BizHawk/mednafen/TODO

!!! Make sure all legal notices and licenses for DLLs are included in the Win32 binary distribution !!!

Document the major changes in Mednafen 0.9.0 over Mednafen 0.8.x somewhere(NEWS?).  A list to get started:
	Sega Genesis/Mega Drive emulation, from Genesis Plus, Genesis Plus GX(EEPROM emulator, general information), Yabause(68K emulator), Gens(YM2612 emulator), and
	of course the Z80 emulator core from FUSE already used in SMS and NGP emulation.

	Virtual Boy emulation(using the V810 emulator forked(and optimized and otherwise improved!) from Reality Boy ages ago for use in PC-FX emulation, along
	with original code for everything else).

	Multi-threaded CD-ROM reading code.

	Significantly more accurate(though not entirely perfect, and slower) PCE emulation.
	
	The "pce_fast" emulation module, which is a fork of 0.8.x optimized for speed, at the expense of *generally*-unimportant accuracy.

	Various accuracy improvements to PC-FX emulation, including ADPCM IRQ generation(fixes Sparkling Feather),
	correcting the ADPCM decoding algorithm, a new slightly more accurate V810 emulation mode to fix glitches in
	Chip Chan Kick and Kishin Douji Zenki FX, greatly improved accuracy of KING->RAINBOW block transfers, and an 
	interrupt controller fix that allows for the BIOS PhotoCD player to work.  The 2 switches on gamepads are also emulated
	now.

	A new experimental/unstable video driver, "overlay".

	"Slow forwarding" feature.  Mednafen is too good for "slow-mo". ;b

	Support for WSR(WonderSwan sound rip format) playback.

	Support for RTC in GBA games(requires a "type" file).

!!! Initialize statically-allocated Blip_Buffer objects(via clear() method) and reset any corresponding last-value variables
on game load for each emulated system.

Rewrite MDFNFILE in file.cpp to separate classes, ArchiveReader(takes a directory or archive filename) and ArchiveMemberReader(reads a file in a directory or a file
in an archive).

Fix C68K generator to work correctly when cross-compiling to a platform whose binaries can't run(through automatic emulation)
on the build platform.

In drivers/video.cpp, see if replacing "SDL_FillRect(screen, NULL, 0);" in ClearBackBuffer() with custom software-only
code to clear it fixes framerate degradation problems on Windows when env var SDL_VIDEODRIVER=directx and setting vdriver=sdl
Additionally, optimize ClearBackBuffer() to not clear parts of the screen that will be overwritten by normal non-OSD
blitting, and do testing for the vdriver=overlay case.

In drivers/video.cpp, allocate video scaler pixel buffers separately from the SDL_Surface itself, and realloc() to larger
sizes when necessary(but never realloc() to smaller sizes), and free() only on video deinitialization.

Add a setting to disallow joystick(or keyboard even) input when focus is lost.

Switch hotkey/command key processing over to be event-based rather than instantaneous-state based.  Consider implications for ckdelay, hotkey
logical AND mode.

Systems with RTC emulation: Fix code to NOT use localtime(), which has consistency issues due to relying on external state(IE time zones), at
			    any time other than virtual power toggle/init.

YM2612:
	Timer fixes?

Internationalization issues:
	Radix(decimal point) character:
		Remove world_strtod().
		Use trio functions solely for string<->floating-point conversion(in settings code), and ensure that trio is using the current locale.

	Alter similar translatable strings in various emulation modules to be the same strings.
	
	Ensure all important translatable strings are marked as such.

	Update Russian and German translations.

	Filesystem character coding vs UTF8 internal encoding vs locale.

	Replace string/ConvertUTF.cpp with iconv() usage.

	Wrap iconv() in C++ classes.

Documentation:
	Document systems with a debugger available.

Netplay:
	Simplify name resolution and eventually remove hackish-sort-preference for IPv4 addresses over IPv6(in 2014 maybe?), split functionality
	into different components:
		NetServer (maybe, to support *potentially* lower-latency P2P-style netplay)
		NetClient (keep, and rework, select() type functionality in this class)
		|-NetAddress
		|-NetConnection

	Consider handling a recv() return value of 0 differently.


Settings:
	Quote(and escape) string-type settings in the configuration file.

	Protect settings structures and objects with mutexes(for thread safety when using MDFNI_SetSetting() while the emulator is running)

	Optional manually-created per-game setting override files(or one big database file per system?).

	Save some emulation-related settings in save states, and temporarily(until another save state load, or a virtual power toggle via F11) make
	them active.  Consider implications for netplay, state rewinding, and PSX memory card emulation(slots can be disabled via settings).
	Would require revamping large amounts of code.
	

