***************************************************************************

                           Older GILDAS News
                         (J.Pety, 01-feb-2013)

***************************************************************************

Dec07
-----

- General:
     * All the code used to handle projections (WCS related aspects) has
       been factorized into the gwcs library. The SFL projection has been
       added.

- Build:
     * Introduced the "integ-clean" target which recursively removes the
       "built" directories wherever it is called. It also removes the
       $gagsrcdir/integ directory when called from $gagsrcdir. It should be
       used with caution as the user can loose several hours of
       compilation...

- MAPPING:
     * The READ MASK and SUPPORT /MASK commands now enable the definition 
       of disconnected deconvolution supports through the use of bitmap
       instead of polygons.
     * SUPPORT /RESET now resets the definition of the current support to
       empty space.
     * MAPPING now accepts to treat mosaic of up to 999 fields.

- Contrib:
     * GAUSSCLUMP now works under 64 bits machines. It is now part of the
       standard GILDAS compilation.

Nov07
-----

- Build:
     * A major overhaul of the gildas build system has been implemented to
       add the following functionalities:
         o Being able to compile exactly the same sources:
              = Under different systems (eg pc-redhat9-ifort, pc-redhat9-g95,
                x86_64-fedora6-ifort,...).
              = With different compilation options, eg standard options,
                debug options, profiling options, using SYSV, using POSIX,
                etc...
           This way, you can check the modifications you are doing on the
           sources on different systems available to you without having to
           commit first.
         o Put libraries, binaries and other compilation products in a
           small number of directories to simplify the changes of some
           environment variables (PATH, LD_LIBRARY_PATH, etc...) and to
           simplify the reading of the make output.
         o Get a cleaner separation between sources and compilation
           products, eg when you update your working copy, you will not be
           annoyed anymore by files unknown to the cvs repository.
         o Enable the compilation of a particular package against an
           already compiled version of GILDAS.
     * This implied the following changes: 
        o The default is now to link executables against shared libraries
          even though both static and shared libraries will be
          produced. The possibility to link all gildas executables with
          static libraries is maintained. To do this, just use the
          following commands
                    shell-prompt> source admin/gildas-env.sh -o staticlink
                    shell-prompt> make
        o The foo_pro: (e.g. greg_pro:, astro_pro:, class_pro:) have been
          replaced by:
             = Empty string when dealing with procedure name (e.g. "@
               greg_pro:p_load.greg" has been replaced by "@ p_load.greg"),
               the resolution of the absolute name being done with the
               macro#dir: path.
             = gag_pro: when dealing with help files and init files (e.g.
               "run extrema greg_pro:extrema.init" has been replaced by
               "run extrema gag_pro:extrema.init").
        o The gag.dico.gbl has been simplified so that all the file
          positions will depend only of two environment variables
          "GAG_ROOT_DIR" and "GAG_EXEC_SYSTEM".
        o The compilation products will not be anymore in the current
          source directory. Instead they will be gathered in a directory
          tree named $gagsrcdir/integ (integ stands here for
          integration). This directory tree has a structure almost
          identical to the executable (installation) tree.
        o Compilation intermediate products (eg preprocessed files) are
          stored in local subdirectories named "built/machine-os-compiler".
     * Everything else did not change. In particular:
        o Standard users outside IRAM, who compile themselves GILDAS, should not
          either be affected.
        o The executable tree structure.
        o The commands to compile:
            shell-prompt> source admin/gildas-env.sh
            shell-prompt> make
            shell-prompt> make install

- Kernel (SIC and GREG):
     * The Post-Script files produced by GILDAS were misinterpreted by
       recent printer drivers (e.g. the ones shipped with Fedora Core 5 and
       6). This produced oversized prints. This has been fixed with the
       side effect of printing an additional blank page... To be followed.
     * Improved the behavior of the SIC\SORT command to enable sorting a
       complete 2-D table in just one command: SORT T[XCOL] T will sort the
       2-D T variable according to column XCOL. More than one table can be
       sorted on the same command line if needed.
     * Added the /CLIP option to the DRAW TEXT X Y "Text" command to avoid
       writing the text if the (X,Y) point is out of the box. The idea is
       that DRAW TEXT /CLIP will be used to annotate markers plotted (or
       not) by DRAW MARKER.  It could be blindly used to designate sources
       in a field by their names, without knowing if their position is
       within or not the current limits (e.g. DRAW TEXT Ra Dec "Source"
       /USER ABS /CLIP).

- ASTRO:
     * Introduce SET PDBI 1995|2000|2006 command, that selects the PdBI
       frontend/backend status corresponding to years 1995 (old receivers,
       500 MHz bandwidth), 2000 (580 MHz bandwidth), 2006 (new receivers
       and new IF processor). Default is 2006.
     * The behaviour of the LINE command now depends on this year:
         o 1995, 2000: 'old' LINE command;
         o 2006: 'new' LINE, NARROW, SPECTRAL, PLOT commands.

- MAPPING:
     * Fixed the previously incorrect printout of the baseline length in
       Kwavelength. It included a factor 2*pi which was put to compute
       pixel size.
     * GO SETUP now uses A0 and D0 to compute the phase center instead of
       RA and DEC previously. This avoids resetting the phase center
       (A0,D0) each time from the pointing position (RA,DEC).
     * LET MYCLEAN%SHOW NO avoids on-the-fly plots when deconvolving with
       GO CLEAN
     * The new widget interface has been largely improved.

Oct07
-----

- Kernel (SIC and GREG):
     * On-line help has been completed.

- ASTRO:
     * Adaptated to take into account the new 2mm receivers for Plateau de
       Bure Interferometer.

- MIRA:
     * Introduced rotated G3 for polarimetry.
     * Added view /derot.

- CLASS90:
     * Add an option /METHOD to command FIT\DISPLAY to allow a display of
       the fit results of the given method without having to change the
       minimization METHOD.

- CLIC:
     * Now writes one visibility per correlator input (or IF) for both line
       and continuum, instead of an average when the frequency setup was
       identical on each correlator input. The reason is that two IFs will
       have different Tsys, hence different weights (continuum and line).

Sep07
-----

- Kernel (SIC and GREG):
     * The /FORMAT option of the SAY command now is able to format arrays.

- CLIC:
     * Implement PDBI\PLOT /SPURIOUS to plot the location of possible
       spurious lines (3 and 4 times FLO1REF), sometimes observed on Rec 1
       on some antennas.

- MAPPING:
     * Tasks uv_fit.f and uv_fit-s.f no longer re-use existing .uvfit files.
       This avoids problems of inconsistent headers of the .uvfit files
       when changing the number of fitted functions.

- MIS:
     * Fixed UV_FIDELITY and AZIMUTH_AVERAGE crashes under g95 due to
       forgotten initialization.

