0.14

Program:
- Fixed bug in Mario Bros. high score loading: it was recovering the high
  score table, but the high score at the top of the screen was not set
  accordingly.

- Mario Bros. and Donkey Kong 3 now also preserve the score distributions
  (use F1 to see them)

- High score support in Galaxian, Time Pilot, Centipede and Millipede.
  However, high scores don't seem to work in Millipede (you are not asked
  to enter your name).

- Fixed bug which sometimes caused 6502 games not to run (actually they did
  run, but interrupts didn't happen).

- Since the US version is better, I renamed amidar to amidarjp, and
  amidarus to amidar.

- Renamed pleiades to pleiads.

- The change in video modes suggested by ue303ey@sunmail.lrz-muenchen.de
  didn't work on some systems, so I reverted to the previous one.

- Ron Fries provided a new version of his Pokey emulator which fixes problems
  in the interface with the SEAL audio library (signed vs. unsigned samples).

- Ron Fries also added preliminary sound support to Donkey Kong! It uses some
  very bad samples for now. The samples are distributed in a separate archive
  - put them in the dkong subdirectory.

- Richard Davies updated the audio routines in Phoenix, sound is now closer
  to the original and noise is emulated.

- Brad Oliver and Mirko Buffoni provided a driver for Vanguard, based on Brian
  Levine's Vanguard emulator.

Source:
- driver.c was getting much too large and confused. I moved the GameDriver
  structures inside the single driver/xxxx.c files. This has the additional
  benefit that people sending me drivers will not forget to include the ROM
  loading addresses ;-)

- I also changed the GameDriver structure, moving there fields from the
  MachineDrivers structure. The idea is that MachineDriver should describe
  the hardware, while GameDriver the software. Therefore things like colors
  (which are usually stored in a PROM) and dip switch settings go into
  GameDriver; decode_color_prom, on the other hand, goes into MachineDriver
  because it is an hardware function.
  It could be argued that input_ports belong to the hardware; but sometimes
  we have keys tied to dip switches (test switches and so on) whose function
  is determined by the software. Moreover input_ports contains the default
  values for all ports, including dip switches, and those are definitely
  software related.