2002-05-10  Dirk Mueller  <mueller@kde.org>

	* removed board_gen subdir and test_* files as they appear to 
	contain code that is possibly copyrighted by Microsoft. 

From 1.6.1 to 1.6.2
-------------------

* The beginning of the mhash optional utilization.

* Fixed the solution optimization-related bug that I saw when running
kpat. mtest works fine.

From 1.6.0 to 1.6.1
-------------------

* Removed CARD_DEBUG_PRES from the makefile.

From 1.5.18 to 1.6.0
--------------------

* Some corrections in the documentation.

* Made the makefile a release one.

* Modified config.h for release. (COMPACT_STATES and all)

From 1.5.17 to 1.5.18
---------------------

* Documented the "-opt" switch in USAGE and modified to reflect the
fc-solve and fc-solve-debug merge.

* Got libfcs.a to compile with the FCS_WITH_TALONS #ifdef turned off.

From 1.5.16 to 1.5.17
---------------------

* Placed all the talon-related code in an #ifdef.

From 1.5.15 to 1.5.16
---------------------

* Klondike solving is now working apparently flawlessly, but it does not
seem to get to actually solving the board. Maybe with the talon present 
there are too many possible states for it to work.

From 1.5.14 to 1.5.15
---------------------

* Got the Klondike Solving to work, but with some kind of free()-related
error in Win32. In Linux it runs without segfaulting, but it seems to get
stuck often.

* Applied the NetBSD MD5 patch.

From 1.5.13 to 1.5.14
---------------------

* Quite a few of Klondike-related modifications. The tests to handle
the Klondike talon are not working yet, though.

From 1.5.12 to 1.5.13
---------------------

* Made the cards in the Klondike board generated by 
make-aisleriot-freecell-board hidden.

* Created a cache for the talons in instance.

* Added some macros to handle the Klondike talon in state.h

From 1.5.11 to 1.5.12
---------------------

* Modified make-aisleriote-board so it will output Klondike boards too.

* Fixed a bug in which make-aisleriot-board switched between Clubs and
Spades (I've got to backport this to 1.4.x)

* Modified make_pysol_freecell_board.py to generate PySol's Klondike board.

* Added game generation for several other Klondike clones that use the same 
layout into make_pysol_freecell_board.py.

From 1.5.10 to 1.5.11
---------------------

* Replaced mymalloc with plain malloc.

* Got the source to compile with COMPACT_STATES.

* Got the source to compile with DEBUG_STATES.

From 1.5.9 to 1.5.10
--------------------

* Fixed the memory leaks caused by the path optimization and possibly
several indirect stacks related ones that were there earlier.

* Fixed a bug in check_and_add_state that affected DFS scans.

From 1.5.8 to 1.5.9
-------------------

* Got suspend and resume to work with the path optimization.

* Made library bindings to the path optimization.

* Added the talon parameter to check_state_validity in lib.c.

* Merged fc-solve and fc-solve-debug into one source.

* Cleaned up the Makefile.

* Updated Makefile.lite

From 1.5.7 to 1.5.8
-------------------

* Added some comments in many places.

* First (and possibly non-working) implementation of the BFS path
optimization.

* Got the BFS optimization to work. With PySol board No. 980662 it reduced
the number of moves from 3724 to 219 (!).

From 1.5.6 to 1.5.7
-------------------

* Fixed a bug in the Soft-DFS scan. Sometimes when doing the last test, the
scan left the current depth, while there were other states waiting. 
(Backported into 1.4.5)

From 1.5.5 to 1.5.6
-------------------

* Added comments to fcs_hash.c.

* Modified the rehashing code in fcs_hash.c so that the linked list items 
that belong to the old hash will be recycled. I don't seem to run out of 
hash optimizations, do I?

* Removed some redundant code in alloc_instance() that was specific to
FCS_STATE_STORAGE_INDIRECT.

* Added some comments to intrface.c.

* Removed a redundant function (soft_dfs_add_state), a redundant #if 0 and
some other stuff from intrface.c.

From 1.5.4 to 1.5.5
-------------------

* Many changes, too numerous to mention or recall, but the end result is
that solving Gypsy games is working. However, it is not operational as it
generates too many intermediate states and does not seem to finish.

* Some comments added, mainly at scans.c.

From 1.5.3 to 1.5.4
-------------------

* Modified fcs_stack_compare_for_comparison so it will correspond to the
order of fcs_stack_compare_for_stack_sort.

* Specifiying type of talon in the command line arguments.

From 1.5.2 to 1.5.3
-------------------

* Changed fcs_initial_user_state_to_c so it will read Talons. Not tested
yet.

From 1.5.1 to 1.5.2
-------------------

* Fixed some broken indentation in lib.c.

* Added some parenthesis arenthesis in lib.c's if's.

