This is the reason why eclipse version number jumped from 2.0 to 3.0. Each time a new mode or instrument is supported, the version number is increased. This time it has to be, with the support of ISAAC/SOFI data reduction. Most users on these instruments will make use of the jitter mode, which is an observation technique requiring very delicate data processing procedures. Specialized routines have been written to handle this data flow in the right way, producing publishable images from a set of raw data frames.

What are the changes brought to eclipse to allow processing of jitter modes? First, SOFI and ISAAC produce huge amounts of data every night, and handling a single frame set means juggling with about a gigabyte of FITS files. Virtual memory handling was almost unavoidable to allow fast processing without memory saturation. Another issue was the number of parameters to set up to run the command. The unusually high number of values to provide prevented the use of command-line options which would have been unusable. An initialisation file does the job of parameter setting, making use of the routines provided by the SFL library (see below).

• long options for all commands

All eclipse commands accept now long options. Short options are still supported, they remain the same as they were in the previous version, for obvious compatibility reasons. It is now recommended to use long options for script readability.

• options common to all commands

Some options are now available to all eclipse commands:
• -v or --verbose
• -D or --debug
• -M value or --memory value
• -L or --license
• -h or --help
The meaning of each option is quite obvious from its name. They are explained in little more details in the eclipse-commands manual page.

• new command: 3color

A new command has been released, named 3color. It allows to build 3-colour composite images from 3 FITS images of equal dimensions. The produced output is coded on 24 bits in PPM format. See 3color manual page to learn more about this command.

• arcube and cstcube superseded by ccube

The two obsolete commands, arcube and cstcube have now officially been removed from the eclipse distribution. Use ccube instead.

• new mode for average

The average command offers a new mode to stack cubes to images. The filtering option allows to do configurable non-linear filtering on the 3rd dimension.

• deadpix --clean replaces cleandead

The cleandead command has been removed from the eclipse distribution. Use deadpix --clean instead.

• detpeak

The peak detector has seen its working domain increase, by the addition of the possibility to pre-process the input images with a 5x5 smearing filter. This enables to detect very small and bright peaks in the image which were not previously detected by the base algorithm. Side-effect: bad pixels are almost always detected with this new method. Use with care!

detpeak has also now different possible output formats, namely a simple format with values separated by tabulations, and a saoimage compatible output allowing to load the points directly from this utility.

• new command: dumppix

dumppix is a command allowing to dump pixel values from a FITS file to stdout in ASCII format. It is useful to convert FITS pixel values to ASCII, which is normally read by any kind of data analysis package. Beware: it produces enormous amounts of data and should be used with care.

• filt

Digital filters offered by filt are now specified by name and not by a number. See filt manual page for more precisions. Also: the morphological filters have been significantly optimized to get a factor 2 to 3 in speed. The median filter has received special attention, bringing a speed enhancement of a factor 6 to 7.

• fwhm

This command now offers an ISAAC output compatible mode. This is mostly useful for ISAAC templates.

• pipeline -> adopl

To avoid confusion between the VLT pipeline tools and eclipse commands, the pipeline command has been renamed adopl (Adonis pipeline), since it only handles Adonis data anyway.

• shiftxy now superseded by warping

The shiftxy command is not distributed any more. Use warping instead.

• stcube

It is now possible to specify a rectangular region in an image over which statistics will be computed, instead of doing it always on the whole image. Another option allows to compute only the median image value, but it is not possible to combine it with rectangle or bad pixel map options.

• strehl

The strehl command has been revised. First, whenever a Strehl ratio is found outside of the interval [0..1], a warning is produced. Apparently, it is very important to provide extremely precise input parameters for a correct Strehl ratio computation. A new option -g generates an image of the ideal PSF used for the Strehl computation, it may help to see what went wrong with false measurements.

• warping

Some new options have been added to this command. It is now possible to request integer shifts or integer rotations (rotations of +90, +180 or -90 degrees) for which only pixel swapping operations are needed. In these cases, since no resampling is done, the process is considerably faster.

A family of interpolation kernels is now offered for user specification. The --kernel option allows to change the interpolation kernel. Another option requests warping to show the tabulated kernel values (see --showkernel).

• fitsort

fitsort did not handle HIERARCH ESO keywords for which the equal sign (=) was stuck to the keyword. Corrected.

• stcube

stcube sometimes displayed false values for MIN and MAX fields when taking into account a bad pixel map. Corrected.

• FITS double precision

Misuse of scanf and print functions prevented the precision of any floating point value in a FITS header from having more than a few digits precision. Now the precision has been extended to read all FITS values as double precision numbers, which should retain at least 16 significant digits.

• Other bugs

Many, many other small bugs have been corrected along thanks to user feedback. Thanks to all of you for these helpful comments!