Super Metroid: Replace random module with world random in variaRandomizer

[Mysteryem]

What is this fixing or adding?

Replaces direct use of the random module with passing around the world's .random attribute instead.

Fixes different generation on webhost compared to local generation, assuming no yaml options are using random options because all random option resolution is currently nondeterministic on webhost.

The new arguments for passing self.random through to the various variaRandomizer classes and functions have been added to the end of the arguments for each relevant function, except utils.randGaussBounds where the random argument is more like a self argument in a method, so has been added as the first argument instead.

The multiworld seed is now passed through to variaRandomizer instead of it generating its own seed.

How was this tested?

I started by removing all the random module imports and ran many generations with 5 almost fully random yamls, making adjustments to pass through the world's .random whenever generation crashed because it was trying to access the random module still. This helped me see the function calls through the code to determine what classes and functions I would have to pass the world's .random through.

Once generation was succeeding and I could not find any more direct use of the random module, I ran a generation of the template yaml on archipelago.gg (specifically the template yaml has no options set to random so should generate deterministically) and generated with the same seed locally and determined them to now produce the same results.

Additionally, I ran these changes along with the WIP test in #4410, and the failing test for Super Metroid would now pass.

I have not done any tests that go as far as connecting the client to a multiworld and patching the ROM, so I recommend more testing should be done by people more familiar with Super Metroid AP.

[PoryGone]

@lordlou

[Exempt-Medic]

This has conflicts btw

[lordlou]

Looks good to me, thanks!

[beauxq]

I think passing self.multiworld.seed might not be what we want.
This is an issue I've run into in other games.

If there are 2 Super Metroid players in the same multiworld, with similar enough settings, they might find that they both have certain things randomized to the same result.
For example: 2 Super Metroid players both turn on area rando, and they both end up with the exact same area connections - which is not what they expect.

I haven't tested whether this is the case (if not with area rando, then it might be with something else).

It might be better to pass self.multiworld.seed + self.player, or maybe just something from self.random.randrange

[Mysteryem]

The only use of the seed that I could find is in APPostPatchRom to call romPatcher.writeSeed(seed). The values written are used for the seed display (picks four enemy names to display at the top of the screen) after the save file selection screen.

Set into patcherSettings["seed"]

Archipelago/worlds/sm/variaRandomizer/randomizer.py

Line 749 in 051518e

"seed": self.seed,

romPatcher is created and customPostPatchApply(romPatcher) is called, calling SMWorld.APPostPatchRom(romPatcher)

Archipelago/worlds/sm/variaRandomizer/randomizer.py

Lines 759 to 773 in 051518e

	if args.rom is not None:
	# patch local rom
	# romFileName = args.rom
	# shutil.copyfile(romFileName, outputFilename)
	romPatcher = RomPatcher(settings=patcherSettings, magic=args.raceMagic, player=self.player)
	else:
	romPatcher = RomPatcher(settings=patcherSettings, magic=args.raceMagic)
	if customPrePatchApply != None:
	customPrePatchApply(romPatcher)
	romPatcher.patchRom()
	if customPostPatchApply != None:
	customPostPatchApply(romPatcher)

romPatcher.writeSeed() is called with the seed as its argument

Archipelago/worlds/sm/__init__.py

Lines 774 to 775 in 051518e

	if not romPatcher.settings["isPlando"]:
	romPatcher.writeSeed(romPatcher.settings["seed"]) # lol if race mode

Writes seed values starting at 0xdfff00

Archipelago/worlds/sm/variaRandomizer/rom/rompatcher.py

Lines 498 to 503 in 051518e

	def writeSeed(self, seed):
	random.seed(seed)
	seedInfo = random.randint(0, 0xFFFF)
	seedInfo2 = random.randint(0, 0xFFFF)
	self.romFile.writeWord(seedInfo, snes_to_pc(0xdfff00))
	self.romFile.writeWord(seedInfo2)

https://github.com/theonlydude/RandomMetroidSolver/blob/master/patches/common/src/seed_display.asm uses that same 0xdfff00 address.

@lordlou would know better if the seed controls anything else, and whether it should be the same for multiple SM players in the same multiworld (so use the multiworld's seed) or if it should be different, but still deterministic for each individual SM player in the same multiworld (so use each world's .random to create a seed). The PR is currently using the multiworld's seed.