Fixed model sub-sampling issues with large objects, including check-image resizing.
Fixed initialization of DETMODEL size guess (thanks to Eli Rykoff).
Fixed image extension pb in double weighting/single image mode.
Improved debug info for memory allocations in FFT and FITS libraries.
Fixed crashes with DETMODEL magnitudes on null measurement weight maps.
Added Eli Rykoff's fix to discard sources with NaNs in the fitted positions.
Improved MKL autoconfiguration.
Improved vectorization in FITS data transformations and FFT-based convolutions.
Pushed SExtractor version number to 3.18.8.

Added control of the number of threads in MKL calls.
Added support for negative NTHREADS (= max number of threads).
Pushed version number to 2.18.1.

Added 2nd order moments measurements of the convolved model.
Fixed issues introduced in a recent updates with check-images.
Implemented new algorithms for SPREAD_MODEL and SPREADERR_MODEL (no optimization yet and the scaling is still variable however).
Added experimental SNR_WIN measurement parameter.
Fixed crash with empty ASSOC lists.
Pushed version number to 2.13.6.

Preliminary implementation of a RESCALE_WEIGHTS configuration switch.

Added new ASSOCCOORD_TYPE configuration parameter for ASSOCiating detections either in PIXEL or WORLD coordinates.
Fixed sampling issues for Sersic, de Vaucouleurs and exponential model-fitting.
Version number pushed to 2.13.2.

Pushed version number to 2.13.1.

Added more complete headers to all files.
Clarified licensing issues.
Updated FITS library.
Updated configuration files.
Added configuration files to the list of data that get installed in ${prefix}/share/sextractor.
Added background offset and point-source model components: FLUX_BACKOFFSET,FLUXERR_BACKOFFSET,FLUX_POINTSOURCE,FLUXERR_POINTSOURCE,MAG_POINTSOURCE,MAGERR_POINTSOURCE.
Tweaked model-fitting convergence parameters.
Fixed catalogue source ordering/number issue in dual-image mode.
Pushed version number to 2.13.0.

Added --disable-model-fitting configure option to turn-off model-fitting features and to allow compiling SExtractor without requiring the ATLAS and FFTW libraries.
Version number pushed to 2.12.0.

Reverted minimization algorithm back to double precision (while keeping single precision for data and parameters in "bounded" space).
Added specific handling of "Jacobian calls" in model-fitting.
Sped up the resampling process in model-fitting.
Tuned up minimization algorithm in model-fitting.
Added shortcuts to the computation of the local PSF model. 
Fixed triggering issues with some model-fitting parameters (thanks to S.Serrano).
Removed CLASS_STAR_MODEL and DISK_PATTERN* measurement parameters.
Added automatic tracking of the PSF FWHM from the local PSF model if SEEING_FWHM is 0 for CLASS_STAR.
Fixed excessive WIN parameter computation time on large ring-shaped objects.
Fixed model-fitting issue with null half-light radii (thanks to S.Desai and B.Armstrong).
Fixed aperture display value error in FITS-LDAC catalogue header.
Removed redundant update_tab() calls in FITS library.
Fixed comment copy issue for slashes within strings in FITS library (thanks to F.Schuller).
Added check of PSF values; SExtractor will now exit in error in the case where all PSF components sum up to 0.
Replaced computation of model second moments and ellipticities with analytical estimates.
Improved the accuracy of peak and average model surface brightness estimates.
Fixed some compiler warnings.
Pushed version number to 2.11.7.

Improved real-time display.
Fixed various issues with XSLT in double image mode.
Pushed version number to 2.11.0.

Improved processing time display.

Fixed error ellipse measurements in PSF-fitting (but works only for PSF_NMAX=1 currently).
Simplified PSF parameter handling (removed PSFDISPLAY_TYPE configuration option).
Simplified triggering of local north axis computation.
Allowed model axis ratios to exceed 1 to facilitate convergence.
Fixed alignment bug for symmetric models.
Restricted maximum allowed range for model coordinates.
Pushed version number to 2.10.0.

Changed CONCENTRATION_MODEL to SPREAD_MODEL.
Added SPREADERR_MODEL measurement (currently a crude estimate).
Optimized FFT calls.
Set PSF_NMAX to 2 by default.