PyBoy•5mo ago

Save states not functioning properly when sound flag is used

Hello, I've been testing out Pyboy with Pokemon Crystal, and was using the --sound flag to enable the sound. I used Z to store the save state which it did. But when I came back around and tried reloading it with the --sound flag. the sound was gone and an error was present

14 Replies

Sas2kOP•5mo ago

OSError: exception: access violation reading 0x0000016DFCE50000
Exception ignored in: 'pyboy.core.sound.Sound.enqueue_sound'
Traceback (most recent call last):
  File "C:\Users\+++++\AppData\Local\Programs\Python\Python312\Lib\site-packages\pyboy\__main__.py", line 172, in main
    while pyboy.tick():
          ^^^^^^^^^^^^
OSError: exception: access violation reading 0x0000016DFCE50000

OSError: exception: access violation reading 0x0000016DFCE50000
Exception ignored in: 'pyboy.core.sound.Sound.enqueue_sound'
Traceback (most recent call last):
  File "C:\Users\+++++\AppData\Local\Programs\Python\Python312\Lib\site-packages\pyboy\__main__.py", line 172, in main
    while pyboy.tick():
          ^^^^^^^^^^^^
OSError: exception: access violation reading 0x0000016DFCE50000

the library seems to be trying reference an invalid pointer/file

Bækalfen•5mo ago

I've tried a few things now, and I cannot seem to replicate it on Mac and Linux. I can see you might be on Windows? But that should of course also work And it works fine if you don't load a saved state? Could you run PyBoy with --log-level=DEBUG and paste the whole output from the terminal?

Sas2kOP•5mo ago

yeah, it works when I run an entirely new instance

pyboy.core.cartridge.cartridge DEBUG    Loading ROM file: 2097152 bytes
pyboy.core.cartridge.cartridge DEBUG    Cartridge type: 0x10 - MBC3, SRAM, Battery, RTC
pyboy.core.cartridge.cartridge DEBUG    Cartridge size: 128 ROM banks of 16KB, 4 RAM banks of 8KB
pyboy.core.cartridge.base_mbc  DEBUG    RAM loaded.
pyboy.core.mb                  DEBUG    Cartridge started:
MBC class: MBC3
Filename: C:\Users\++\Downloads\Pokemon - Crystal Version (USA).gbc.ram
Game name: PM_CRYSTAL
GB Color: False
Cartridge type: 0x10
Number of ROM banks: 128
Active ROM bank: 1
Number of RAM banks: 16
Active RAM bank: 0
Battery: True
RTC: True
pyboy.core.mb                  DEBUG    Cartridge type auto-detected to CGB
pyboy.core.lcd                 DEBUG    Starting CGB renderer
pyboy.plugins.base_plugin      DEBUG    WindowSDL2 initialization
pyboy.plugins.base_plugin      DEBUG    Scale: x3 (480, 432)
pyboy.core.mb                  DEBUG    Loading state...
pyboy.core.mb                  DEBUG    State version: 12
pyboy.core.cpu                 DEBUG    State loaded:
A: 01, F: 20, B: D6, C: 01, D: 44, E: 5F, HL: D150, SP: C0E3, PC: 045F ()
Opcode: [HALT]  Interrupts - IME: True, IE: 00001111, IF: 00000001
LCD Intr.: 0, LY:0, LYC:0
Timer Intr.: 0
halted:True, interrupt_queued:False, stopped:False

pyboy.core.cartridge.base_mbc  DEBUG    RAM loaded.
pyboy.core.mb                  DEBUG    State loaded.
OSError: exception: access violation reading 0x0000016AD47B9000
Exception ignored in: 'pyboy.core.sound.Sound.enqueue_sound'
Traceback (most recent call last):
  File "C:\Users\++\AppData\Local\Programs\Python\Python312\Lib\site-packages\pyboy\__main__.py", line 172, in main
    while pyboy.tick():
          ^^^^^^^^^^^^
OSError: exception: access violation reading 0x0000016AD47B900

