Version 0.9.10

- Migration advices (due to some API and compilation process changes)

  * as MLGMPIDL is not included anymore, you may want to consider
    - either apron-dist-XXX.tgz packages
    - or svn co http://svn.cri.ensmp.fr/svn/apron/apron-dist/trunk apron-dist
    that include it.
	
  * Makefile.config: 
     new variables HAS_SHARED and OCAMLMKLIB

  * C API: 
     ap_pkgrid_manager_alloc has a new signature

  * OCaml API: 
     PolkaGrid.manager_alloc_(loose|strict) are replaced by
      PolkaGrid.manager_alloc with a new signature.

  * OCaml compilation and Makefiles
     Typical commands

      ocamlc -o mlexample.byte 
        bigarray.cma gmp.cma apron.cma box.cma -cclib "-lboxMPQ" mlexample.ml
      ocamlopt -o mlexample.opt
        bigarray.cmxa gmp.cmxa apron.cmxa box.cmxa -cclib "-lboxMPQ" mlexample.ml
      are replaced by

      ocamlc -o mlexample.byte 
        bigarray.cma gmp.cma apron.cma boxMPQ.cma mlexample.ml
      ocamlopt -o mlexample.opt
        bigarray.cmxa gmp.cmxa apron.cmxa boxMPQ.cmxa mlexample.ml

- Distribution/Packaging

  * MLGMPIDL not included anymore.
    If you want to include it with apron, consider: 
    - either apron-dist-XXX.tgz packages
    - or svn co http://svn.cri.ensmp.fr/svn/apron/apron-dist/trunk apron-dist

  * Optional interface to PPL upgraded to 0.10 (but still support only
    polyhedra and linear congruences). No need with version 0.10 to apply a
    patch (see ppl/README for more details.

- Compilation and linking process

  * Support for dynamic libraries added (new HAS_SHARED flag in
    Makefile.config)

  * IMPORTANT CHANGE for OCaml compilation:
    (see migration advices)

    - OCaml library files box.cma, polka.cma become now
      boxMPQ.cma, boxRll.cma, polkaRll.cma, etc... 
      
       No need any more to use the -cclib option to choose the
        underlying number representation.

       Former box.cma, oct.cma, polka.cma are soft links to MPQ
        versions (for compatibility

    - Use of dynamic libraries when possible, with -dllib feature of
      ocamlc for bytecode executables.

       No need for custom bytecode interpreter and toplevel
       Enable the use of pretty-printer depending on C code in ocamldebug

- API

  * Better (correct) signature for Abstract0.permute_dimensions:
    old: permute_dimensions : 'a Manager.t -> 'a t -> Dim.perm option -> 'a t
    new: permute_dimensions : 'a Manager.t -> 'a t -> Dim.perm -> 'a t

  * pkgrid/PolkaGrid product library: new signature for the manager
    allocation function

  * New ap_dimchange2_t datatype, for specifying addition followed by removal
    of dimensions. Generated by ap_environment_dimchange2 function and used by
    ap_abstract0_apply_dimchange2 function.

  * Octagon domain now supports hashing

  * Newpolka library: real canonical form for strict polyhedra, compatible
    with the hash function.

  * OCaml API for domains (box,oct,polka,ppl,polkaGrid): 
    new functions allowing type conversion (for instance between
    Oct.t Apron.Manager.t and 'a Apron.Manager.t).

- Bugs corrected: several

- Internal API

  * Change of signature for itv_join (now requires itv_internal_t* object);
    
Version 0.9.9:

- API: 

  * Important change: tbool_t replaced by bool in tests in flags
    exact and best (C modules ap_abstractX and ap_manager, OCaml
    modules AbstractX and Manager)

    In source code, 
    * tbool_top,tbool_false should be replaced by false,
    * f(...)==tbool_true by f(...)
    * f(...)!=tbool_true by !f(...)

  * MPFR numbers added as possible numerical types (besides GMP
    MPQ and C double)

  * hash functions added in most places:
     * bound_hash, itv_hash, linexpr0_hash, texpr0_hash,
       ap_abstract0_hash, ap_abstract1_hash, ap_environment_hash,
       ap_var_hash
     * in libraries, hash implemented for itv and newpolka 
       (for others, raises not implemented).
     * in OCaml API, used as a custom hash function for
       corresponding abstract types.