Aug07
-----
 
- General:
     * The following include files have been renamed from
         ./kernel/include/gbl_convert.par                 to ./kernel/include/gbl_convert.inc
         ./packages/class/lib/class_version.par           to ./packages/class/lib/class_version.inc
         ./packages/class/lib/structure.pat               to ./packages/class/lib/structure-t.inc
         ./packages/class/lib/structure.par               to ./packages/class/lib/structure-r.inc
         ./packages/clic/lib/clic_version.par             to ./packages/clic/lib/clic_version.inc
         ./packages/clic/lib/clic_procedure.par           to ./packages/clic/lib/clic_proc_par.inc
         ./packages/clic/lib/clic_procedure.inc           to ./packages/clic/lib/clic_proc_com.inc
         ./packages/otfcal/lib/cal_version.par            to ./packages/otfcal/lib/cal_version.inc
         ./packages/pako/lib/inc/commands/pakoVersion.par to ./packages/pako/lib/inc/commands/pakoVersion.inc
       and ./packages/class/lib/convert.par has been removed because it was
       unused anymore.
     * Removed ULTRIX, VMS and G77 (obsolete) specific code.

- Kernel (SIC and GREG):
     * The task MOMENTS now correctly handle input file names which include
       a directory.
     * "GO AREA" now also works when FIRST=LAST=0.

- MIRA:
     * An ambiguity in raw data file sorting has been removed when both
       100khz and 1mhz connected.

- CLASS90:
     * Minor bug fix of the PLAIT task.

- CLIC:
     * Fixed the definition of the observatory frequency when creating UV
       tables for the new generation receiver. For information, the
       previous definition had a negligible effect (even though is was not
       exact).

- MAPPING:
     * Fixed the definition and use of supports in the new window interface.
     * Fixed deconvolution of a mosaic in the new window interface.

Jul07
-----

- Build:     
     * MIRA is now part of the default compilation.

- Kernel (SIC and GREG):
     * Better support of the frequency and velocity information when
       reading FITS files of data cubes.
     * The private include files of the kernel have been transferred into
       F90 modules. If this affect your program compiled against GILDAS,
       please contact us at gildas@iram.fr.
     * The E, G and D descriptors are now recognized in the "LET /FORMAT"
       command.
     * The arguments of a procedure are now stored in the PRO SIC
       structure. 
     * The "$ command line" is now forbidden in loops: please, use the
       SYSTEM "command line" equivalent.

- MIRA:
     * A warning is now issued when trying to calibrate a
       front-end/back-end combination with calibration data obtained with
       another front-end/back-end combination.
     * The OVERRIDE command now affects the whole MIRA session and not
       anymore just the next calibration (to be used with precautions).

- CLASS90:
     * Apex backend started with "AP-" are now automatically recognized.
     * Bug fix of the extended FITS support.
     * Maximum file name length is now set by a single variable currently
       set to 512 characters.
     * The GREG command has been debugged in CLASS90.

- CLIC:
     * On-line documentation has been updated to take into account changes
       made to support the new generation of PdBI receiver.
     * New calibration scheme of the Water Vapor Radiometer has been
       implemented. 
     * Some work to ensure compatibility of the current version of CLIC
       (fitted to the new generation of PdBI receivers) with data from
       the old generation of receivers (<2007).

- MAPPING:
     * "GO UVCIRC" produces circularly averaged uv tables both for single
       dish and mosaics.
     * The conversion from radian to sexagesimal notations have been fixed
       for |dec| < 1 deg.

Jun07
-----

- Kernel (SIC and GREG):
     * The possibility to make variable aliases has been introduced:
          DEFINE ALIAS A B
       makes A an alias of the B variable.
          DELETE /VAR B
       removes both A and B. But
          DELETE /VAR A
       removes only alias A.
          HELP DEFINE ALIAS
       for more help.
     * Conversion to sexagesimal coordinates has been fixed to avoid that
       the second and minute fields reach 60.
     * It is now possible to group buttons in menus in widgets.
     * Every new instance of a menu bar which already exists now destroys
       the previous instance.
     * "GO FFT" computes the spatial Fourier transform and modulus of a
        spectra cube.
     * The AZIMUTH_AVERAGE task now also works on spectra cubes instead of
       just one plane of a spectra cube.

 - CLASS90:
     * The reading of the input files used to setup the fit parameters has
       been fixed.

 - CLIC:
     * The flagging of a large number of data is now working again implying
       the possibility to use the tool to assess the quality of the
       calibrated visibility data.
     * This version of CLIC is almost stable for the data produced by the
       new generation of PdBI receivers.

 - MAPPING:
     * The creation and merge of the short-spacings have been fixed into
       the new window interface.

May07 (skipped release to tune the new interprocess communication)
-----

- General:
     * Introduction of a new library called libgwidget which regroup all
       the platform specific code (e.g. MOTIF or WINDOWS widgets). This
       means that linking outside programs against gildas may require to
       add -lgwidget to the list of libraries.

 - Kernel (SIC and GREG):
     * "GO VIEW" now also displays information about the studied data cube
       (file name, source name, line name, etc...)
     * Character fonts are now customizable under unix/linux by the
       standard resource mechanism. For instance, to change the fonts used
       by Gildas:
          shell-prompt> xlsfonts | grep-v [-]
          ... Long list of simple fonts ...
          shell-prompt> emacs ~/.Xdefaults
          ... You add the following lines to change the font to the 9x15 font
                gag_menu*fontList: 9x15
                gag_toolbar*fontList: 9x15
                *gag_dialog*fontList: 9x15
                *gag_file*fontList: 9x15
                *gag_choice*fontList: 9x15
                *gag_help*fontList: 9x15
                *gag_popup_menu*fontList: 9x15
          shell-prompt> xrdb-load ~/.Xdefaults
          shell-prompt> greg
       From this point on, the menu will always use the indicated font
       unless you change again the content of the ~/.Xdefaults file.
     * Suppression of command syntax checking when creating a procedure by
       BEGIN procedure / END procedure. This is (in particular) to enable
       definition of command symbol inside a BEGIN/END procedure.
     * Some bug fixes in the new interprocess communication introduced in
       Apr07.

 - TELCAL:
     * TELCAL is now working under 64 bits machine when compiled in 64 bits
       mode. 

 - CLASS90:
     * "DUMP /SECTION section" enables to dump only the content of a
       particular section.
     * DUMP now also displays the doppler factor in the SPECTROSCOPIC
       section. 

 - CLIC:
     * Debugging of this version of CLIC adapted to the new generation of
       PdBI receiver.