* Modified card.c, state.h and state.c to read flipped cards. (in
preparation for the Gypsy/Klondike type games integration)

* Added support for generating Gypsy games in make_pysol_freecell_board.py

* Added a gypsy_talon and gypsy_talon_len members to instance.

* Changed deck to suit or foundations as appropriate in the code. It
compiles and runs in all three modes of storage.

* Started covering the state storage in Architecture.txt.

From 1.5.0 to 1.5.1
-------------------

* Added the comments in fcs.h

* Removed the macro INDIRECT_STATE_STORAGE and replaced it with 
FCS_STATE_STORAGE == FCS_STATE_STORAGE_INDIRECT.

Looks good.

From 1.4.4 to 1.5.0
-------------------

* Added an optimize for caching option when adding elements to the 
internal hash. It is on in the code.

From 1.4.3 to 1.4.4
-------------------

* Placed the Priority Queue and BFS Queue de-allocation in 
free_instance() instead of finish_instance().

* Placed the Architecture.txt and Architecture.summary.txt documents in
devel_docs.

From 1.4.2 to 1.4.3
-------------------

* Made sure tat finish_instance is not called if user_solve was not called
in lib.c

From 1.4.1 to 1.4.2
-------------------

* Applied the patch from Markus Oberhumer. Basically, some spelling errors
and some const stuff.

From 1.4.0 to 1.4.1
-------------------

* Set proto_solution_moves and solution_states to NULL in two places
they were not set to. (Thanks Markus Oberhumer!)


From 1.3.42 to 1.4.0
--------------------

* Win32 adaptation to the new module organization. 

* And finally, a new major version!

From 1.3.41 to 1.3.42
---------------------

* Another patch from Stephan Kulow - this time regarding int32_t in md5.h.

* Replaced malloc.h by stdlib.h.

* Made the inclusion of search.h into an #ifdef dependant on libredblack only.

* A makefile fix regarding DLFLAGS.

From 1.3.40 to 1.3.41
---------------------

* Applied Stephan Kulow's new patch. Some const char * stuff and additions
to the user_ library.

* Set solution_states to NULL in one place it was missing in lib.c

From 1.3.39 to 1.3.40
---------------------

* Set proto_solution_moves and solution_states to NULL where they were
freed, and added a call to free for them in finish_instance.

* Some changes to the documentation.

* Set solution_states to NULL in the main function of fc-solve.

* Ditto for test_mulit.c.

From 1.3.38 to 1.3.39
---------------------

* Added memory-leak fprintf's to the move stack creation/deletion functions.
They are cancelled with an #ifdef, but I'm keeping them for the while.

* De-allocated the move stacks that were left by the previous test function
in soft_dfs_do_solve_or_resume.

From 1.3.37 to 1.3.38
---------------------

* Added a deallocation to the states_to_check arrays between
num_solution_states to dfs_max_depth.

From 1.3.36 to 1.3.37
---------------------

* Added de-allocation routines to the soft-dfs stuff. Right now, mtest
crashes from time to time, but I have to figure out why.

* Well, the soft-dfs cleaning up is working and no segfaults happen with
mtest or fc-solve.

* I placed the total move stack creation on a separate function because it
is used in both solve_instance and resume_instance.

From 1.3.35 to 1.3.36
---------------------

* Modified the file USAGE to mention the method choosing and the A* 
weights.

* Added a soft_dfs_num_free{stacks,cells} array to the instance data
structure, and modified the algorithm to use it so it won't re-calculate
them time after time.

* Note: I don't seem to free the soft_dfs related arrays. So I should.

From 1.3.34 to 1.3.35
---------------------

* Removed some unnecessary "#if 0"'s and "#if 1"'s.

* Added Soft-DFS support for the other state storage implmentations.

* Created a preprocessor FCS_STATE_STORAGE enum.

* Converted internal hash to use FCS_STATE_STORAGE.

* Converted libavl's AVL and Red-Black tree to use FCS_STATE_STORAGE

* Converted libredblack to the enum.

* Converted glib's tree and hash to FCS_STATE_STORAGE.

* Converted DB-File to FCS_STATE_STORAGE

From 1.3.33 to 1.3.34
---------------------

* Some head comments to the new files.

From 1.3.32 to 1.3.33
---------------------

* Placed the interface functions of the instance handling routines in
intrface.c.

* Placed the Scan-specific functions in scans.c.

* Renamed dfs.c to caas.c.

From 1.3.31 to 1.3.32
---------------------

* Removed calls to free for the solution_states in the Soft-DFS unresume
function.

From 1.3.30 to 1.3.31
---------------------

* Removed the allocation of instance->solution_moves in solve_for_state(). A
minor bug.

From 1.3.29 to 1.3.30
---------------------

* Various changes to make the code more ANSI C/ANSI C++ compatible. Thanks 
should go to Stephan Kulow!

