+Known bugs:
+
+- some positions show problem in the evaluation algorithm
+- minishogi repetition rule, different from standard shogi, is
+ not implemented
+- with no piece on board, computer generates invalid move
+
Target for v1.5:
-- refactor dspwrappers into a driver-like structure
-- integrate H.G.Muller's work on the xboard protocol on top of that
-- deprecate xshogi in favor of xboard
+- use ~/.gnushogi for binary book and hash file
+
- make minishogi a run-time option rather than a compile-time one,
using the "variant" mechanism of the xboard protocol
- switch cli parsing to getopt
- hunt for extern's and prototypes spread all over the source
-- use automake
- hunt for more hardcoded variant-specific constants
- position of captured pieces in curses mode
- compile with bound-checker
- add autosave mode to ease hunt for segfaults ?
-- use 2D array for the board, get rid of the (i)nunmap stuff
+- ? use 2D array for the board, get rid of the (i)nunmap stuff
- fixup build procedure to support parallel make runs
- investigate those preexisting "overflow in implicit constant
conversion" warnings
- minishogi patterns
- tune difficulty levels
- other tunings to investigate
+ * eval.c has lots of heuristic code specific to standard shogi that
+ probably should not been used
* PromotionZoneDistanceBonus
* OPENING_HINT
- add minishogi stuff to the doc
- use valid minishogi moves in online help
+
+Improvements
+
+- find out why the engine is so weak
+ - make it battle against Mutz-era gnushogi, and other engines, and
+ compare their evaluation results
+ - see what needs updating in book
+- express available commands as display-dependant, instead of using if(XSHOGI)
+ - make "help" list commands available for current display
+- make X and raw two different display drivers
projects / gnushogi.git / blobdiff