* Use I8255 or MSM6253 devices for most non-JAMMA inputs
* Separate service and test inputs in Multi-32 and twin-unit sets
arescue: Four coins are recognized, not just two
ga2: Use MB8421 device for V25 communications
orunners: Assign keys to some 2P inputs
-----------------------------------
coco_cart: Added Super LOGO and removed all Dragon cartridges
dragon_cart: Moved all Dragon cartridges from coco_cart
dragon_cass: All known cassette dumps
dragon_flop: All known floppy dumps
dragon_flex: FLEX System and other Compusense FLEX releases
dragon_os9: OS-9, BASIC09, C, Dynacalc, Pascal, RMS, Stylograph
dgnalpha_flop: OS-9, NitrOS9, C, Dynacalc, Pascal, RMS, Stylograph, etc.
- Added joysticks to all machines
- Added floppy sounds
dragon200e: Added chargen ROM, not working (possibly mc6847 issue)
- Demoted to MACHINE_NOT_WORKING
d64plus: Added crtc with chargen ROM and second screen output, not yet hooked up
- Demoted to MACHINE_NOT_WORKING
dgnalpha: Added older Boot v0.4 ROM
- Replaced 5.25" floppy drives with 3.5", and added sounds
The NeoGeo driver exposes a number of different memory regions as vectors. Accessing the pointers was done through a '&vector[0]' pattern. This caused problems when the region was size zero; I changed the code to return null pointers in these scenarios.
This was reported by Robbbert in response to recent regressions (the command line was 'mame aes bjourney'), but it seems to be present in MAME 0.184.
* Made a number of device_image_interface members private
This also fixes a number of get_default_card_software() implementations that were accessing m_file incorrectly
* Remove duplicate prototype
* Revert entries(), indirect_entries(), shadows_enabled() and hilights_enabled() to return the configuration parameters instead of accessing the live state. The thought behind the implementation change was that palette devices could potentially determine the number of entries from the sizes of devfind objects. The regressions caused by this have been worked around, but it was probably a bad idea in the first place.
* Zero-entry palettes are no longer valid. The code that tried to support them was basically left over from when every running machine had a single global palette.