* Added a library function to tell what is the current depth.

From 1.3.28 to 1.3.29
---------------------

* Added some calls to fcs_move_stack_destory() in the pre-mature breakup
of move_freecell_cards_to_founds and move_top_stack_cards_to_founds.

From 1.3.27 to 1.3.28
---------------------

* The multi-line macros in state.h are now sorrounded by curly brackets 
("{ ... }").

From 1.3.26 to 1.3.27
---------------------

* Found out why the Soft-DFS and hard-DFS solutions of PySol 980662 are
different - it's because of the state-is-not-solveable stuff. So everything
is OK. BTW, it seems to make matters much better in case of this board.

* Fixed num_times to correspond with the standard Hard-DFS definition.

* Removed DIRECT_STATE_STORAGE. Less #ifdefs and less things to think about
, so it seems like a good idea.

From 1.3.25 to 1.3.26
---------------------

* The problem exists on RedHat 7.0 (or RH6.2) too.

* Solved - current_state_indexes was not set to zero when advancing to a 
higher depth.

From 1.3.24 to 1.3.25
---------------------

* Fixed a bug in Soft-DFS that the states starting from the first successful
one were not checked.

* Fixed a bug in Soft-DFS that the move stack of the index+1 was selected 
in the current depth. Now it's operational.

* There's a bug when solving the notorious PySol 980662.

* When I ran it on Mandrake 6.0 with MS board No. 24, it get into an
infinite loop that shuouldn't happen in any case. When I link it with
ElectricFence it rans fine. - What gives?

From 1.3.23 to 1.3.24
---------------------

* Added the perl scripts and the module for analyzing the malloc/free
memory leak.

From 1.3.22 to 1.3.23
---------------------

* Added the test_multi.c target to test for several games in one run.

From 1.3.21 to 1.3.22
---------------------

* Made sure that the Soft-DFS specific code is called in the move top stack
cards to founds and move freecell cards to founds test. It causes a segfault
or something like that and I have to find out why.

From 1.3.20 to 1.3.21
---------------------

* Fixed a fc-solve-debug bug that in soft-dfs the states were printed once
for each test.

From 1.3.19 to 1.3.20
---------------------

* Fixed some bugs. The improved Soft-DFS is working, but takes too much 
memory. I have some idea how to improve it, but I'm not sure they will be
good in the long run.

From 1.3.18 to 1.3.19
---------------------

* Added support for -s -i to the improved soft-DFS code.

From 1.3.17 to 1.3.18
---------------------

* A non-working but on the other hand non-segfaulting version of the 
improved soft-DFS code.

From 1.3.16 to 1.3.17
---------------------

* Implemented the improved soft-DFS do_solve_or_resume function.

* TODO: Cleanup and modify the tests accordingly.

From 1.3.15 to 1.3.16
---------------------

* Added re-allocation and init to NULL to the new soft_dfs variables.

* Added an states_to_check move stacks variable.

* Wrote pseudocode for the new asynchronous soft-DFS algorithm.

From 1.3.14 to 1.3.15
---------------------

* Added soft_dfs_{states_to_check,num_states_to_check,current_state_index} to
instance. Improved Soft DFS coming through!

* Added soft_dfs_test_indexes. Just for good measure, so I won't collect too
many states in the stack of arrays.

* TODO: Make them grow in size according to the depth.

From 1.3.13 to 1.3.14
---------------------

* Added the seqeunce length over renegade cards' weight to A*. It seems to 
make matters only worse in most cases, so I disabled it by default.

From 1.3.12 to 1.3.13
---------------------

* Ported the fcs_move_* macros to FCS_DEBUG_MOVES.

From 1.3.11 to 1.3.12
---------------------

* Implemented A* and BFS resume.

* Removed some parameters from freecell.c's solve and resume functions.

From 1.3.10 to 1.3.11
---------------------

* Merged the code of the A* and BFS scans.

* Increased the A* rating to INT_MAX.

From 1.3.9 to 1.3.10
--------------------

* Added pqueue.c and pqueue.h to Makefile.lite

* Added the cleaning code for the move stacks for the state with locations.

* Deallocated the priority queue and the BFS linked list.

* Fixed a segfault in BFS, by not deallocating the move stack in the
appropriate place.

From 1.3.8 to 1.3.9
-------------------

* Modified unresume_instance() so it would do the deallocation only
in DFS and Soft-DFS.

From 1.3.7 to 1.3.8
-------------------

* Implemented a move as a 4-byte data-structure, and modified the code 
accordingly.

From 1.3.6 to 1.3.7
-------------------

* Added a call to free for hash->entries in fcs_hash.c.

* And another Internal Hash bug - num_elems should be assigned to the
rehashed hash too.

From 1.3.5 to 1.3.6
-------------------

* Input the A* weights from the command line. No documentation for it yet.

