Summary of important user-visible changes for version 3.2:
---------------------------------------------------------

 ** Compatibility with Matlab graphics has been improved.

    The hggroup object and associated listener callback functions have
    been added allowing the inclusion of group objects.  Data sources
    have been added to these group objects such that

           x = 0:0.1:10;
           y = sin (x);
           plot (x, y, "ydatasource", "y");
           for i = 1 : 100
             pause(0.1)
             y = sin (x + 0.1 * i);
             refreshdata();
           endfor

    works as expected.  This capability has be used to introduce
    stem-series, bar-series, etc.  objects for better Matlab
    compatibility.

 ** New graphics functions:

      addlistener         dellistener  ezsurf      ishghandle
      addproperty         ezcontour    ezsurfc     linkprop
      allchild            ezcontourf   findall     plotmatrix
      available_backends  ezmesh       gcbf        refresh
      backend             ezmeshc      gcbo        refreshdata
      cla                 ezplot       ginput      surfl
      clabel              ezplot3      gtext       waitforbuttonpress
      comet               ezpolar      intwarning

 ** Improvements to the debugger.

    The interactive debugging features have been improved.  Stopping
    on statements with dbstop should work correctly now.  Stepping
    into and over functions, and stepping one statement at a time
    (with dbstep) now works.  Moving up and down the call stack with
    dbup and dbdown now works.  The dbstack function is now available
    to print the current function call stack.  The new dbquit function
    is available to exit the debugging mode.

 ** New experimental OpenGL/FLTK based plotting system.

    An experimental plotting system based on OpenGL and the FLTK
    toolkit is now part of Octave.  This backend is disabled by
    default.  You can switch to using it with the command

        backend ("fltk")

    for all future figures or for a particular figure with the command

        backend (h, "fltk")

    where "h" is a valid figure handle.  Please note that this backend
    does not yet support text objects.  Obviously, this is a necessary
    feature before it can be considered usable.  We are looking for
    volunteers to help implement this missing feature.

 ** Functions providing direct access to gnuplot have been removed.

    The functions __gnuplot_plot__, __gnuplot_set__, __gnuplot_raw__,
     __gnuplot_show__, __gnuplot_replot__, __gnuplot_splot__,
     __gnuplot_save_data__ and __gnuplot_send_inline_data__ have been
     removed from Octave.  These function were incompatible with the
     high level graphics handle code.

 ** Improvements to fsolve.

    The fsolve function now accepts an option structure argument (see
    also the optimset function).  The INFO values returned from fsolve
    have changed to be compatible with Matlab's fsolve function.
    Additionally, fsolve is now able to solve overdetermined systems.

 ** Object Oriented Programming.

    Octave now includes OOP features and the user can create their own
    class objects and overloaded functions and operators.  For
    example, all methods of a class called "myclass" will be found in
    a directory "@myclass" on the users path.  The class specific
    versions of functions and operators take precedence over the
    generic versions of these functions.

    New functions related to OOP include

      class  inferiorto  isobject  loadobj  methods  superiorto

    See the Octave manual for more details.

 ** Block comments.

    Commented code can be between matching "%{" and "%}" markers, even
    if the commented code spans several line.  This allows blocks code
    to be commented, without needing to comment each line.  For
    example,

    function y = func (x)
      y = 2 * x;
    %{
      y += 1;
    %}
    endfunction

    the line "y += 1;" will not be executed.

 ** The Control, Finance and Quaternion functions have been removed.

    These functions are now available as separate packages from

      http://octave.sourceforge.net/packages.html

    and can be reinstalled using the Octave package manager (see
    the pkg function).

 ** Special treatment in the parser of expressions like "a' * b".

    In these cases the transpose is no longer explicitly formed and
    BLAS libraries are called with the transpose flagged,
    significantly improving performance for these kinds of
    operations.

