all Ureka installs of IRAF on MacOSX Mavericks and CentOS6.

We're trying to do some astrometry with msccmatch, and the astrometry succeeds, but IRAF barfs with a floating point overflow when trying to write the modified header. Note that neither the image nor the catalogue is the problem. We've tried other images too, and also tried renaming the files. The sequence below SUCCEEDS for Scisoft and other installs of IRAF before 2.16, but fails for Ureka installs with IRAF = 2.16. The kind folks at STScI tested a patch upgrade to their Ureka install with IRAF 2.16.1, but the same error is produced.

Any ideas on how to fix? Files and command sequence, with output below.

Ken

Two files (a FITS file and a catalog) can be found here:

http://star.pst.qub.ac.uk/~kws/ureka_problems/

Here's the command sequence in IRAF (/tmp/atlas is the location of the two files).

ecl> cd /tmp/atlas
ecl> ls
01u56681c0282o.WEB2MASS_J.test  01u56681c0282o.fits
ecl> mscred
This is the initial release of the IRAF FITSUTIL package
to include support for FITS tile compression via 'fpack'.
Please send comments and questions to seaman@noao.edu.

     ccdhedit       darkcombine    mimpars@       mscblkavg      mscdisplay     mscfocus       mscjoin        mscpupil       mscshutcor     mscstat        mscwfits       setinstrument  zerocombine    
     ccdlist        flatcombine    mkmsc          msccmatch      mscexamine     mscgetcatalog  mscmedian      mscrfits       mscskysub      msctoshort     msczero        sflatcombine   
     ccdproc        irmfringe      mscagetcat     msccmd         mscfinder      mscimage       mscotfflat     mscselect      mscsplit       msctvmark      rmfringe       xtalkcor       
     combine        irmpupil       mscarith       mscctran       mscfindgain    mscimatch      mscpixarea     mscsetwcs      mscstack       mscwcs         rmpupil        xtcoeff        

mscred> unlearn msccmatch
mscred> msccmatch.accept = yes
mscred> msccmatch.outcoords = "01u56681c0282o.WEB2MASS_J.used"
mscred> msccmatch.usebpm = no
mscred> msccmatch.nsearch = 1000
mscred> msccmatch.search = 50.0
mscred> msccmatch.rsearch = 0.50
mscred> msccmatch.nfit = 4.
mscred> msccmatch.rms = 5.0
mscred> msccmatch.maxshift = 20.0
mscred> msccmatch.fitgeometry = 'general'
mscred> msccmatch.reject = 3.
mscred> msccmatch.update = yes
mscred> msccmatch.interactive = no
mscred> msccmatch.fit = no
mscred> msccmatch.cbox = 11
mscred> msccmatch.csig = 2.0
mscred> msccmatch.cfrac = 0.7
mscred> msccmatch.verbose = yes
mscred> msccmatch.listcoords = no
mscred> msccmatch 01u56681c0282o.fits "01u56681c0282o.WEB2MASS_J.test"
MSCCMATCH:
 01u56681c0282o.fits:
   1523 input coordinates
   361/1523 coordinates out of bounds
   search using up to 1000 objects:
   search using up to 1000 objects:
   search found offsets of (1, 1) pixels and rotation 0.36 degrees
   1523 input coordinates
   364/1523 coordinates out of bounds
   20/1159 not centroided
   Fit coordinates:
     input number of coordinates = 1139
     average shift = (-0.35, 1.33) arcsec
     tangent point shift = (-0.04, 0.24) arcsec
     fractional scale change = (1.000, 1.000)
     axis rotation = (0.311, 0.303) degrees
     rms = (0.342, 0.429) arcsec
ERROR: floating point overflow
    called as: `msccmatch (input=01u56681c0282o.fits, coords=01u56681c0282o.WEB2MASS_J.test)'
mscred>


Mike Fitzpatrick wrote on May 10, 2014



I don't see the error in the standard v2.16.1 release so this must be something peculiar to the distribution itself (which has custom mods of it's own). The Ureka folks will have to comment on a fix, however you could simply overlay the working MSCRED binaries from the ftp://iraf.noao.edu/iraf/extern directory to continue. If you still have problems then the error is probably coming from one of the core system tasks and using the standard release is the only option (but again, you can download the main iraf system tarball and simply replace the binaries).

Mike Fitzpatrick wrote on May 10, 2014


This problem appears to have already been fixed in v2.16.1, from the images$Revisions file:



imcoords/src/t_ccmap.x

There were missing arguments that will cause "refpoint=user" to crash

with a floating point overflow. (9/21/12, Valdes)



I can't explain why the ST update didn't work, you can get a new x_images.e binary (that goes into your $iraf/bin. directory) from ftp://iraf.noao.edu/iraf/v216/PCIX/support/


ksmith wrote on May 13, 2014


Many thanks. Replacing the x_images.e binary in both my MacOSX and linux64 installs fixed the problem. I've sent a message to that end back to ST (Ureka).

Last post on May 13, 2014