From 1.3.4 to 1.3.5
-------------------

* Made the A* weights run-time paramters.

* Placed them in an array.

From 1.3.3 to 1.3.4
-------------------

* Modified "new_size" to "new_size+1" in pqueue.c - This solves the segfault
on Linux.

From 1.3.1 to 1.3.3
-------------------

* Fixed a bug with assigning the parent and the move stack to the new state.

* Added the priority queue code.

* Working A* code. When running on Linux segfaults on PySol board No.
980662.

From 1.3.0 to 1.3.1
-------------------

* Added the parent and moves_to_parent members to
fcs_state_with_locations_t.

* A non-working BFS implementation.

From 1.2.0 to 1.3.0
-------------------

* Made the choice between DFS and Soft-DFS a runtime option.

From 1.1.14 to 1.2.0
--------------------

* Checked that it works with libredblack and libavl (and both). Looks good.

* A fix for md5.h to make it compile cleanly on Solaris.

* Updated the file config.h.dkds.

* Modified the files README and INSTALL.

From 1.1.13 to 1.1.14
---------------------

* Removed the fcs_tree macros and enabled an application to use libavl's 
AVL tree for the states and its red-black tree for the stacks (or vise
versa).

From 1.1.12 to 1.1.13
---------------------

* Modified the makefile to enable the different stack allocation mechanisms.

* TODO: test it against a working installation of libavl and libredblack.

From 1.1.11 to 1.1.12
---------------------

* Implemented indirect stack storage with Glib's balanced tree.

* Implemented indirect stack storage with Glib's Hash Table.

From 1.1.10 to 1.1.11
---------------------

* Implemented indirect stack storage with libavl (AVL and Red-Black trees).

* Implemented indirect stack storage with libredblack.

* TODO: Modify the makefile to support this.

From 1.1.9 to 1.1.10
--------------------

* Added a common function for SOFT_DFS resume and solve_for_state.

* Coded a macro fcs_caas_check_and_insert which contain the state storage
implementation specific details. Now there is one common check_and_add_state
function.

From 1.1.8 to 1.1.9
-------------------

* Made sure that if only kings can be on empty stacks, then I won't descend
into the rest of the function in move_sequences_to_free_stacks.

* Some Soft-DFS fixes to make the resume work. Not finished yet.

* A possibly working Soft-DFS resume_solution function. TODO: merge the code
 of it and solve_for_state into one function.

* Added the "#include <string.h>" line to the file lib.c.

* Tested the Soft-DFS resume. Looks good!

From 1.1.7 to 1.1.8
-------------------

* More robust argument handling. No more segfaults - hopefully.

From 1.1.6 to 1.1.7
-------------------

* Improved move_sequences_to_free_stacks so it will also move sequences
that are under some cards. (quite rare but possible.)

* TODO: If only kings can be on empty stacks then don't continue unless
this_card is king.

From 1.1.5 to 1.1.6
-------------------

* Improved the move_freecell_cards_on_top_of_stacks so it will move freecell
cards on top of non-top stack cards.

From 1.1.4 to 1.1.5
-------------------

* Added the Win32 makefiles.

From 1.1.3 to 1.1.4
-------------------

* Soft-DFS organization and integeration with the main solution model. No
resume though.

From 1.1.2 to 1.1.3
-------------------

* A working implementation of Soft-DFS (i.e. without using function recursion)
. Now to make it conform to the regular conventions.

From 1.1.1 to 1.1.2
-------------------

* Removed the check for cards under parent in
move_stack_cards_to_different_stack because it is already handled by the 
c and seq_end routines.

From 1.1.0 to 1.1.1
-------------------

* Fixed the seq_end code of move_stack_cards_to_different_stack for
sequences at the end of the stack.

* The c loop of move_stack_cards_to_different stacks now skips to seq_end+1,
so it saves iterations.

* Added a "#include <string.h>" line to preset.c.

From 1.0.2 to 1.1.0
-------------------

* Added a CREDITS file.

* Changed the makefile to delete test-lib and libfcs.a.

* Modified move_stack_cards_to_different_stacks so it will also move source 
sequences that are below some cards.

* Checked if Freecell Solver solves MS-Freecell board No. 7477 and that Seahaven Towers
board Tom Holroyd sent me. It does.

* Checked half of the solution of 7477 to see if it's valid. It looks OK.

From 1.0.1 to 1.0.2
-------------------

* Changed fcs_card_deck() to fcs_card_card_num() in the checks for whether
to return FCS_STATE_ORIGINAL_STATE_IS_NOT_SOLVEABLE. A major bug!

From 1.0.0 to 1.0.1
-------------------

* Added an "#include <string.h>" line to fcs_hash.c.

* Added a u_int32_t typedef for WIN32 in md5.h/

* Added an initialization of debug_iter_output in alloc_instance().