Video:
	Refactor hq2x, scale2x, 2xsai, etc. filtering code into classes based off an abstract base class.

	Add a "tblur.handle_gamma" setting that, when enabled, will cause gamma-correct temporal blurring.

Sound:
	Handle sound being disabled.

	Open sound device in the game thread, and close it in the game thread, since the game thread is what writes to it.

Backup memory:
	SRAM delta log and/or backing up of previous save game files.

PCE_FAST:
	MWR latching!


PCE:
	<AamirM> <Ki_> one thing about timer inside the HuC6280
	<AamirM> <Ki_> the 7-bit timer is decremented when MSB of the prescaler changes from 1 to 0.
	<AamirM> <Ki_> (assuming the prescaler is an up counter)
	<AamirM> <Ki_> the prescaler seems to be a 10-bit counter
	<AamirM> <Ki_> which is reset to zero when zero is written to the control bit ($0c01.bit0)
	<AamirM> <Ki_> if the control bit is written to zero when the prescaler is greater than 512 (i.e. the MSB is 1),
	<AamirM> <Ki_> prescaler is reset and the MSB changes from 1 --> 0
	<AamirM> <AamirM> ah..so the timer decrements
	<AamirM> <Ki_> exactly
	<AamirM> <Ki_> and if the timer value was zero at that point, /TIRQ will be asserted

	Fix HuC6280 op TMA with multiple bits set.

	Determine startup state of various chips(including PSG).

	Fix sprite 0 IRQ timing.

	Ensure all state is being saved to save states.

	Fix VRAM->VRAM DMA, SAT DMA timing, and CPU-halting handling with new information from tests on a real system.

	Fix broken SuperGrafx window emulation.

	If power/reset is triggered in the debugger, force a return from MDFNI_Emulate(); either that or clear
	the audio buffer on power/reset.  To fix a potential sound buffer overflow issue.  (Actually, that won't happen
	as-is, but

	Emulate disabled button/output when CLR is active on gamepad reads(might need to adjust the multitap code
	to buffer data in this case?).

	Double-check timer IRQ semantics vs real system(what reads and writes acknowledge IRQs, cli-then-ack behavior, 
	mask register write effective delay, etc.).

GBA:
	GSF multiboot support, and make sure it works(and GSF loading in general) on big-endian platforms.

SMS:
	Rewrite VDP execution loop so events happen on the correct cycles more or less, and to remove the ugly IRQ-versus-
	IRQ-bit-status-poll hacks.  Use "z80_tstates" to do this.

	Simplify and improve(including TH latching of VDP H-counter) IO emulation, and use Genesis controller/gamepad
	code(abstract it out of the Genesis code, into hw_input/sega_db9/* or something similar)

	Galactic Protector input when FM sound is enabled.

	Actually save/restore YM2413 data in save states.

NGP:
	Implement low-level BIOS and auxillary hardware emulation, with information provided by Judge.

VB:
	Check if Red Alarm's sound effects are supposed to be totally wonky when various sound effects "overlap".

	Look into getting VIP SCANRDY status bit emulated correctly.

Genesis:
	Fix VDP H-counter read value(write test program if necessary).

	Add Z80 IRQ ack hack which can be controlled by a setting.

	Event system in C68K, fix possible problems with excessively long 68K instructions and the Z80 IRQ
	not being handled.

	Mouse emulation is borked, fix it.

	Implement "Virtua Racing" SVP emulation via code supplied by notaz.

PC-FX:
	cdrom/scsicd.cpp
		Allow commands that require medium access to complete successfully if there's no disc present and 
		the "alloc size" field is 0.

		Generate SCSI errors when reserved fields are set to something other than 0(especially cdb[1] and the final cdb
		entry).

		Complete mode page emulation: especially the CD audio control page, tests need to be done on the real
		system to see if the volume fields have any effect, and if it's possible to have both L and R channels active
		(0x3) on a single output port, or if only one is allowed per output port.

		Determine what data command 0xDE, mode 0x03, match 0xB0 is returning(the PhotoCD player uses it, it may
		contain information encoding the LBA of the filesystem).
		Doing this will require obtaining various CD-i, CD-i Ready, and PhotoCD discs for examination.

		Command 0x42(READ SUB-CHANNEL):
			Add proper MCN and ISRC support.

			Figure out the weirdness when the ISRC is requested for a data track.

			Implement Q subchannel interpolation(perhaps globally).
	
			Only return Q subchannel data for audio operations?

			Emulate audio completion(or lack thereof) status correctly(if completed successfully, 
			0x13, and from then on 0x15, I believe).

			Investigate audio status code when scanning.

	mempeekhw

	movbsu, cache load/store breakpoints

	Figure out 1M/4M DRAM mode differences on KING [low priority]

	Set pcfx.rainbow.chromaip default to 0 if the glitch in Deep Blue Fleet when viewing a ship from the side doesn't occur on a real PC-FX.