 ** Single Precision data type.

    Octave now includes a single precision data type.  Single
    precision variables can be created with the "single" command, or
    from function like ones, etc.  For example

      single (1)
      ones (2, 2, "single")
      zeros (2, 2, "single")
      eye (2, 2, "single")
      Inf (2, 2, "single")
      NaN (2, 2, "single")
      NA (2, 2, "single")

    all create single precision variables.  For compatibility with
    Matlab, mixed double/single precision operators and functions
    return single precision types.

    As a consequence of this addition to Octave the internal
    representation of the double precision NA value has changed, and
    so users that make use of data generated by Octave with R or
    visa-versa are warned that compatibility might not be assured.

 ** Improved array indexing.

    The underlying code used for indexing of arrays has been
    completely rewritten and so the indexing of arrays is now
    significantly faster.

 ** Improved performance for reduction operations.

    The performance of the sum, prod, sumsq, cumsum, and cumprod
    functions has been significantly improved.

 ** Diagonal and permutation matrices.

    The interpreter can now treat diagonal and permutation matrices as
    special objects that store only the non-zero elements, rather than
    general full matrices.  Therefore, it is now possible to construct
    and use these matrices in linear algebra without suffering a
    performance penalty due to storing large numbers of zero elements.

 ** 64-bit integer arithmetic.

    Arithmetic with 64-bit integers (int64 and uint64 types) is fully
    supported, with saturation semantics like the other integer types.
    Performance of most integer arithmetic operations has been
    improved by using integer arithmetic directly.  Previously, Octave
    performed integer math with saturation semantics by converting the
    operands to double precision, performing the operation, and then
    converting the result back to an integer value, truncating if
    necessary.

 ** Improvements to the norm function.

    The norm function is now able to compute row or column norms of a
    matrix in a single call, as well as general matrix p-norms.

 ** New functions for reading and writing images.

    The imwrite and imread function have been included in Octave.
    These functions require the GraphicsMagick library.  The new
    function imfinfo provides information about an image file (size,
    type, colors, etc.)

 ** New functions for computing some eigenvalues or singular values.

    The eigs and svds functions have been included in Octave.  These
    functions require the ARPACK library (now distributed under a
    GPL-compatible license).

 ** Changes to strcat.

    The strcat function is now compatible with Matlab's strcat
    function, which removes trailing whitespace when concatenating
    character strings.  For example

      strcat ('foo ', 'bar')
      ==> 'foobar'

    The new function cstrcat provides the previous behavior of
    Octave's strcat.

 ** Specific sparse matrix functions removed.

    The following functions, which handled only sparse matrices have
    been removed.  Instead of calling these functions directly, you
    should use the corresponding function without the "sp" prefix.

      spatan2     spcumsum  spkron   spprod
      spchol      spdet     splchol  spqr
      spchol2inv  spdiag    splu     spsum
      spcholinv   spfind    spmax    spsumsqk
      spcumprod   spinv     spmin

 ** New QR and Cholesky factorization updating functions.

      choldelete  cholshift   qrdelete  qrshift
      cholinsert  cholupdate  qrinsert  qrupdate

 ** New quadrature functions.

      dblquad  quadgk  quadv  triplequad

 ** Other miscellaneous new functions.

      addtodate          interp1q                    rectint
      bicgstab           isdebugmode                 regexptranslate
      cgs                isfloat                     restoredefaultpath
      command_line_path  isstrprop                   roundb
      contrast           log1p                       rundemos
      convn              lsqnonneg                   runlength
      datetick           matlabroot                  saveobj
      display            namelengthmax               spaugment
      expm1              nargoutchk                  strchr
      filemarker         pathdef                     strvcat
      fstat              perl                        subspace
      full               prctile                     symvar
      fzero              quantile                    treelayout
      genvarname         re_read_readline_init_file  validatestring
      hypot              reallog                     which
      idivide            realpow
      info               realsqrt

See NEWS.3 for old news.