* Added an initialization of debug_context.debug_iter_state_output in 
main().

From 0.11.14 to 1.0.0
---------------------

* Some changes to the documentation.

* New Initial digit!

From 0.11.13 to 0.11.14
-----------------------

* Some changes to the TODO file.

From 0.11.12 to 0.11.13
-----------------------

* Added the files "config.h.dkds" and "config.h.freecell"

* Added documentation about them and about INDIRECT_STACK_STATES to 
INSTALL.

* Fixed some typos in INSTALL.

From 0.11.11 to 0.11.12
-----------------------

* Optimized the check_and_add_state() of libredblack. Now there is only one
call to the rb functions instead of two.

* Checked that libredblack is working. It does.

* Modified the file "README".

* Modified the file "INSTALL".

* Added headers to the new files.

From 0.11.10 to 0.11.11
-----------------------

* Fixed the function move_top_stack_cards_to_founds and 
move_freecell_cards_to_founds.

* Fixed the resume code in the check_and_add_state() function of 
DIRECT_STATE_STORAGE. I have to fix all the other storages!

* Fixed all the check_and_add_states(). libredblack was not tested yet
because I don't have it installed here.

From 0.11.9 to 0.11.10
----------------------

* Added the lib.c and fcs_user.h files to be the foundation of the 
user library.

Not tested yet.

From 0.11.8 to 0.11.9
---------------------

* Added the ignore_osins (ignore original state is not solveable) paramter
to the sfs tests.

* Modified the code of move_top_stack_cards_to_founds and 
move_freecell_cards_to_founds to make use of this parameter.

From 0.11.7 to 0.11.8
---------------------

* Added a de-allocation of the moves stack in solve_for_state_resume.

* Added the function freecell_solver_unresume_instance().

* Modified main.c to make use of the new interface.

From 0.11.6 to 0.11.7
---------------------

* Added a de-allocation to the memory allocated for the stacks in
INDIRECT_STACK_STATES.

* Added the premature-termination of a solving process and its resuming.
Checked it and it works OK.

From 0.11.5 to 0.11.6
---------------------

* Made sure the loop over the decks in sfs_move_top_stack_cards_to_founds
took consideration of d. (it was a bug).

* Handled the case of sequences built by rank.

* Same for move_freecell_cards_to_founds.

From 0.11.4 to 0.11.5
---------------------

* Added the FCS_NON_DFS compile-time option, which is currently applicable
only for libavl and for COMPACT_STATES or DEBUG_STATES. 

Not recording the states only seem to make matters much worse.

From 0.11.3 to 0.11.4
---------------------

* Implemented the function fcs_move_to_string

* Implemented the "-m" option.

From 0.11.2 to 0.11.3
---------------------

* Changed sequences_are_built_by_suit to sequences_are_built_by with three
options: rank, suit and alternate colours.

* Added the Baker's Dozen preset.

* Added the Good Measure preset.

* Added the Cruel preset. (I don't support the re-deal feature for the 
while).

From 0.11.1 to 0.11.2
---------------------

* Modified the presets code so it will not exceed the hard-coded limits
of stacks, freecells, etc.

From 0.11.0 to 0.11.1
---------------------

* Code cleaning and organization. Made sure all STORAGE's worked OK with
the INDIRECT_STACK_STATES.

From 0.10.0 to 0.11.0
---------------------

* Changed "die_schalange" to "die_schlange".

* Implemnted INDIRECT_STACK_STATES. At the moment, present only with
libavl.

* Solution of several Die Schlange boards is fine.

From 0.9.15 to 0.10.0
---------------------

* Added the new options to the help string.

* Modified the file USAGE.

* Modified the file board_gen/make_pysol_freecell_board.py to generate
boards for the new games.

* Modified user_state_to_c to input several decks in the foundations.

* Fixed a nasty bug in fcs_Check_state_Validity()


From 0.9.14 to 0.9.15
---------------------

* Added the preset module and interfaced it with main().

From 0.9.13 to 0.9.14
---------------------

* Added opening comments to move.h, move.c and dfs.c

* Added some comments in dfs.c.

* Added an opening comment in fcs_hash.c and fcs_hash.h

* Added some extern "C" { ... } wrapping around the header file.

* Removed the check to FCS_CANONIZE_STATE in main,c

* Added some initializations in freecell_solver_alloc_instance().

* Removed many of the local variables of main(). I now use instance
and debug_context to store all of the dynamic information.

From 0.9.12 to 0.9.13
---------------------

* Added the fcs_move_stack_get_num_moves() function.

* Added the fcs_move_stack_duplicate() function.

* Replaced the stack swallowing code in fcs_move_stack_normalize with 
  a call to fcs_move_stack_swallow().

* Updated Makefile.lite.