pyboy.core.cartridge.cartridge DEBUG    Loading ROM file: 2097152 bytes
pyboy.core.cartridge.cartridge DEBUG    Cartridge type: 0x10 - MBC3, SRAM, Battery, RTC
pyboy.core.cartridge.cartridge DEBUG    Cartridge size: 128 ROM banks of 16KB, 4 RAM banks of 8KB
pyboy.core.cartridge.base_mbc  DEBUG    RAM loaded.
pyboy.core.mb                  DEBUG    Cartridge started:
MBC class: MBC3
Filename: C:\Users\++\Downloads\Pokemon - Crystal Version (USA).gbc.ram
Game name: PM_CRYSTAL
GB Color: False
Cartridge type: 0x10
Number of ROM banks: 128
Active ROM bank: 1
Number of RAM banks: 16
Active RAM bank: 0
Battery: True
RTC: True
pyboy.core.mb                  DEBUG    Cartridge type auto-detected to CGB
pyboy.core.lcd                 DEBUG    Starting CGB renderer
pyboy.plugins.base_plugin      DEBUG    WindowSDL2 initialization
pyboy.plugins.base_plugin      DEBUG    Scale: x3 (480, 432)
pyboy.core.mb                  DEBUG    Loading state...
pyboy.core.mb                  DEBUG    State version: 12
pyboy.core.cpu                 DEBUG    State loaded:
A: 01, F: 20, B: D6, C: 01, D: 44, E: 5F, HL: D150, SP: C0E3, PC: 045F ()
Opcode: [HALT]  Interrupts - IME: True, IE: 00001111, IF: 00000001
LCD Intr.: 0, LY:0, LYC:0
Timer Intr.: 0
halted:True, interrupt_queued:False, stopped:False

pyboy.core.cartridge.base_mbc  DEBUG    RAM loaded.
pyboy.core.mb                  DEBUG    State loaded.
OSError: exception: access violation reading 0x0000016AD47B9000
Exception ignored in: 'pyboy.core.sound.Sound.enqueue_sound'
Traceback (most recent call last):
  File "C:\Users\++\AppData\Local\Programs\Python\Python312\Lib\site-packages\pyboy\__main__.py", line 172, in main
    while pyboy.tick():
          ^^^^^^^^^^^^
OSError: exception: access violation reading 0x0000016AD47B900

Here's the debug statement

Bækalfen•5mo ago

Nothing out of the ordinary Could you try cloning the repo and inside the repo do python3 -m pyboy ... as you did before. It should give some more details

Sas2kOP•5mo ago

ah I'll try that, in a moment also, the sound compatibility is still in beta right?

Bækalfen•5mo ago

It's still in beta, but it shouldn't crash. Just not be perfectly accurate, and sometimes stutter a little

Sas2kOP•5mo ago

yeah

krs013•5mo ago

I might have an idea what's happening here On load state, the cpu restores its cycle counter to what it was before, causing the sound device to try to process more samples than it's supposed to.

    def tick(self, _cycles, double_speed):
        cycles = _cycles - self.last_cycles
        self.last_cycles = _cycles

        if double_speed:
            self.clock += cycles // 2
        else:
            self.clock += cycles

    def tick(self, _cycles, double_speed):
        cycles = _cycles - self.last_cycles
        self.last_cycles = _cycles

        if double_speed:
            self.clock += cycles // 2
        else:
            self.clock += cycles

self.last_cycles is initialized to 0 instead of the value of pyboy.mb.cpu.cycles (which probably isn't loaded when the Sound object is made anyway) to make matters worse, there's this:

        nsamples = self.clock // self.sampleclocks

        for i in range(min(2048, nsamples)):
            ...

        nsamples = self.clock // self.sampleclocks

        for i in range(min(2048, nsamples)):
            ...

I'm not sure why that's min instead of max for the number of samples to generate--I think we've gone back and forth on it before, but whatever the reason is, if nsamples is greater than 2048, it will overrun the sample buffer. Then the access violation is caused in Sound.enqueue_sound(self):

        sdl2.SDL_QueueAudio(self.device, self.audiobuffer_p, 2 * nsamples)

        sdl2.SDL_QueueAudio(self.device, self.audiobuffer_p, 2 * nsamples)

when SDL_QueueAudio(...) tries to access a values indexed off of self.audiobuffer_p that is outside of the buffer and possibly allocated memory. I don't know when I'll have time to revisit --sound 🥲 but for now it should be sufficient to change that min to max on line 171, and test to make sure that doesn't break anything.

Bækalfen•5mo ago

I think the min is right. It bounds it to be ≤2048

krs013•5mo ago

Oh right, right. I think I messed it up the first time too haha Been a long week Shoot, guess I don't have anything to offer here then

Bækalfen•5mo ago

I also get confused with min/max 😅

BET Adsorption•4mo ago

Best support thread ever

Bækalfen•4mo ago

I’ll be waiting for your PR I think I've inadvertently fixed this in the next version. I can only reproduce it on master

Sas2kOP•3mo ago

ah nice, good job!

Gaming

Programming

Save states not functioning properly when sound flag is used

Did you find this page helpful?