 - MAPPING:
     * "SUPPORT /RESET" removes the current support.
     * The UV_FLAG task now only changes the sign of the weights while it
        was putting the whole visibility spectrum to zero.
     * Introduction of the UV_PROPER_MOTION task which corrects position
       (and thus phases) to take into account source proper motion and
       bring the reference to J2000.
     * A fully new interface has been introduced for mapping and
       deconvolution with the following goals:
        o Having the same functionalities through commands, procedures and
          widgets. In particular, ensure that what is available through
          widgets is available through procedures (eg short-space
          processing).
        o Being able to process single-field and mosaic observations in
          similar way.
        o Ensure that widgets guide new-comers through good imaging and
          deconvolution practices.
        o Ensure that minimum consistency checks are done (in particular
          when producing the short-spacings.)
       It is still a beta version that will be fully debugged in the coming
       months (see above).
 
Apr07 (skipped release to tune the new interprocess communication)
-----

 - General:
     * The GRAPHIC program is removed from the GILDAS distribution. The
       user should use GREG for operation related to lmv cubes and MAPPING
       for operation related to uv tables (imaging and deconvolution). The
       .GRAPHIC extension is still recognize by GREG and MAPPING but this
       functionality may disappear in the future.
     * The NIC program is removed from the GILDAS distribution. MOPSIC is
       the recommended program to reduce bolometric data from the IRAM-30m
       telescope.
     * Introduction of a fully new interprocess communication (POSIX
       complient). This should be transparent for the users except for
       possible remaining bugs. As a major consequence, widgets are now
       available under the CYGWIN operating system.

 - Build:
     * The python binding is now automatically activated under linux 32
       bits machine if both python and numeric or numpy are found.

 - Kernel (SIC and GREG):
     * Fix of a minor bug concerning the superimposition of contours and
       images.
     * The "HSV control widget" to interactively tune the color look-up
       table is back under linux.

 - CLIC:
     * Debugging of this version of CLIC adapted to the new generation of
       PdBI receiver.

 - MAPPING:
     * The WIPE interface has been factorized.
         o All variables are now part of the WIPE sic structure.
         o WIPE is now the only command with INFO|APER|BEAM|ERROR|SOLVE 
           as 1st argument.
     * The uvfit interface has been changed:
         o Introduced uvfit structure to prefix all the variables
           associated with the uv fitting.
         o "go uvfit" is now calling the UV_FIT-S task by default. You need
           to type "go uvfit nag" to get the obsolescent NAG version.
         o Now both the task uv_fit-s and the procedure p_uvfit.map are able
           to fit a linear combination of up to 4 functions but the widget
           takes input for only 2.
         o Changed the way the kinds and ranges of the fitted parameters are
           input into p_plotfit.map. Now uvfit%xtype and uvfit%ytype are
           strings arrays of dimension 6, each string should read like "kind
           min max", e.g. "velo * *". The number of plotted parameters are
           uvfit%nx and uvfit%ny. The number of functions plotted are
           uvfit%np and the order in which they are plotted is stored in
           uvfit%order.
     * Introduced the UVCOV button in the "UV action widget".
     * The FITS command is now able to read the UVFITS format produce by
       MIRIAD. There are still minor problems in exporting UVFITS to
       MIRIAD.

Mar07
-----

 - Build:
     * It is now possible to compile GILDAS under MACOSX/intel processor
       with the ifort compiler.

 - Kernel (SIC and GREG):
     * The MASK task has been debugged.

 - CLASS90:
     * The "SET WINDOW" command now takes effect without any call to the
       PLOT command.

 - CLIC:
     * Debugging of this version of CLIC adapted to the new generation of
       PdBI receiver.
     * "HEADER /PLOT" plots the spectral/correlator configuration
       of the current scan.

Feb07
-----

 - Kernel (SIC and GREG):
     * Bug fix in the handling of the loop variable. This bug implied the
       removal of other SIC variables.

 - CLASS90:
     * Introduction of the difference between the natural telescope
       resolution (MAP%BEAM parameter) and the wanted gridded resolution
       (MAP%RESO parameter) of the XY_MAP gridding command.
     * The MODEL command now works even when no previous observation has
       been loaded: the X axis is computed (irrespective of the existence
       or not of previous R buffer, or of the presence of new supplied X
       array).
     * "FITS WRITE" do not produce anymore a segmentation fault when the
       output file is already existing.
     * Fixed the list of spectra which is kept in the AVERAGING command.
     * Line fitting has been debugged.

 - CLIC:
     * Major upgrade of CLIC to be able to treat data coming from the new
       generation of PdBI receivers. This version is not yet stable and the
       backward compatibility has not been fully tested. So we advise to
       use the Jan07 release of GILDAS to reduce data acquired before 2007.

 - External contributions:
     * GAUSSCLUMPS documentation has been written.

Jan07
-----

 - Kernel (SIC and GREG):
     * Now the REPROJECT task is able to use a template image that has a
       different velocity/frequency axis. A side effect is that the header
       of the template image is not anymore (wrongly) fully copied into the
       output image.

 - TELCAL:
     * Minor changes to enable convergence with negative amounts of water
       vapor.

 - CLASS90:
     * Now the unit of the output cube will be either "K (Ta*)" or "K
        (Tmb)" in the XY_MAP command.
     * The resolution section is now only filled in when the telescope
        name is known in the XY_MAP command.
     * Added UASO as synonym for KITT (USAO is the new name for the 12m
       Kitt-Peak telescope).
     * New optional argument ANGLE to the "DRAW LOWER UPPER MOLECULE"
       command.
     * The documentation of the internal CLASS format has been updated.
     * The new command MERGE merges several irregularly continuum drifts
       into a single drift, without resampling nor averaging. The
       integration time is the sum of each int_time. Tsys and tau are
       time-weighted averages. No consistency check is done for the moment:
       you can merge drifts from any TELESCOPE...  A new type of plot,
       namely SET PLOT P, draws the R buffer with the current marker.


Dec06
-----

 - Build:
     * Now the paper size (A4 by default) may be changed by exporting
       GAG_PAPER_SIZE=US_LETTER before sourcing gildas-env.sh.

 - Kernel (SIC and GREG):
     * The command "SIC\SORT VarKey [Var1 ... VarN]" reorder N+1 variables
       according to the order of variable VarKey. Accordingly, the
       particular case GREG1\SORT is now obsolescent. Please, do not use it
       anymore. 
     * "COMPUTE Loc LOCATION Array Value" find from a sorted array Array
       the integer array Loc[2] so that:
              Array[Loc[1]] < Value < Array[Loc[2]]
              Loc = 1 if Value < Array[1] & Loc = n if Value > Array[n]

 - CLASS90:
     * Added the /REDUCED and /OBSERVED keywords to the FIND command.
     * The AVERAGE command is now able to add frequency switch folded
       spectra to position switch spectra.

 - CLIC:
     * The production of the report is now using LaTeX2e.