From 0.9.11 to 0.9.12
---------------------

* Modified user_state_to_c() so it will accept empty freecells by
specifiying them with a "*" or "-".

* Modified check_state_validity() so it will report in case there's a hole
in one of the stacks.

* Modified main() accordingly.

* Created the fcs_move_stack_normalize() function, so end-users of the
library woould not need to canonize the intermediate states.

From 0.9.10 to 0.9.11
---------------------

* Wrote the perl script and the one state per line output seems to be good
on many test boards I checked.

* Changed the arguments of fcs_apply_move so freecells_num will come before 
stacks_num and decks_num.

* Removed the part about recursion states state output in main, and made
the moves state output to stdout.

* Fixed a bug in which fcs_move_stack_destroy was called twice in main for
the same stack.

* Changed the move stack growth to the maximum of 16 and 1/8 of its current 
size.

* Placed check_and_add_state() inside the file dfs.c.

From 0.9.9 to 0.9.10
--------------------

* Implemented the FCS_MOVE_TYPE_CANONIZE solution and it seems to be
working.

* Added some missing fcs_move_stack_reset() calls.

* Output from the 0000.board seems fine, but diff is not very helpful. I'm
going to write a perl script to convert each state into a single line, so it
will be more sensible. I'll also try playing a game.

From 0.9.8 to 0.9.9
-------------------

* Added a move loop and dump into main.c.

* I realized that the problem with the move bugs is that the move refer to
the canonized indexes of the stacks and freecells. I have two ideas how to
solve this problem:

1. Add a reverse_stack_locs and reverse_fc_locs array to the
state_with_locations and use it to lookup the real index.

2. Add a FCS_MOVE_TYPE_CANONIZE move and operate on the canonized indexes
of the stacks and freecells.

Currently I'll try #2, and see if it works out. Anyway I'm going to close a
version so I'll later may be able to start from this position and try #1.

From 0.9.7 to 0.9.8
-------------------

* Fixed a place in move_stack_cards_to_a_parent_on_the_same_stack
where I should have used fcs_is_parent_card.

* Started implementing the moves and move stacks.

* Added move support (probably buggy) into freecell.c.

From 0.9.6 to 0.9.7
-------------------

* Fixed a bug in the parsing of the "FC: " line if there is a trailing
whitespace.

* Added the internal hash implementation with the MD5 hash function. It runs
really quickly. (faster than libavl's AVL tree).

From 0.9.5 to 0.9.6
-------------------

* Added the empty_stacks_fill option in the instance struct.

* Added appropriate bindings to the main function.

* Added the code for the Empty stacks filled by kings only and Empty stacks
not filled at all.

* Tested on Forecell board No. 26, and the solution is correct.

From 0.9.4 to 0.9.5
-------------------

* Tested the solver on a Baker's Game board. Looks good!

* Added the startegy documents inside the devel_docs sub-dir. I'll
remove them towards the final release.

* Added code for Berkeley DB files. The database creation is not working for
some reason.

* Implemented the unlimited (or relaxed) sequence move. It can solve game
#11982 of MS-Freecell!

From 0.9.3 to 0.9.4
-------------------

* Added a sequences_are_built_by_suit member to instance, and modified main
to support it.

* Added the fcs_is_parent_card macro in freecell.c and changed all the
parent-child checking to it.

* Added an if->return statement to handle the sequences built by suit
stuff in moving cards to the foundations.

From 0.9.2 to 0.9.3
-------------------

* Changed check_state_validity to reflect the multiple decks.

* Changed the bitwise or in move_top_stack_cards_to_founds and 
move_freecell_cards_to_found to bitwise xor. It was a bug!, and exists
in version 0.2, 0.4, 0.6, and 0.8 too.

* Fixed a bug in the main function in inputting the number of stacks. It
also exists in FCS 0.8.

* Modified the freecell.c functions to make use of the multiple decks.

From 0.9.1 to 0.9.2
-------------------

* Added decks_num to freecell_solver_instance

* Added decks_num in the main function and in the debug_context.

* Added decks_num support to the parseable part of state_as_string.

* Added decks_num support to the non-p part of state_as_string.

From 0.8.0 to 0.9.1
-------------------

* Added the MAX_NUM_DECKS macro and modified state.h accordingly.

From 0.7.21 to 0.8.0
--------------------

* Added the section about changing stacks, freecells and cards per stack
maximal numbers in INSTALL.

* Changed MAX_NUM_INITIAL_CARDS_IN_A_STACK back to 7.

* Changed the version number to 0.8

* Changed the makefile to be a release one.

From 0.7.20 to 0.7.21
---------------------

* I realized that storing the solution states is done centrally in 
solve_for_state. Thus, I removed it from the macros, and from one of the
tests functions in which it was present.