	Determine correct relative volume levels between ADPCM, CD-DA, and PSG.

	Investigate reported discrepancy between real PC-FX and Mednafen when uploading waveform to PSG by rewriting
	the channel number/index before each sample write(on a real PC-FX, it doesn't work; IE, likely either the sample
	writes are being ignored, or the waveform memory index is being reset on channel number/index write).

	V810 MUL or MULU overflow flag logic is likely wrong(can't remember which off hand ATM; confirm, and fix).

NES:
	Add some kind of protection against FDS corruption when exiting the emulator in the middle of a disk write.

	Fix apparent sound skew(lost cycles?) messing up blargg's DMC tests.

	Fix weird graphical glitches in Quattro Sports: BMX Simulator

	nes/boards/vrc7.cpp (slot save state saving)

Safety fopen() replacement for preventing overwriting of files.

smem_seek() offset signedness fix

Reexamine the code for choosing which emulation module to use.  Highest priority to file extension(as long as the extension is unique to a system,
"bin" is no unique for example), and special magic test functions if that fails.

Investigate reports of undefined symbols(in V810 FPU emulation code, and the WonderSwan x86 disassembler, and possibly other places).  There ARE references to
undefined symbols, but the symbols that reference them are never referenced themselves.  We should probably fix this.

Add dummy copy constructors to lots of our new classes.

TODO:  Add a "SaveStateMD5" member to the MDFNGI struct, to allow the sharing of battery-backed save files between PCE and PCE_FAST
without allowing save states to be shared.

TODO: Clean up multithreaded CD-ROM reading code.  Add fixes for drives that don't support raw subchannel reading and/or raw reading.  Add fixes
for drives that return borked(from the wrong sectors) subchannel data.  Add fixes for drives that refuse to read garbage mode 1 sectors(or fix readahead to not read
the pregap between audio and data tracks).

TODO: Do locking in drivers/remote.cpp

TODO: Consider locking in MDFNI_printf() due to the indentation global variable.

Alter MD5 hash of games based on the module's name for save states etc.?

Emulation "module" namespace encapsulation.

Save state timestamp-related variables consistency checks to avoid the possibility of buffer overruns and, less importantly, locking up the emulator
by use of malicious save states.

Investigate excessively long pause time after buffer underrun with ALSA and Intel HDA.

Rewrite the cheat interface, and add support for multiple patches per named cheat.

Movie recording non-overwriting.

Clean up intialization/deinitialization code so that more than one game can be loaded per Mednafen run.  Including cleaning up the
 sound file player interface to not trash a system's MDFNGI struct. 
 (This is a lot harder than it sounds! :b)

Debugger(general):
	Disassembly address forcible resynchronization has inconsistent results across different emulated systems; make
	them consistent.

	assert() in cdrom/scsicd.cpp on line 2629 is being triggered; "loading a save state while
	stopped in the debugger brings up the message"

	Breakpoint logger mode(instead of breaking, output the instruction address that caused the breakpoint,
	and which breakpoint was triggered(assuming that address+breakpoint pair wasn't outputted to the text file already).

	Add opcode breakpoints to PC-FX and NES debuggers.

	Debugger memory usage map and per-address-space breakpoints.

	Support for multi-line entries in the log viewer.

	fix various debugger problems(first, register updates when in stepping mode, or when new values are inputted in the register editor)

	Up-to-date register updates(for things like timer counter registers) during debugger while in step mode.

	Sprite viewer

	BG map viewer

	Full undefined behavior logging

Input macro recording.

Return -1 on game load error in main(), -2(maybe!) for unknown format.



************
**Old TODO**:
************

The following games are broken to some extent:

Crystalis:		Mostly working, but the screen jumps around during
			dialogue.  It apparently resets the MMC3 IRQ counter
			mid-scanline.  It'll require low-level PPU and MMC3
			IRQ counter emulation to function properly.

Kickmaster:		IRQ occurs a few CPU cycles too late, interferes with NMI routine,
			and causes the game to lock up after the second boss.  Luckily, there
			are passwords, at least.
			*"FIXED" BY HACK*

Star Wars(PAL Beam Software version):
			MMC3 IRQ occurs when it shouldn't, similar to the problem in
			Kickmaster.
			*"FIXED" BY HACK*

*** General Features:

 Fix possible UNIF crashes(if no PRGx or CHRx chunks exist, it may crash,
 due to changes made in 0.92).

*** Emulation:

 Figure out what mapper 113 really is.

 Sound frame count stuff on PAL games(is it correct?).
 
 Fix Zapper emulation(one version of Chiller still doesn't always work correctly).