Tasks for Heroes.  If you think something is missing here,
please send a note to <heroes-discuss@lists.sourceforge.net>.

* Use the level-loading library in the level editor, so as to get rid
  of most struct in src/struct.h (and their ugly ATTRIBUTE_PACKED).
  This require makeing the level editor work with libhlvl (see
  src/lvl/README for more details).

* Add `FIXME: obscure' comments at will, so I know where comments or
  rewrite is really important.  (Actually this can be done by anyone
  reading the code.)

* Check dependencies.  Heroes >= 0.11 requires heroes-data >= 1.1,
  Heroes >= 0.18 requires heroes-data >= 1.4.  Heroes will abort with a
  "cannot open file X" message on error.  These is definitely some room
  for improvement here.

* The code is VERY UGLY and needs to be cleaned up.

* hedlite.c should be cleaned.

* If possible, Get rid of GETWORD, SETWORD and Cie, see src/endian.h.

* The default installation path on BeOS should not be /usr/local (?).  I need
  some more input about this.

* There are several comments in French.  They should be translated.

* Check every place were filenames are used, use buffer allocated dynamically.

* Split heroes.c

* Split hedlite.c

* const.[ch] should not exist.

* Use {u,s}{8,16,32}_t types where it's important.

* __attribute__ ((packed)) is a gcc-isms, something must be done to avoid
  them. `something' may be using a different level format, in which case
  the `heroeslvl' above would be helpful.

* Add an option that add a new life each time you start a new level.
  Suggested by Samuel Tardieu (sam@infres.enst.fr).

* Programmable vehicles (for CPU controlled ones) would be nice.  It
  would be fun to link with Guile and control the vehicles in scheme.
  Or in Python.  Or in C (using shared libraries like pluggins).

* The external editor could be ported too, this
  would need many many more work.  (It's probably easier to rewrite it
  from scratch or from hedlite.c.)

* Network support.

* Direct X11 support would be nice.

* What is the effect of removing the DMODE_SOFT_MUSIC from MikMod options ?

* Would Hermes be faster than LibGGI's crossblit ?

* Redesign the way menus are handled internally.

* Write assembly routines for stretching code.

* Use makeheaders ? (provided I can get in touch with its maintainer)

* Load GGI, SDL, and MikMod libraries dymically (i.e. select
  available library at run-time versus configure-time).  This is probably
  easier to do now, with the new way media libraries are handled.

* Proof-read the documentation.

* Full screen support with Allegro.  Is it possible?

* It would be nice to have the argument processing and argument
  documentation of `heroes' and `heroeslvl' generated automatically
  from a definition file (well, two definition files, one for each
  program) using AutoGen.  AutoGen ships with a tools called AutoOpts
  that performs a similar task, but I find the result horrible.  Our
  command line parsing code should follow the Gnits standards and use
  getopt_long.  (Such alternate tool is on the TODO list of AutoGen,
  so it might be contribued back.)

* The end-scroller used to be an end-scroller :) Now it's just a
  screen that display "THE END".  That's not really rewarding for the
  player.  It would be nice to have something more delighting here.

* Heroes-data should ships with 100 levels, shouldn't it? (There is
  only 94 levels, IIRC.)

* There are some "extra" levels in heroes/src/extralvl/ on CVS.  Do
  they deserve some packaging?

* There are some hq modules packaged in http://heroes.sf.net/dl/misc/
  Should they be packaged in a more "official" way?

* Some files #include too much headers (for historical reasons), this
  slows down compilation times.  That's quite unrelated to Heroes, but
  a tool that can figure which #include are superfluous would be
  appreciated by several people I have talked to.

* The "AI", if I may call this "AI" (see `ia_*' functions in heroes.c, but
  have some aspirin handy) is ABSOLUTELY HORRIBLE.  It's using a very
  stupid and inefficient recursive algorithm.  `A*' is not usable as-is
  with the current path evaluation functions, but maybe these heuristics
  can be changed too.  A better "AI" have been required several times in
  the comments about Heroes on www.happypenguin.org.

* Heroes is relocatable (the manual explains that), but heroeslvl isn't.

* The paragraph formater (parafmt.c), has two minor issues: it does
  not handle a paragraph with leading spaces (i.e. first line is
  indented) properly, and will probably fail to format a paragraph where
  one word is larger than one of the line (even if that's not the line
  the word is finally put on).

* Improve the test suite (whatever you think it can mean).

* Add more items to this TODO list.