* I changed the constants 0 and 1 to FCS_STATE_WAS_SOLVED and 
FCS_STATE_ORIGINAL_STATE_IS_NOT_SOLVEABLE.

* The makefile distributed with 0.7.20 had been set to HASH_STATE_STORAGE.
Ouch!
I modified it to be INDIRECT_STATE_STORAGE.

* changed the constants 0 and 1 in solve_for_state to FCS_STATE_WAS_SOLVED
and FCS_STATE_IS_NOT_SOLVEABLE.

From 0.7.19 to 0.7.20
---------------------

* Finished the file USAGE and fixed README. 

From 0.7.18 to 0.7.19
---------------------

* Added support for glib's balanced binary trees.

From 0.7.17 to 0.7.18
---------------------

* Started to write the USAGE file

* Added the hash-table state storage (which is as slow as hell)

From 0.7.16 to 0.7.17
---------------------

* Added an INSTALL file.

* Modified the README file.

* Added a README in the board_gen sub-dir.

From 0.7.15 to 0.7.16
---------------------

* Added the ability to consider "T" and "0" as card 10.

* Added the same for the output.

* Modified the help text to include the new switches.

* Changed the shorts in card.c to ints.

* Wrote a new and improved makefile to handle all those state storage 
  mechanisms. The old makefile is preserved in the file Makefile.lite

From 0.7.14 to 0.7.15
---------------------

* Modified the makefile and headers so libredblack will be compiled with
the -lredblack flag.

* Removed redblack.c and redblack.h from the archive. (I don't want anything
in here to divert from the public domain).

From 0.7.13 to 0.7.14
---------------------

* Implemented the variable number of cards in a stack.

From 0.7.12 to 0.7.13
---------------------

* Added freecells_num and stacks_num to call to canonize_state() 
in DIRECT_STATES

* Removed the call to free inside solve_for_state().

* Various fixes so fcs will compile smoothly with the trees.

* Timed the trees vs. indirect on 1000 sample boards. The trees were 
approximately 33% faster.

From 0.7.11 to 0.7.12
---------------------

* Added freecells_num and stacks_num to call to canonize_state() 
in INDIRECT_STATES

* Bug fixes related to variable number of stacks and freecells.

* Variable stacks and freecells num now works.

* TODO: Variable number of cards in stacks.

From 0.7.10 to 0.7.11
---------------------

* Changed 8 to instance->stacks_num. It doesn't work yet.

* TODO: Fix it.

From 0.7.9 to 0.7.10
--------------------

* Modified user_state_to_c.

* Fixed an ouput bug in state_as_string().

From 0.7.8 to 0.7.9
-------------------

* Fixed the state_as_string -p option to properly handle a variable
number of freecells. Next we have to check if it can handle a variable
amount of stacks.

From 0.7.7 to 0.7.8
-------------------

* Some more variable nubmer of freecells and stacks modifications.

* Tested that it works with less than 4 freecells.

* TODO: Write a proper not -p output state_as_string function.

* TODO: Implement variable number of stacks.

* TODO: Fix user_state_to_c

From 0.7.6 to 0.7.7
-------------------

* Started implementing the variable nubmer of freecells and stacks.
Still a lot to do though, especially in the algorithm.

From 0.7.5 to 0.7.6
-------------------

* Using avl instead of libredblack.

TODO:

 *TREE_IMPLEMENTATION macros in config.h
benchmarking the results.
Makefile ifdefs and stuff for handling the avl library.

From 0.7.4 to 0.7.5
-------------------

* Cleanups and organization of the macros.

* Defined the macro LIBREDBLACK_TREE_IMPLEMENTATION and there will also be
AVL_AVL_TREE_IMPLEMENTATION AND AVL_REDBLACK_TREE_IMPLEMENTATION.

From 0.7.3 to 0.7.4
-------------------

* Implemented REDBLACK_TREE_STATE_STORAGE (in a macro-wise ugly way).

* Modified the makefile to compiled redblack.c too.

* Changed if (!solve_for_state(...)) to 
if (solve_for_state(....) == FCS_STATE_WAS_SOLVED) in the 
check_and_add_state functions.

From 0.7.2 to 0.7.3
-------------------

* Fixed a bug in the board generation of
(pi_)?make_microsoft_freecell_board.c . The pseudocode given as reference 
had a mistake!

From 0.7.1 to 0.7.2
-------------------

* Pre-calculated the maximal sequence length in 
sfs_move_sequences_to_free_stacks().


From 0.6.2 to 0.7.1
-------------------

* Changed the calc_max_sequence_move to the (fc+1)<<fs calculation. It
enabled me to implement it as a macro.


* Moved the board generation programs to the board_gen sub-dir of the 
distribution.

* Added make_microsoft_freecell_board.c, 
pi_make_microsoft_freecell_board.c (buggy), 
make_pysol_freecell_board.c 
(should not be used because of 64-bit problems) 
and make_pysol_freecell_board.py (O.K.)