Nov06
-----

 - Kernel (SIC and GREG):
     * Implemented variable communication between SIC and PYTHON.
     * First introduction of the PYTHON command.
     * PYTHON functions can be called directly from SIC, e.g. "SIC> LET A
       BESSEL(2)" will call the python Bessel function if it exists and the
       python process has been started from SIC.
     * "CHANGE LUT STATIC" make the LUT command change the color look-up
       table of the to-be-drawn images to enable the possibility of having
       several color look-up tables on the same plot. "CHANGE LUT DYNAMIC"
       enables to get back to the current default, i.e. a LUT command
       modifies the color look-up table of the whole display whenever it is
       called in a procedure.

 - CLASS90:
     * "SET MODE ALL" is now also working for the Z axis.
     * "GO ALIGNMENT" and "GO EFFICIENCY" commands are replacement of the
       old "@ ALI_CHECK" and "@ TEL_CHECK" procedures.
     * Debugging of the case of an irregularly sampled frequency axis.

Oct06
-----

 - Kernel (SIC and GREG):
     * LOGICAL variables can now be output with the "SAY /FORMAT" command.
     * The $ character is not anymore allowed in variable names to ensure
       variable name compatibility with PYTHON variables.  As a
       consequence, the SIC$... GTV$... GREG$... and POLY$... have been
       replaced by the structures SIC%... GTV%... GREG%... and POLY%...

Sep06
-----

 - Outside contributions:
     * A gravitational lensing simulation task GLENS is now available in the
       $gagsrcdir/contrib directory.

 - ASTRO:
     * The PdBI sensitivity estimator has been upgraded for the new
       generation of receivers.
     * The definition of the spectral setup (RF, IF and correlator) has
       been rewritten for the new generation of receivers.

 - CLASS90:
     * Reordered initialization to avoid creating the log and messages
       files before the creation of the $HOME/.gag directory tree.

 - MAPPING:
     * Serious bug fix in the ALMA simulator: In the mosaic case, only
       snapshot were simulated whatever the integration time. All the
       known past and present users have been notified. The
       implications for the results of the ALMA memos using this simulator
       is negligible because all the simulations were snapshot simulations
       anyhow.

Aug06
-----

 - CLIC:
     * Since ??, the weights of the visibilities computed by the TABLE
       command were wrong when no resampling was applied. This was only a
       scaling factor (=number of channels in the spectral unit), so all
       maps were correct. But the absolute values of the weights were
       wrong, giving a wrong rms in the map headers.

Jul06
-----
 
 - Build:
     * GILDAS can now be compiled under MacOSX with an Intel processor.

 - Kernel (SIC and GREG):
     * "LET I /INDEX H H2 H+" now correctly resolve the ambiguity between
        H, H2 and H+.
     * Protect against NaN in bitmap.

 - CLASS90:
     * Added FCRAO (14m dish) in the list of known observatories.
     * The RESAMPLE command now avoids an infinite loop when searching for
       blanked values at spectrum edges
     * BASE LAST is working again for sinusoidal baselines.

Jun06
-----

 - Kernel (SIC and GREG):
     * The EXAMINE command now display variable values for all
       structure components but arrays. Recursive behavior is available.
       For instance:      
         SIC> exa struct
         STRUCT         ! Structure GLOBAL    
         SIC> exa struct%
         STRUCT%        ! Structure LOCAL  1  
         STRUCT%SUB%REALVAR =    0.000000             ! Real    LOCAL  1  
         STRUCT%SUB     ! Structure LOCAL  1  
         STRUCT%REALARRAY is a real array of dimensions      2
         STRUCT%STRINGVAR       ! Character*   16 LOCAL  1  
         STRUCT%INTEVAR  =            0               ! Integer LOCAL  1  
         STRUCT%REALVAR  =    0.000000                ! Real    LOCAL  1  
       Only the first 15 characters of a string are displayed with trailing
       dots if needed.
     * First try to use the external FFTW library.

 - CLASS90:
     * GO WHERE is a quick way to plot the spatial coverage of the spectra
       belonging to the current index.
     * The BASE command has been speed up by a factor 4 after profiling.
     * The new COPY command copies the whole index from INPUT to OUTPUT
       file.

 - CLIC:
     * New solvers for antenna base pointing and focus have been
       implemented. They enable a better estimation of the error bars.

May06
-----

 - Kernel (SIC and GREG):
     * Support for NaNs and Infs has been fixed for SUN machines.
     * NaNs, Infs and blanking values are now supported in COMPUTE command
     * New support of "virtual" images and headers, i.e. SIC variables like
       images and headers, but not associated to a disk file. The syntax is
       "DEFINE IMAGE Name[dim,...] * Status", where "*" instead of a
       filename indicates the "virtual" status of the image.
     * Improved display of structure content by the EXAMINE
       command.

 - CLASS90:
     * The GREG command is again available.
     * The FITS class format has been extended to handle variable length
       arrays in binary tables and extensions.
     * XY_MAP now enables griding in a rotated frame compare to
       EQUATORIAL or GALACTIC.

Apr06
-----

 - Release fix:
     * A bug in the axis labeling has been introduced in
       gildas-src-apr06a.tar.gz. Users are advised to get the correct
       gildas-src-apr06b.tar.gz (or later) version.

 - Legacy:
     * In LAPACK, avoid optimizing DLAMC3 which is used to test machine
       precision.

 - Kernel (SIC and GREG):
     * COLUMN
     * MFIT has been rewritten to enable the use of more fitting methods
       and to feed back the estimation of the uncertainties.

 - ASTRO:
     * Added an extremely simple sensitivity estimator for PdBI in ASTRO
       menu. 

 - TELCAL:
     * Implementation of a new interface for pointing with a finer
       granularity, e.g. possible access subscan per subscan.
     * Implementation of double beam fitting.

 - CLASS/CLASS90:
     * SOURCE, LINE and TELESCOPE search-filters are now case insensitive.

 - CLIC:
     * The TABLE command is not crashing anymore under 64 bits machines.
     * The translation table of the logical vs physical antenna number is
       now clearly printed when using "SET ANTENNA|BASELINE" or "SHOW
       ANTENNA|BASELINE". The plot antenna labeling is by default
       logical unless "PLOT /PHYSICAL" is typed.

Mar06
-----

 - Build:
     * GILDAS now supports the ifort compiler (version 9.0 and higher) in
       full 64 bit mode under x86_64 machines (ie AMD64 compatible
       processors).

