VIPS changed from 7.10 to 7.12 (not exhaustive, see ChangeLog for details)

Non-backwards compatible changes
================================

- IMAGE->Hist is deprecated, use im_history_get() instead
- im_append_Hist() removed
- IMAGE->Bbits is deprecated (now ignored)
- im_region_local() replaced by im_region_buffer()

VIPS enhancements
=================

- new backwards and forwards compatible VIPS file format: it has a new 
  metadata system that efficiently shares and propogates ICC profiles,
  EXIF data, etc. & whatever
- speed ups: new pixel buffer cache saves and reuses computations, uses liboil 
  where possible for a small speedup
- SMP scaling: simplified lock system improves SMP scaling, double-buffered 
  image writes overlap computation 
- lower mem use: new mmap window cache shares file mappings, history buffers 
  share and reuse image history metadata
- built-in support for new image formats: OpenEXR, Analyze7, CSV
- RGB16 and GREY16 image hints
- uses GOption for much better command-line argument handling
- better C++ binding: namespaces, VError derives from std::exception, image
  and number vector constants
- python binding 
- gcc sentinel attributes added
- added GValue as an arg type
- added im_benchmark() and friends for testing
- new functions: 
	im_analyze2vips(), im_benchmark*(), im_buildlut(),
	im_concurrency_get/set(), im_contrast_surface(),
	im_contrast_surface_raw(), im_copy_from(), im_copy_morph(),
	im_cp_desc_array(), im_cp_descv(), im_extract_areabands(),
	im_flood_blob_copy(), im_get_option_group(), im_grid(),
	im_header_exists(), im_header_map(), im_history_get(),
	im_invalidate(), im_isscalar(), im_lineset(), im_linreg(),
	im_meta*(), im_msb(), im_norm_dmask(), im_project(),
	im_raw2vips(), IM_REGION_ADDR_TOPLEFT(), im_tile_cache(),
	im_video_test()


VIPS changed from 7.8 to 7.10 (not exhaustive, see ChangeLog for details)

Non-backwards compatible changes
================================

- no longer generates vips-config script, instead it makes .pc files for
  pkg-config ... replace any "vips-config" with "pkg-config vips-7.10"
- origin hint semantics changed ... it now records the position in the output
  image of the origin of the input image
- all float to int casts on pixel values are now floor() rather than rint(),
  for consistency with C casting semantics

VIPS enhancements
=================

- nip reworked for gtk+-2.4, now called nip2
- now based on glib, so VIPS no longer has it's own thread, plugin and data
  structure libraries
- i18n support, although not quite complete yet
- new error message API to make i18n possible (compatibility macros mean the
  old API still works)
- new 'start up VIPS' call, im_init_world(), sets up i18n and threads
  (optional to call it, but i18n won't fully work unless you do)
- C programs now just need "#include <vips/vips.h>", no other includes
  required
- wraps the Pango font rendering library, see im_text()
- new function flag: NOCACHE ... means the result of the call should not be
  cached, useful for video grab functions and paintbox functions
- im_embed() now supports a wider range of embedding styles, including a fast
  edge stretch
- all area operations use the new embed style to expand their input rather 
  than their output, so there are no more black borders on filtered images
- other new functions: im_render(), im_cache(), im_extract_bands(), 
  im_copy_swap(), im_rint(), im_make_xy(), im_init_world(), im_error(),
  im_warn(), im_diag(), im_rank_image(), im_replicate()
- many fixes and improvements to old funcs
- configure and portability improvements

Bug fixes
=========

- all serious bug fixes got back-ported to 7.8, I think


VIPS changes from 7.6 to 7.8 (again, not exhaustive, see ChangeLog for details)

Non-backwards compatible changes
================================

- output format options specified in filenames: new prototypes for load and
  save tiff and jpeg
- C++ API name changes ... error => VError
- include path change
    - C includes <vips/vips.h>
    - C++ includes <vips/vipscpp.h>
- im_extract*() numbers bands from zero, not from 1 ... also other funcs which
  refer to bands (eg. im_lrmosaic() etc.)
- mosaic funcs have extra max blend width parameter

