
 0.1  first version, two functions "matrix2poly" and "scdd". 

      "matrix2poly" the former is deprecated and should be removed

      "scdd" successfully does conversion between H-representation and
      V-representation (both ways) using real arithmetic (as opposed
      to other two possibilities: integer arithmetic or GNU multiple
      precision arithmetic)

      To do: other arithmetic modes are essential

 0.2  second version, function "matrix2poly" removed.

      functions "d2q", "q2d", "q2q", and "z2q" added -- conversion between
      IEEE floating point and GMP rationals.

      "scdd" now takes GMP rational arguments as well as IEEE reals.

      function "validcdd" -- validates H-representations and V-representations

      functions "validcdd", "makeH", "addH", and "dropH" added --
      construction and adjustment of H-representations

 0.3  many changes

      completely redesigned "cdd objects", now former "linearity" component
      is represented in first column of matrix, and former "representation"
      component is now "representation" attribute of matrix, so "cdd objects"
      are just matrices having a certain form (first column of
      H-representation is zero-or-one valued, first two columns of
      V-representation are zero-or-one valued)

      removed "dropH", replaced "addH" by "addHeq" and "addHin"
      former "dropH" can now just be done by matrix indexing,
      e. g., if "qux" is cdd object, then qux[-3, ] drops row 3.
      added argument "x" to "makeH" so "addHeq" and "addHin" can
      be implemented simply as calls to "makeH"

      function "qmq" added -- subtraction of GMP rationals -- useful
      for exact comparison.

      also completely revised and expanded description in inst/doc/design.tex

 0.4  GPL, Makevars

 0.5  changed name of package from cdd to rcdd per request from Fukuda

 0.6  added maximal, all.intersect

      added convex hull example to scdd (how absurdly simple, how stupid
         of me not to have seen that).

      added qpq, qxq, qdq

 0.6-1  fixed bug in nonred.c, using uninitialized variables, found by
        valgrind

      cleaned up qoq.c, removed unnecessary initialization mpq_init
      always initializes to zero

      added autoconf

 0.6-2  autoconf changes

          * changed LIBS to LDFLAGS following autoconf conventions

      scdd (and scdd_f) change, H-representation now allowed to be
      matrix with zero rows (represents the whole space).  Important
      that this be allowed because scdd can produce such a representation
      (when the whole space is the convex polytope in question).
      OTOH an empty V-representation (representing the empty set)
      is NOT allowed because (1) scdd does not produce such and (2)
      error exits when given such.

      fixed z2q so it preserves attributes (expecially "dim"), just like
      the qoq functions.

      added rudimentary vignette

      modified error messages as per Writing R Extensions, Sec 1.8

 0.6-3  autoconf changes, now checks for libraries

 0.7  Gave up and copied all of Fukuda's code into the src directory.
      Now we should build under windoze.

      added NAMESPACE file

      fixed a bunch of complaints about the CHAR macro in Rinternals.h
      returning (const char *)

      fixed bug in documentation of scdd function

      copied cddlibman.pdf from cddlib distribution into inst/doc

      added qsum and qprod to GMP arithmetic functions

 0.7-1  added qsign and qmatmult to GMP arithmetic functions

      fixed ANOTHER bug in the documentation for scdd (before 0.7 V-rep
      was correctly explained but H-rep was incorrectly explained, a
      sign mistake; in 0.7 H-rep was correct, but V-rep made incorrect,
      sign change applied to both was wrong, now hopefully both correct).

 0.7-2  inserted dd_free_global_constants(); in scdd.c and scdd_f.c

 0.8  added lpcdd which does linear programming

      added init.c to "register routines" to protect against argument
      mismatch calling C from R

 0.9  added redundant which eliminates redundant rows from H-representation
      or V-representation

 0.10 added allface which produces list of faces from H-representation

      added several sections to package vignette describing new functions
      added since 0.7

 0.10-1 changed "typeof(hrep)" to "class(objgrd)" in computing default
        in signature of R function lpcdd as per e-mail from Brian Ripley

 0.10-2 fixed PROTECT/REPROTECT bugs in allfaces*.c and redundant*.c

        added BUGS section to lpcdd man page.

 1.0  upgraded from cddlib 0.94d to 0.94f.  This fixes a bug we found
      in lpcdd: the dual.direction returned when the problem had no
      feasible solution was incorrect.

      deleted BUGS section to lpcdd man page.

      minor improvements to vignette.

 1.0-1  minor changes requested by Uwe Ligges to make the Windows build work.

 1.1  fix major, major bug in lpcdd, when optimal solution found and when
      problem has equality constraints just 1.0-1 crashes when any Lagrange
      multiplier negative (which is allowed for equality constraint).

      add linearity, qneg, qinv, qabs.

      update vignette to use qneg, etc. as appropriate.

      more tests exercising lpcdd.

 1.1-1  fix major, major bug in lpcdd, when problem infeasible and when
        problem has equality constraints 1.1-1 crashes when any Lagrange
        multiplier negative (which is allowed for equality constraint) --
        essentially same bug fix as in 1.1 except for infeasible case
        (not quite the same, corner case of that case, but still a bug;
        also not quite a crash, "Can't happen" error message, but still bad)

        add bug2.R to tests to illustrate bug fix

 1.1-2  fix bug ? on p. 16 of vignette
        add version numbers to sec 1 of vignette