 - Kernel (SIC and GREG):
     * "COMPUTE med MEDIAN array" now set the median of array into med.
     * "CLEAR ALPHA" is not working anymore under the gnome window manager
       (named metacity) shipped with fedora core 4. This is *not* a bug of
       gildas. GILDAS is correctly sending to X the XRaiseWindow
       command. However, this request is now ignored by metacity (but this
       is a general tendency of window manager to ignore these
       requests). This is a collateral damage of the idea that new
       applications should not steal the focus. For instance, if you
       launch many applications when you log in and then you start working
       in an xterm. If the applications take time to popup, the focus will
       stay in your xterm where you are working. The focus stealing
       prevention may be disabled under KDE but not under the version of
       metacity shipped with fedora core 4. The XRaiseWindow is just a
       request. Nothing in the X standard says it should be honored. For
       more information:
         http://lists.freedesktop.org/archives/xorg/2006-January/012420.html
         http://mail.gnome.org/archives/desktop-devel-list/2004-December/msg00306.html
         http://www.motifdeveloper.com/tips/tip18.html

 - MAPPING:
     * "WRITE CCT" has been fixed. It now writes "niter" components (the
       user set limit of clean components) instead of the found number of
       clean component of the last deconvolved velocity plane.
     * "GO CCT" now plots cumulative flux as a function of clean components
       only up to the last found component instead of the last written
       component. This is useful when the deconvolution stops before
       reaching the number of clean component limit (niter).

Feb06
-----

 - General:
     * An annoying bug appeared at exit of GILDAS programs (namely an
       incorrect cleaning of the semaphore resources) in
       gildas-src-feb06a.tar.gz. Users are advised to get the correct
       gildas-src-feb06b.tar.gz (or later) version.

 - Build:
     * GILDAS was compiled with success on a SUN opteron (64 bits
       intel-like architecture) with solaris 10, in restricted 32 bits
       mode.

 - Kernel (SIC and GREG):
     * NaNs and Infs are now supported in LOG, SQRT and other mathematical
       operations.
     * "DEFINE FITS" is now able to understand HIERARCH keywords.

 - TELCAL:
     * The beta release of TELCAL, a GILDAS library implementing focus and
       pointing fitting is now introduced in the GILDAS main compilation.
       This library is now used by MIRA.

 - CLASS90:
     * The beta release of CLASS90, version of CLASS fully rewritten in
       FORTRAN90 with improved OTF support is now introduced in the GILDAS
       main compilation. Standard users will thus been able to use CLASS90
       by just typing

             shell-prompt> class90

       Old CLASS is still available through its standard name
 
             shell-prompt> class

       For more information, please look at

             http://www.iram.fr/GENERAL/reports/class-evol1.pdf

 - ASTRO:
     * Added the leap second of 2006 January 1 from Bulletin C 30 of IERS.

 - CLIC:
     * The RESAMPLE option of the TABLE command has been improved to avoid 
       blank channels at the edges between two contiguous correlator
       windows.
     
 - MAPPING:
     * New UV_RESAMPLE command and tasks that enable velocity/frequency
       resampling of a UV table.
     * "GO UVALL" error bars have been fixed. Note that the error bars are
       correct for real and imaginary part but just approximate for
       amplitude (i.e. amplitude is a positive measure whose probabilistic
       behavior is NOT Gaussian, see Thomson et al. for more details). In
       the same process, "go uvall" structure has been clarified. Moreover,
       "LET DO%TRACK YES" will try to colorize the visibilities depending
       on the track they belong to (up to 7 colors/tracks available). The
       algorithm currently works only on unsorted tables but happily enough
       MAPPING does not write the sorted tables without you explicitly
       requesting it.
     * "GO UVCOV" will plot the uv coverage of the first channel without
       having to input "let first 1; let last 1; let xtype u; let ytype v;
       go uvall". "go uvcov" even tries not to modify your current "go
       uvall" input (ie first, last, xtype, ytype).

Jan06
-----

 - Kernel (SIC and GREG):
     * Bug fix: Typing end does not crash anymore GILDAS programs.

 - ASTRO:
     * The SOURCE command now ignores all unsupported keywords instead of
       issuing an error message.

 - CLIC:
     * Internal changes for a better support of g95.


Dec05
-----

 - ASTRO:
     * Positions of the new stations on the NS and EW track extensions
       (doubling the largest baseline to almost 800 m) have been inserted.

 - CLIC:
     * New configurations (to accommodate extended baselines) have been
       introduced. First observations using the new configuration started
       in November 2005. New names are 6Aq, 6Bq, 6Cq, 6Dq, where 'q' stands
       for 2006 (ipb filename convention). Old configurations used since
       2002 are 6Ap, 6Bp, 6Cp, 6Dp.

Nov05
-----

 - Build:
     * GILDAS now supports the g95 compiler in full 64 bit mode under
       x86_64 machines (ie AMD64 compatible processors).

 - Kernel (SIC and GREG):
     * Integrated spectrum velocity scale is now correct even when the
       velocity increment is negative.

Oct05
-----

 - ASTRO:
     * Now the ASTRO language has priority over other languages to resolve
       ambiguities of ASTRO\HEADER and ASTRO\CONVERT with VECTOR\HEADER and
       GREG\CONVERT.
     * The tool to find the PdBI calibrators now uses the correct name of
       the calibrator catalog.

 - MAPPING:
     * "go cct" now correctly plots the header.

Sep05
-----

 - General:
     * Just set the SIC_INITWINDOW logical variable to NO in your
       $HOME/.gag.dico file to suppress the automatic popup of the plotting
       window and program widget menu. Then typing "@ define" at the
       program prompt will pop both of them up at your own will.
     * Now the GILDAS bash_profile or login files output the selected
       version *only* when there exists a prompt (i.e. the shell is
       interactive). This is to avoid problems with secure remote shell
       commands.

 - Kernel (SIC and GREG):
     * "go view" key controls have been changed as follow:  1) "E" now exit
       instead of extract. 2) "X" now extract a subcube.  3) "C" is now
       window dependent, i.e. it toggles from absolute to relative
       coordinates either in the images or the spectra, not both
       anymore. 4) "P" now make an color EPS file in a "ha" subdirectory.
     * SIC now enables the definition of up to 32 languages and 1024 commands.

 - ASTRO:
     * Updated dut1 according to IERS recommendation.
     * The GAG_ASTRO_OBS logical name was introduced. When the OBSERVATORY
       command is given without argument, the GAG_ASTRO_OBS logical is
       translated to obtain the observatory name.
     * GAG_ASTRO_OBS is set by default to "BURE" in the gag.dico.gbl
       file. To override this default at compilation time, users have to
       "export GAG_OBSERVATORY=USER-OBS-NAME" just before sourcing
       gildas-env.sh.
     * The "astro_local.astro" procedure now calls the ASTRO/OBSERVATORY
       command without argument. The procedure behavior then depends on the
       fact the chosen observatory is BURE or VELETA.

 - MAPPING:
     * The UV_AVERAGE task now uses the correct reference frequency in all
       cases.

Aug05
-----

 - Build:
     * Removed the g95 "-std=f95" switch to avoid problems with equivalence
       statements.
     * Minor modifications to ensure correct compilation under
       sparc-solaris-f90.
 