VIPS enhancements
=================

- rewritten and much fancier ip (see separate docs), now called nip ... old ip
  (slightly fixed) still there
- mmap windows remove 2GB image size limit
- images have an origin hint field
- icc transforms supported via wrapping of Little CMS
- much, much faster Fourier transforms via wrapping of libfwfft
- ppm/pgm/pbm read/write
- C++ API enhancements: new constructors, more operator equivalences
- new colourspace: Yxy
- support for D50 colour temperature images
- new functions: im_image(), im_system(), im_version*(), im_blend(),
  im_copy_set(), im_flood_blob(), im_icc_*(), im_open_local_array(),
  im_header_*(), im_sign(), im_ceil(), im_floor(), im_remainderconst*(),
  im_print(), im_guess_prefix(), im_remosaic(), im_invertlut(), Yxy funcs,
  color temperature funcs, im_clip2fmt(), im_blend(), im_lab_morph(),
  im_histnorm(), im_histcum(), im_video_v4l(), im_LabS2Lab(), im_Lab2LabS()
- new type: IMAGEVEC
- header is much faster
- ip/nip split to separate packages
- better vips2dj
- better cygwin support, mingw support too
- Mac OS X and Windows build support
- new set of sensible #define names, old names can be turned off with
  -DIM_NO_VIPS7_COMPAT
- many configure improvements

Bug fixes
=========

- speed ups
- fixes to implicit image format conversions 
- im_zoom() spots integer overflow
- im_LabS2LabQ() rounding on a/b slightly broken for a/b == 0
- fixes to refcounting in C++ API
- mask casts in C++ were broken




VIPS Changes for 7.4 to 7.6

Non-backwards compatible changes
================================

- im_histplot() has new (and better) rules
- im_andconst(), im_orconst() and im_eorconst() now take a
  double, not an unsigned char, as their constant argument type
- im_global_balance_float() renamed as im_global_balancef() to be
  more consistent with other functions
- im_global_balance_search() removed ... you can do this efficiently
  with an ip macro now
- new parameter "gamma" to im_global_balance() and
  im_global_balancef() ... sets the gamma for the input device with
  which the images were captured, set to 1.0 for old behaviour
- im_malloc() now returns void *, not char *

Bug fixes
=========

- tiny memory leak in im_list_remove() fixed
- oops, the value for sRGB in the header was wrong, now 22
- missing function descriptor for im_rank_raw()
- im_dECMC_fromLab() was broken
- im_erode() and im_dilate() output incorrect error messages
  if mask elements were not one of 0, 128, 255
- im_rotate_*mask*() functions were not preserving scale and offset
  values

Package enhancements
====================

The whole thing is now GPL, with GNU configure

ip changes
==========

- better display control bar
- matrix operations now appear in the workspace
- new UI elements: input options and toggles
- better LUT optimisation --- arithmetic operations on UCHAR images should be
  much faster
- new macro language --- same as old one, but faster and much more powerful
- all standard menus rewritten in new macro language, much nicer!
- batch operations with Group/Ungroup
- now uses GDK for drawing in image windows, much better colour handling on
  8/16-bit graphics cards
- image repaint is now threaded for an about-linear speedup as you add
  more CPUs
- many interface improvements and polishes

VIPS enhancements
=================

- VIPS now uses POSIX threads (7.4 used Solaris threads) and has been rejigged
  to provide a nice threading API to functions which call the library
- im_open() now knows about LSB- and MSB- first VIPS images, and
  silently converts them to native order ... it also now ignores case when
  deciding what format to write
- new parameter type IM_INPUT_REALVEC, a vector of doubles
- new set of functions with vector constants: im_lintra_vec(),
  im_powtra_vec(), im_expntra_vec(), all relational
  (im_equal_vec(), im_notequal_vec(), etc.), all boolean
  (im_and_vec() etc.)
- new flag for function descriptors: "IM_FN_PTOP" set for point-to-point
  functions ... makes ip use LUTs for operation
- im_tiff2vips() now reads and writes 16-bit images, and knows about zip
  (deflate) compression
- convenience functions im_extract_band() extracts a band from an image;
  im_extract_area() extracts an area
- im_list_member() tests for list contains object
- new functions im_write_*mask_name(), im_create_*maskv()
_ new functions im_remainder() and im_profile()
- fourier filtering, forward transform, reverse transform, make power spectrum
  all free memory more quickly, making them suitable for very large images
- new functions im_isMSBfirst() and im_amiMSBfirst() test images
  and this processor for MSB first byte order
- im_malloc() now prints low-on-memory warnings to stderr with
  im_warning() for easier bug catching
- D65_X0 updated for latest recommedations, D60, D93,
  D55, D50, A, B, C added
- minor fixes to the C++ interface to bring it in line with the newer ANSI
  standards
- more and more comprehensive prototypes in "proto.h" to aid C++ (use of
  "const" etc.)
- im_and*(), im_or*() and im_eor*() can now work on any
  integer image