2003/12/29 - 0.6.1
   * All calls to the calloc function have been replaced with mxCalloc() and the 
     same for the function free() with mxFree() due to a problem with the 
     Matlab memory manager.

   * Updated lower and upper bounds definition of the structural variables. If 
     some structural variables are not lower (upper) bounded you can define 
     an lb (ub) array with -Inf (Inf) value in the position of that variable. 
     If the i-th structural variable is unbounded (free) define lb[i]=-Inf 
     and ub[i]=Inf.
     If all structural variables are free LB=-Inf UB=Inf and LB=[], UB=[] are 
     equivalent definitions.
   
   * Free bounds bug fixed. If lower and upper bounds are empty ('[]') the mex 
     interface sets the bounds to LPX_FR. Thanks to Johan Lfberg.


2003/11/28 - 0.6
   * The mxGetNzmax function to get the number of non-zero elements 
     has been replaced with the last element of the jc array. 
     Actually the mxGetNzmax returns nzmax, an integer value      
     indicating the number of elements in the ir, pr. 
     Its value is always greater than (or equal to) the number of
     nonzero elements in a sparse mxArray. 
     Thanks to Johan Lfberg. 

   * Added the fault hook function. When in GLPK something goes wrong 
     the 'fault' function is called and the program execution is 
     terminated crashing Matlab. The hook redirects the output error
     directly to the Matlab shell and avoids Matlab from crashing.
     
   * Added the print hook function. Now all output sent by GLPK to the 
     standard output has been redirected to the shell of Matlab. To 
     avoid too much information on standard output the 'msglev' option 
     has been set at 1 (error messages only) as default.
     
   * The mex interface has been splitted in three separated files: 
     glpsets.h contains some useful definitions
     glpkfun.h contains the routine glpk for solving the problems,
     glpkmex.c contains the mexfunction
     
   * glpktest1.m, glpktest2.m, glpksparse.m updated.


2003/11/14 - 0.5.9
   * the default value of the scale parameter has been set to 1 
     (instead of 3). Until 4.2 the default value was 3. In 4.2 the 
     default value is 1. It was changed because equilibration 
     scaling seems to be much more appropriate for badly scaled 
     problems.

   * In the precompiled version the PDF parameters' guide has been 
     replaced with a .m file.

   * (under Windows) The libglpk.a of GLPK 4.2 used to produce the 
     glpkmex.dll has been compiled with mingw. The code is smaller 
     than cygwin code with the -mno-cygwin option.
     In both Windows and Linux, the -g -O2 options of the compiler 
     have been replaced with -O3 (maximum optimization, no debug infos).   

   * Added a matlab help file, glpkparams.m, for the set of params that
     can be passed to the mex interface.

2003/09/30 - 0.5.8

   * Added a list of parameters that the user can pass to the mex 
     interface.
   
   * Removed the warning section in the mexFunction in case the output 
     code is between 204 and 214.

2003/09/29 - 0.5.7

   * Added some informations when you type glpkmex with no argument, 
     i.e., '>>glpkmex'.
   
   * If the problem is not feasible you get now more output details
     This is very useful when you run Matlab with JVM.
   
   * The problem is saved before it is solved so if matlab 
     crashes you can study the output with glpsol (or CPLEX).

2003/09/16 - 0.5.6

   * Option SAVE changed. Instead of printing the solution of the 
     problem now you can print the problem itself. In detail, 
     lpx_write_lpt() has replaced lpx_print_sol(). The output file 
     (in CPLEX LP format) can be used with both glpksol and cplex.
   

2003/08/26 - 0.5.5 

   * Previuos version