 - Legacy code:
     * Suppression of some redundant routines between blas, lapack and
       slatec.

 - Kernel (SIC and GREG):
     * Bug fixes for port to 64 bit machines in particular the plotting of
       images.

 - ASTRO:
     * APEX has been added to the list of known observatories.

 - CLIC:
     * Increased to 20 the maximum number of files that can appear in the
       "Open Raw Data File Widget".

Jul05
-----

 - Kernel (SIC and GREG):
     * All SIC variable started with "GTV$" has been put into the structure
       "GTV%". For instance, GTV$NOFAIL is now called GTV%NOFAIL.
     * Update documentation of "go bit", "go map", "go lmv" and "go nice".

 - OTFCAL:
     * Major synchronization between GILDAS and 30m versions of
       OTFCAL. From now on the 30m is using the GILDAS version of OTFCAL.

 - CLASS:
     * The (user-defined) descriptive coordinate system is now better
       described in the header of spectra. This is the first step toward a
       better use of this kind of coordinate system.

 - MAPPING:
     * Debugged the definition of the synthesized beam by the user when the
       user does not trust the MAPPING default.

Jun05
-----

 - General:
     * First round of documentation of the previously undocumented tasks.
     * Lot's of cleaning to prepare for new features in future releases.

 - MAPPING:
     * UV_ASCAL, UV_SELFCAL and UV_MCAL tasks have been respectively
       renamed to UV_REF_POINT, UV_REF_GAUSS and UV_REF_MODEL.
     * Antenna-based calibration inside the UV_GAIN task is working again.

May05
-----

 - MAPPING:
     * Bug fix (starting with gildas-src-may05c.tar.gz): in apr05 release,
       global SIC variables BLC and TRC were incorrectly set to a non-zero
       value when cleaning. This lead to strange behaviors when changing
       the size of the dirty image in the same MAPPING session. MAPPING
       does not change these values anymore as it should be.

 - NIC:
     * Bug fix: ONOFF command is working again.

Apr05
-----

 - General:
     * Unfinished developments to improve the fitting routines of CLIC
       accidentally arrived in gildas-src-apr05a.tar.gz. Users are advised
       to get the correct gildas-src-apr05c.tar.gz version.

 - Build:
     * Compilation success with g95 under MAC OSX. This version passed
       minimum tests currently available.
     * Compilation success with ifort under linux on an AMD 64-bits
       processor.

 - Kernel (SIC and GREG):
     * Better cleaning of allocated memory to avoid memory leaks.
     * Command RGDATA fixed when fed with 3 arrays of identical size.
     * "SIC SEARCH" is marked as obsolescent, i.e. user can still use it
       but it will disappear in the future. Message indicates possible
       replacement inside GILDAS.
     * There is NO more limitation in the number of file lines that a 
       single COLUMN command is able to load.
     * The widget associated to REPROJECT now correctly states the
       possibility to change the Epoch of equatorial coordinates.
     * Variables associated to GREG3 commands (IMAGE, KILL, SPECTRUM)
       are now defined only at first call of one of those commands.

 - ASTRO:
     * Fixed the ATM [INTERPOLATE|MAKE|SAVE] command.

 - CLASS:
     * Plait algorithm for OTF data is now available as the PLAIT task.
     * Now polygons may be used to define baseline windows for OTF data.
     * Fixed behavior of ACCUMULATE when R and T have different spectral
       resolutions and the weights are equal.

 - CLIC:
     * First step implemented to work with the "new" ATM library developed
       by J.Pardo at CSIC.

Mar05 (skipped)
-----

 - General:
    * Some in-depth changes were needed in the kernel to avoid memory leak.
      Testing time is required before releasing those changes. It has thus
      been decided to skip Mar05 release.

feb05
-----

 - General:  
     * Following some late minute changes, a bug happened on exit for the
       gildas-src-feb05a.tar.gz version. Users are advised to get the fixed
       gildas-src-feb05c.tar.gz version.

 - ASTRO:
     * Parallactic angle now also computed for planets.

 - CLIC:
     * Added histograms in the visibility quality assessment.

jan05
-----

 - Build:
     * Fixed compilation with IBM xlf under MAC OSX.
     * First compilation success with g95 under MAC OSX. However, there
       still are some problems at run-time.
 
 - Kernel (SIC and GREG):
     * New priority mechanism among languages to avoid undue ambiguities
       (e.g. to enable that the SPECTRUM command inside CLASS will call
       by default the CLASS\SPECTRUM and not the GREG3\SPECTRUM). For more
       details, HELP SIC PRIORITY.

 - ASTRO:
     * Fixed a minor bug in the "LINE /SPECTRAL".

 - CLIC:
     * Fixed normalization of RF bandpass calibration (antenna-based, per
       channel, for continuum units).

dec04
-----

 - General:
     * There is a new GILDAS release under MS Windows! It is named
       gildas-win-dec04a.exe. It is much closer to the standard UNIX/Linux
       version than the previous release (i.e. gildas-win-jun04a.exe). For
       complete instructions, please look at install-gildas-on-windows.txt
       on the usual GILDAS download area.

 - Kernel (SIC and GREG):
     * Task GAUSS_2D fixed.
     * Improvements of MVIEW task.

 - CLIC:
     * Normalized integration time to the number of baselines in CLASS FITS
       file written by the CLIC\FITS command. This enable correct time
       weighting of spectra inside CLASS when the spectra were obtained
       with a different number of antennas.

nov04
-----

 - Build:
     * Support of G95 under linux.
     * Partial support of G95 under Cygwin.
     * Better support of the SUN platform.

 - ASTRO:
     * Behavior change: The beam used in computation is now available
       through the READONLY SIC variable PRIMBEAM (in arcsec). User may
       override the default used beam size by setting the BEAM SIC variable
       to a non-zero value (in arcsec).

oct04
-----

 - ASTRO:
     * There now is a basic but useful PdBI time and/or sensitivity
       estimator.

 - CLIC:
     * Changes in the main calibration widget: new AUTOFLAG button and
       replacement of the EFF button by FLUX1 and FLUX2 buttons.
     * Improvement of the "self-calibration on point source" processing.
     * Cleaning of the 22GHz procedures.

sep04
-----

 - General:
     * Removed reference to all previous email addresses (in particular
       nic@iram.fr and alma-simulation@iram.fr). Users should now use
       the gildas mailing list for bugs, comments and questions.

 - CLASS:
     * User can now smooth up to 50 continuous channel (maybe useful for
       cosmological CO detection).

 - CLIC:
     * Bug fix in the writing of the ALMA-TIFITS format.
     * Major cleaning of the procedures used for calibration of PdBI
       standard observing modes.