From 0.6.1 to 0.6.2
-------------------

* In make_gnome_freecell_board.c removed the declaration of rectify,
and added a call to free at the end of the main() function.

From 0.6 to 0.6.1
-----------------

* Changed num_freestacks to num_freestacks-1 on the following line in
freecell.c:

/* One stack is the destination stack, so we have one     *
 * less stack in that case                                */
while ((calc_max_sequence_move(num_freecells, num_freestacks-1) < cards_num -c) && (c > 0))
{ 
    c--;
}

It was a bug, and it probably exists in freecell solver version 0.4.x and
0.2.x too.


From 0.5.30 to 0.6
------------------

* Checked that fc-solve solves my boards and released a stable release.

From 0.5.29 to 0.5.30
---------------------

* Added an opening comment to main.c and fcs.h, and modified freecell.c's.

* Changed Makefile.release.


From 0.5.27 to 0.5.29
---------------------

* Nothing much. Replaced tabs by 4 whitespaces, cleaned the code a little.

From 0.5.26 to 0.5.27
---------------------

* Placed the main function and other non-library functions inside main.c.
  freecell.c remains with the actual solving algorithm.

From 0.5.24 to 0.5.26
---------------------

* Fixed a bug that instance was freed before num_times is printed.

* Created the following functions:
      freecell_solver_init_instance
      freecell_solver_solve_instance
      freecell_solver_finish_instance

  and replaced their code in main with calls to them.

From 0.5.23 to 0.5.24
---------------------

* Added a help string for tests-order.

From 0.5.22 to 0.5.23
---------------------

* Added the ability to specify the order of the tests. No help string on it
though.

From 0.5.21 to 0.5.22
---------------------

* Splitted the move stack cards to stacks function into two:
  1. Move stack cards to diffrent stacks.
  2. Move single stack cards to a non-immediate parent on the same stack.

From 0.5.19 to 0.5.21
---------------------

* Placed each of the various solve_for_state tests in its own function, 
made an array of those tests and solve_for_state is a loop over them.

From 0.5.18 to 0.5.19
---------------------

* Placed the "remove cards from stack in order to put a card in the decks"
in the beginning of solve_for_state.

From 0.5.17 to 0.5.18
---------------------

* Modified the help string to mention max-depth and max-iter-num.

From 0.5.16 to 0.5.17
---------------------

* Added an optional user-specified limit to a maximal number of iterations 
or depth.

From 0.5.15 to 0.5.16
---------------------

* Eliminated putting the initial board in prev_states[0] in
INDIRECT_STATE_STORAGE.

* Handled instance variables that are specific to DIRECT_STATE_STORAGE or
INDIRECT_STATE_STORAGE.

* Placed the state output on each iteration inside a callback function.

From 0.5.14 to 0.5.15
---------------------

* Placed the command-line options inside the instance struct.

From 0.5.13 to 0.5.14
---------------------

* Added fcs_ prefix to macros.

* Added fcs_ prefix to functions.


From 0.5.10 to 0.5.13
---------------------

* Compiled and ran with DIRECT_STATE_STORAGE and fixed some bugs.

* Added the option to print in canonic or location-preservative order.

From 0.5.6 to 0.5.10
--------------------

* Implemented the remember the original location of every stack and freecell,
so the user will not see them re-arranged all the time.

From 0.5.5 to 0.5.6
-------------------

* More global variables transfered to freecell_solver_instance_t

* Solution boards now stored as state_t's.

From 0.4.1 to 0.5.5
-------------------

* Added the portion about moving a card to a parent card when they
are both in the same stack.

* Created a freecell_solver_instance_t struct to hold all the global
variables.

From 0.4 to 0.4.1
-----------------

* Minor tweaks to get all the files compile on Win32. 


From version 0.2 to version 0.4
-------------------------------

* Replaced qsorting the sort margin with a call to 
SFO_merge_large_and_small_sorted_arrays. Now, the sort margin is kept 
outside the main array of previous states, and it is kept sorted.

* Added a check_and_add_state function that checks the state, and if it 
is not present in the previous states records it adds it there and then
call solve_for_state on it.

* Implemented Compact States, which store the states in much less memory.
(and are faster at least on Pentiums). Program can be compiled with either
compact states or debug states (the old int and short based states).

* Implemented Indirect State storage, in which the pointers to the states
are sorted rather than the whole states data-structures. Added the
state_isa.c module to allocate memory for the states in a 
memory-conservative fashion. Program can be compiled with either indirect
states or direct state storage.

* Fixed some bugs in the user_state_to_c function.

* Added support for specifing the contents of the free cells and the
foundations, and added support for stacks of variable length.

* The check_state_validity function to make sure all cards are present
in the initial board.
