ᗪMinimal Reproducible Code:
ᗪError Traceback:
ᗪI am half asleep, trying to debug this haha, mb if dumdum
ᗪSummary of PyBoy Troubleshooting Steps
My goal is to create an AI to play Pokémon Red using PyBoy. I've encountered a persistent crash when trying to send button inputs after loading a save state. Here's what I've tried so far:
- Initial Setup & State Reading:
- Installed Python 3.12.3 on Ubuntu (latest).
- Set up a virtual environment.
- Installed PyBoy (currently v2.5.1) and dependencies (, , ).
- Successfully loaded the ROM using PyBoy.
- Found RAM addresses for game state ( for Map ID, for Player Y, for Player X).
- Confirmed I could read these values correctly using in a loop.
- Save State Implementation:
- Implemented logic to create a save state () by playing manually to a desired point (just outside Oak's Lab, Map ID 0) and saving via when stopping the script (using caught by Python's module).
- Confirmed that the script successfully loads this state using on subsequent runs.
- Encountering the Crash:
- Added logic to send button inputs (e.g., , ) using after loading the state, followed by .
- Consistently encountered a crash immediately after the call, before the next could execute.
- The error reported by my block is .
- Enabled full traceback printing, which showed the error originates inside PyBoy's code: , line 337, in .
ᗪ- Debugging Attempts (PyBoy 2.5.1):
- Different Inputs: Confirmed the crash happens with various inputs (, ,
), not just one specific button.A - Graphics Mode: Confirmed the crash happens both with a visible window () and in headless mode (), ruling out SDL window interaction as the primary cause.
- Button Press Timing: Simplified the button press logic from a helper function with multiple ticks down to direct , , , calls. The crash still occurs immediately after the first .
- Tick Alone: Confirmed that calling repeatedly without any preceding call works fine after loading the state. The crash only happens when is involved.
- Initial Ticks: Added a loop to run 10 times immediately after but before the first in the main AI loop. This did not prevent the crash.
- PyBoy Debug Mode: Initialized PyBoy with . This provided verbose startup logs but no additional useful information right before the crash point.
- Dependencies: Confirmed and were already up-to-date using .
- Environment Change Attempts:
- Downgrade PyBoy: Attempted to downgrade to PyBoy 2.1.0.
- This resulted in a indicating binary incompatibility with the NumPy version being installed alongside it by pip (even with ).
- Attempting to manually install an older NumPy version (1.23) first failed because it required to build from source, which is missing in Python 3.12.
- Concluded that using PyBoy 2.1.0 on Python 3.12 is problematic.
- Reverted to PyBoy 2.5.1: Uninstalled older versions/NumPy and reinstalled latest and .
ᗪCurrent Status:
I'm back on PyBoy 2.5.1 / Python 3.12 / Ubuntu, and the crash reliably occurs inside PyBoy's code immediately following a call after loading a state. It appears to be an internal bug or incompatibility within PyBoy under these specific conditions that cannot be fixed by changes to my script's logic.
I'm back on PyBoy 2.5.1 / Python 3.12 / Ubuntu, and the crash reliably occurs inside PyBoy's code immediately following a call after loading a state. It appears to be an internal bug or incompatibility within PyBoy under these specific conditions that cannot be fixed by changes to my script's logic.
BI’m not reading all of that. Maybe keep it a little bit more concise next time.
But the issue is that you’re not supposed to send a string to send_input: https://docs.pyboy.dk/#pyboy.PyBoy.send_input
But the issue is that you’re not supposed to send a string to send_input: https://docs.pyboy.dk/#pyboy.PyBoy.send_input
BSo I can see you used an LLM to write all of this. Please don't use LLMs to submit bug reports to anyone. It's a waste of everyones time to read through something you didn't even take the time to write.