 - MAPPING:
     * Worked around a ifort compiler which made the uv_fit-s task crash.

aug04
-----

 - Build:
     * Updated first-hand documentation: AUTHORS, INSTALL, PROBLEMS,
       README, WISHES. 
     * Definitely removed G77 support.

 - Kernel (SIC and GREG):
     * Polygons may now have up to 500 gons.

 - CLIC:
     * Holography bug fixes.
     * Implementation of on-line 22 GHz atmospheric phase correction
       (Continued).
     * Major upgrade of on-line help.

jul04
-----

 - Build:
     * Compilation under MACOSX 10.3 (32 bits) is now fully debugged.
     * Fedora Core 2 Linux distribution now supported.
     * General include files are now correctly copied to the executable
       branch enabling "easy" compilation of outside program against GILDAS
       general libraries.
     * Improvement in documentation compilation. In particular, HTML
       documents now correctly reference each others. Content must still be
       fully updated.

 - Kernel (SIC and GREG):
     * PS output is now fully EPS compliant (in principle).
     * DEFINE UVTABLE now enables creation of uv table from SIC.

 - ASTRO:
     * New visibility and wind plot for Bure.

 - CLASS:
     * Debugged and documented the CLASS\FITS command.
     * Added draft documentation of CLASS format.

 - NIC:
     * Works again under Windows.

 - MAPPING:
     * New version of the addition of short-spacing has been added in the
       main menu. Be careful: still in test.
     * A bit of cleaning of uv modelization tasks.

 - CLIC:
     * Implementation of on-line 22 GHz atmospheric phase correction.

jun04
-----

 - Build:
     * Debian and Suse Linux distribution now supported.

 - Kernel (SIC and GREG):
     * Exploration procedures of LMV data cube like "go bit", "go view",
       etc... now works almost in any GILDAS programs (in particular
       CLASS).

 - NIC:
     * Minimum maintenance for use with recent data.

 - Alma simulator:
     * Does not bug anymore at computation start.

may04
-----

 - Kernel (SIC and GREG):
     * On-line HTML help (set on with the "SIC HELP HTML" command) is
       working again.
     * Implemented new naming convention for the on-line help file:
       doc-help-language.nroff where program is the name of the doc where
       the hard-copy (PDF,HTML) equivalent of the on-line help will appear,
       and language is the name of the documented language (i.e.
       sic-help-gui.nroff).
     * Now the 3rd argument of SIC_LOAD is the signed number of commands
       and options (negative means library only mode). This enables the
       user to have on-line help on command like TASK\REAL.
     * Added the possibility to plot box with relative coordinates in
       arcsec in the p_box.greg procedure.

 - OTFCAL:
     * First inclusion into the GILDAS distribution.

 - Mapping:
     * "go snr" computes mosaic noise and SNR maps from signal and weight
        files.
     * Reinserted the forgotten uvsort tasks under uv_sort name.

apr04
-----

  - Major bug fix in interferometric noise estimation:
      * Modification to transparently implement a change of interpretation
        of the GILDAS_UVDATA format. The old format was named
        "GILDAS_UVDAT", and contained an error in the interpretation of the
        weights of the UV tables.  The new format is named "GILDAS_UVFIL"
        and enables a correct direct estimate of the noise. All tasks or
        code creating UV Tables has been modified to use the new
        GILDAS_UVFIL format. GILDAS_UVDAT can no longer be written, so any
        code which would be using UV Tables OUTSIDE of this ensemble will
        no longer work if recompiled without modification. This is a
        precaution to avoid inconsistencies.  Appropriate scaling is done
        on-the-fly when reading old data. Old data cannot be modified,
        though. Only new data is being written in the new format. So, if an
        old UV table must be appended, please use the USER\UVT_CONVERT
        command introduced in GREG and MAPPING to easily convert from UVDAT
        to UVFIL formats.

  - Build:
      * Implemented automatic GILDAS releases in the daily and monthly IRAM
        build.
      * Introduced a useful set of bash functions for GILDAS/USERS
        interaction at IRAM Grenoble.
      * Major upgrade of the GILDAS building system. In particular, it does
        not require anymore to be under bash (i.e. any Bourne-compatible
        shell is OK).
      * Successful compilation under sparc-solaris-f90, rs6000-aix-xlf.

  - Kernel (SIC and GREG):
      * Now correctly handle empty BUNIT keyword when creating a FITS file
        from a LMV cube.
      * Re-enable last procedure or task to be called again when "go" is
        called alone.
      * "go bit", "go lmv", "go map", "go nice" prompt for the "name" and
        "type" variables when they are equal to " ".
      * Transfered GreGPLOT shell script content directly into hard.fpp as
        a simplification. Introduced new logical variable GAG_LPR which
        defines the printing command. Changed printer logical variable name
        from "GAG_PLOTTER" to "GAG_PRINTER" for coherency.

  - ASTRO
      * Support of the /LOW_ELEVATION option for HORIZON command.

  - CLIC:
      * Introduced SET WEIGHT BANDPASS ON|OFF command to enable application
        of bandpass calibration applied to weights in output UV table.  
      * Better message for mis-match of source or position in creation of
        UV tables.

  - MAPPING:
      * The "dummy" UV point at 7.5m has been suppressed, and replaced by
        the value of the primary beam size as "MAJOR axis" of the
        RESOLUTION section in the UV table. This primary beam size is used
        by UV_MAP and siblings to specify the default field of view, but
        not propagated further.

mar04
-----

  - General:
      * Implemented new GILDAS standard behavior, i.e. the default opening
        of the program menu and of one plotting window. User can use the
        "-nw" or "-niw" switches to bypass this default. GILDAS manpages
        should be written.
      * Removed obsolescent equivalent statements between BYTE and
        CHARACTERS using the CHTOBY and BYTOCH new subroutines.

  - Build:
      * Dirty patches to support rs6000-aix-xlf and powerpc-darwin-xlf.
      * First verification that the GILDAS dependencies are available 
        on the compilation system.

  - Kernel (SIC and GREG):
      * Added Xform, XML and PNG support.
      * On-the-fly conversion from simple FITS to GDF formats in command
        like "image" or "define image" is back again as a subroutine
        (i.e. not an independent program anymore).
      * The SIC procedures used to implement "go" and "input" are now
        shared by GREG and MAPPING. 
      * Changes in processing of language ambiguities: 1) A fully typed
        command which raises ambiguity for other commands is now used
        without any message (i.e. FIT and FITS); 2) A message is now output
        when two (or more) identical commands exist in different languages.
      * GAG_EDIT and GAG_PLOTTER are respectively set to the $EDITOR and
        $PRINTER shell environment variables. This is the job of the user
        to ensure that they are correctly defined.

  - CLIC:
      * Debugged widget support. In particular, added missing widget help
        files from a very old GILDAS version.

