Bugs that existed in v1.07/v1.08 and needs to be fixed:

- [UNCONFIRMED] repeating sounds are kept alive when you "restart from
  scratch"

- Dink shouldn't take [space] into account for killing a text if the
  text just appeared on the screen (maybe except when the user is
  pressing space continously, showing interest to skip the whole
  conversation)

- When saying a text near the border of a screen, it isn't placed
  correctly the very first time it's displayed

- say_stop*(&sprite, ...) kills other text sprites related to
  &sprite. Unfortunately this prevents those other texts to return to
  their originating script, because of kill_text_owned_by(). This is
  discussed in http://www.dinknetwork.com/forum.cgi?MID=92406 so it's
  also present in v1.08.

- "Dinkedit has, and forever will, crash if you type in really big
  numbers. You should know better."
  (http://www.dinknetwork.com/forum.cgi?MID=57829)

- "Dinkedit may crash if opening certain D-Mods with a dink.ini from
  WDE 1.1-1.4. Try using an older version of Dinkedit, such as that
  included with the v1.07 patch, or use WinDinkedit."
  (http://www.dinknetwork.com/forum.cgi?MID=59442)

- The new 'custom' field in 'struct sp' is initialized with 'new'
  ('malloc' for FreeDink) but isn't freed. In addition it may not be
  initialized in all situations.

  In dinkvar.c, the following functions need to free it:
  - kill_text_owned_by
  - kill_sprite_all
  - update_play_changes
  - place_sprites_game
  - kill_last_sprite
  - place_sprites_game_background
  - fill_back_sprites
  and this function need to initialize it:
  - add_item (set to 1 but without filling in .custom)

  Fix it directly looks error-prone. Maybe we could implement some
  wrapper function to create and remove sprites, rather than
  manipulate one 'active' field directly. Then we might try and find a
  way to enforce this for the programmer.

- When using kill_this_magic(), *pcur_weapon is set to 0, instead of
  *pcur_magic.

For fixed bugs, see NEWS.