  * Functions added
     * OCaml: Linexpr0.of_list, Linexpr0.of_array,
       Environment.lce_change, Environment.rename_perm
          
  * Internal functions added/modified
    * num: num_integer, numrat_inv added
    * apron: spec of ap_texpr0_max_dim modified

- Distribution/Packaging
  * Instruction files REAME.windows and README.mac files, some
    problems corrected.
  * Makefile.config.model simplified

- Bugs corrected: several, in particular:
  * Special cases with abstract values in dimension 0, and null
    arrays in nassign/substitution functions and
    add_dimensions/remove_dimensions
  * ap_fpu_init for Mac
  * itv_lincons_reduce_integer (normalisation of integer constraints)

Version 0.9.8
  * ap_linexpr0_compare
- New C++ interface, still experimental 

- Instructions for compilation under Windows

- Clarification of license issues related to the PPL (under GPL license,
  whereas APRON is LGPL)

- Several bugs corrections

Version 0.9.7

Many changes !

General:

- Addition of the reduced product of linear inequalities (from
  Polka library) and linear congruences (from PPL library).

- Addition of tree expressions and constraints, generalizing
  linear expressions and constraints with
  * non-linear operations (multiplication, division, square root, ...)
  * integer, rational/real and floating-point semantics

C interface:

- API:
  * Addition of ap_abstract1_unify function
  * Modification of the signature of ap_environment_remove

- some (redundant) functions removed from the API requested from
  underlying domains, but still provided through the APRON
  interface (level 0 and 1) for compatibility reasons:
  * of_lincons_array
  * assign_linexpr, substitute_linexpr (superseded by
    assign_linexpr_array, substitute_linexpr_array).

- Suffixes of C libraries names indicating internal number
  representation made more systematic and uniform among various
  underlying libraries:

  * Il, Ill, MPZ: denotes resp. long int, long long int, mpz_t (GMP)
  * Rl, Rll, MPQ: denotes resp. rationals using long int, long
                  long int, and mpq_t (GMP)
  * D, Dl: denotes resp. double and long double

  C library name without suffix (eg, libpolka.a) corresponds to
  the default suffix MPQ.

- Other changes in library organisation:

  * libitv.a integrated in libapron.a 
  * libapron_ppl.a renamed in libap_ppl.a,
    libppl_caml.a  renamed in libap_ppl_caml.a 

OCaml interface
- Linking simplified thanks to the change in C libraries (see above).
- API:
  * Addition of Environment.lce, Abstract1.unify (Abstract1.unify_with) functions
  * Modification of the signature of Environment.remove

Version 0.9.6

OCaml interface
- PPL module
- Parsing extended for congruence equalities and congruence generators.

General:
- Addition of PPL *convex polyhedra* and *linear congruences*
  (grids) abstract domains
- Some internal reorganisation
  apron/ap_linearize module added
- Full support for interval linear expressions and linearisation to
  (quasi-)linear expressions in NewPolka and PPL libraries.  Well
  tested for quasilinear expressions, not yet for full interval
  linear expressions.
- Many bugs corrected.

Version 0.9.5

OCaml interface
- Better type systems, similar to the system used by the [Bigarray] 
  library of the OCaml standard distribution.
- Generation and installation in $(APRON_PREFIX)/bin of apronrun
  and aprontop executables in the main Makefile.

C interface:
- Use of const qualifier removed in the API, except for strings
  (const char *)

General:
- mpfr library now required in addition to gmp library.
- conversions between mpq_t and double are now safer, thanks to the use of
  mpfr.
- itv becomes box.
- package num extended with a new package itv, used currently by box.
- support for linear congruences added in constraints and generators.
- NewPolka offers a layer for linear equalities only, in addition to more
  general linear inequalities.
- Improved Makefiles and Makefile.config.model
- Bugs corrected, in particular in Box and Newpolka.

Version 0.9.4

- Modification of Makefile and Makefile.config.model

- Updated documentation

- OCaml linking: apron.cmo, polka.cmo, oct.cmo... become libraries
  apron.cma, polka.cma,... that reference needed C libraries, with
  the exception of libpolkaXX, liboctXX, ... This is documented in
  READMEs and doc.

- mlapronidl: Added string parsing (module Parser)

- newpolka: 

  * Stronger normalization in widening for strict constraints

  * Bug fixed (some of them were severe)