  - MAPPING:
      * Fixed error bars in "go uvall". 
      * Now the STATISTIC command includes a guard band to decrease the
        effect of increased noise at map edges.

feb04
-----

  - General:
      * Removed obsolescent BYTE statements.

  - Kernel (SIC and GREG):
      * Introduced "go vy" and "go xv" to make easy position-velocity
        plots. 
      * Introduced the "center" variable in addition to the "size" variable
        to be able to position the zoom in "go bit" like procedures.
      * The file of the GREG font definition is now a portable binary file
        (i.e. it is translated on-the-fly from IEEE to EEEI and vice
         versa).
      * Introduced the GTV$NOFAIL logical variable. When set to .TRUE. it
        allows hard-copies to made regardless of whether the output file
        already exists or not.
      * MVIEW task is back.
      * Default hard-copy type changed from "ps fast" to "eps color" because
        only the eps device is fully compliant with the PostScript norm.

  - CLASS:
      * Introduced cube2class.class to create a CLASS file from a spectra
        cube in GDF format.

  - CLIC:
      * Lot's of work to get back pipeline modifications done at Bure
        during work on new gildas into the CVS repository.

jan04
-----

  - Build:
      * Support of ifort, the new version of the Intel FORTRAN90 compiler.
      * Now produce a GILDAS environment file for bash and csh users.

  - Kernel (SIC and GREG):
      * All images are now deleted in a CLEAR TREE action. This raised the
        number of SIMULTANEOUSLY displayed images to 4096.
      * Debugged the histo_cross task.
      * Introduce a Cartesian to polar coordinate transformation procedure
        called car2pol.greg.
      * Now output the task log file to screen after the task end when the
        sic$tee variable is set to false. Default sic$tee value is OS
        dependent because the tee return status is bugged under Linux.
      * New rank (2d instead of 1d) of the SIC "size" variable used to zoom
        in "go bit" like commands.

dec03
-----

  - Build:
      * Improved the building and administration of concurrent GILDAS
        versions: 1) Introduction of the gagadmdir gagsrcdir and gagexedir
        environment variables; 2) Introduction of the define-version.sh and
        define-system.sh scripts; 3) Clearer separation of the gag and go
        script functionalities.
      * Introduced support of the Portland compiler under Linux.

  - Kernel (SIC and GREG):
      * Adapted SIC demo procedure to GILDAS standards.
      * Implemented easy over-plotting capabilities through the "go over"
        command.
      * Almost fully rewritten "go over" procedure to take into account: 1)
        blanking values, 2) rectangular images, 3) zoom possibilities, 3)
        sub-cube extraction, 4) pixel killing/unkilling.  Logics inside the
        loop procedure tries to be simple (instead of clever) for
        readability.
      * The CAR(tesian) projection is now recognized by the GILDAS data
        format. 
 
  - MAPPING:
      * New tasks to calibrated visibilities on a uv model.

nov03
-----

  - General:
      * Added the admin/gildas.el to enable correct font colorization of
        the GILDAS files under Emacs.
      * Standardization of each packages main menu. Verification that help
        files are correctly called.
      * Added the section about the release policy into the FAQ.
      * Adaptations for an experimental proof-of-concept of the python port
        of greg and clic.  

  - Build:
      * Support to build and release monthly version at IRAM.
      * Added protection against used of the building scripts under the csh
        shell. 
      * Introduced the "build" script to easily (re)build GILDAS from the
        CVS repository.
      * Added support of alpha-osf-f90. Build support of 64 bits processor
        in the makefile system to do this.
      * Introduced the "make doc" and "make install-doc" to create and
        install PDF documentation from LaTeX source files.
      * Introduced the "show-path" script to enable easy viewing of long
        path environment variables.
      * Added support for end-user environment customization at end of
        installation procedure.
      * Now GILDAS tries hard to make the executable tree at the same level
        as the source tree. For instance, if your source tree is called
        $HOME/gildas-src or $HOME/gildas, the executable tree will be
        called $HOME/gildas-exe.

  - Kernel (SIC and GREG):
      * Fixed problem with terminal definitions.
      * Fixed string length problem in search of help files.
      * No longer any significant differences between Unix and MS-Windows
        for the handling of environment variables and logical name
        definition files.
      * Improved FITS handling routines.
      * Debugged a complete mis-behavior of LIMITS .. .. .. .. ABSOLUTE
        and PROJECTION commands.

  - ASTRO:
      * Built in support for the new ATM library.

  - NIC:
      * Translated NIC help from .hlp to .nroff format for standardization
        purpose.

  - MAPPING:
      * Recreated an obsolescent GRAPHIC program for compatibility
        reasons. 

From the first import to CVS (05-jun-2003) to first monthly release in new
system (i.e. Oct03, 01-oct-2003):

  - Build:
      * Adapted doc to LaTeX 2e and PDF standards.  
      * Added many missing task help files.
      * Added possibility to automatically make ChangeLog files from CVS
        repository.

  - Kernel (SIC and GREG):
      * Update of SIC on-line help.
      * Introduction of the 1 ENDOFHELP convention to indicate the end of a
        help file. This avoids to have lots of dummy spaces at the end of
        the .hlp files.
      * Support of the back quotes to translated a Logical name.  For
        instance, assuming SPACE_CLIC is a Logical name with translation
        128, then SIC SAY `SPACE_CLIC` will be expanded to SIC SAY 128.
      * Support of multiple default extensions for the Procedures.  
        The syntax is simply 
                SIC EXTENSION [.]Ext1 [[.]Ext2 [[.]Ext3 ....]]]
        It enables generic procedures to be given a generic extension, and
        called in cascade by specific programs.
      * Now automatically create the $HOME/.gag/ directory tree if it does
        not exists.
      * Replaced obsolete GFITS language by a VECTOR\FITS command.

  - ASTRO:
      * The VSOP87 ephemerids file now is a portable binary file (i.e. it
        is translated on-the-fly from IEEE to EEEI and vice versa).

  - CLASS:
      * Replaced obsolete CFITS language by a CLASS\FITS command.

  - CLIC:
      * Common alignment is now checked at run time.
      * Debugged memory allocation problems to trim the memory size to its
         strict minimum.
      * Merged tifits inside clic.  
      * Added support for 64 antennas.  

Before the first import to CVS:

  - One year (part-time) of efforts to:
      * Replace the old home-made preprocessor (fpre) by the standard cpp;
      * Devise a new GILDAS tree and sort the old sources in it;
      * Devise a new makefile system.
      * Configure CVS.

  - The major changes being:
      * The obsolescence of GRAPHIC; its features have been transfered to
        GREG and/or MAPPING; old GRAPHIC procedures should still work in
        GREG and/or MAPPING.
      * The renaming of GILDAS libraries and general include.

Before that: 20 years of history, i.e. 20 years of experience to produce
well tailored softwares!

***************************************************************************
