Linux64/plux.so version 1.9 crashes on unstable connection with muscleBAN devices

[Inzarcon]

Issue and Steps to Reproduce

• Connect to a muscleBAN device using PLUX-API-Python3/Linux64/plux.so and example code of this repository.
• Once connection is established and stable, obstruct the Bluetooth connection (with your hands around the physical muscleBAN or by moving it out of range) for a short time Y depending on sample rate/frequency (see graph below).
• Immediately unblock the muscleBAN to reestablish Bluetooth connection.
• Depending on the length Y of the timeout, now one of three things will happen:
  • (OK) If the timeout Y was shorter than the time in the graph below, the streaming of samples continues without any issues or errors.
  • (OK) If the timeout Y was much larger, it will more likely raise a RuntimeError: The communication with the device was lost.
  • (Not OK) If the timeout Y was only slightly larger than in the graph, one of various C/C++ memory errors occur, and Python crashes completely:
    • munmap_chunk(): invalid pointer
    • malloc(): unaligned fastbin chunk detected
    • malloc(): unaligned tcache chunk detected
    • malloc_consolidate(): invalid chunk size
    • corrupted size vs. prev_size in fastbins
    • Quiet crash with no message at all

Further details

• Crash occurs in both Python 3.9 and 3.13.
• Crash occurs in both Ubuntu 20.04 LTS and CachyOS (Arch Linux-based).
• Crash occurs independent of which muscleBAN device is used.
• Behavior get worse with higher sample rates/frequencies set for the muscleBAN. I tested this with different sample rates:
  

Workaround

• Downgrading the plux.so file from version 1.9 to 1.7 found in the file history of this repository fixes the issue and consistently raises expected RuntimeError: The communication with the device was lost. instead of crashing.

[GRamosPlux]

Dear @Inzarcon,

Thank you very much for sharing this extremely detailed feedback and the respective workaround.
This information will certainly be helpful to the remaining members of our community 🎉

Regarding the fix of the memory issues in the Linux version of the API, sadly, due to the end of the official support to Linux systems, unfortunately, the team is not able to guarantee that this issue will be fixed in a near future, however, as previously mentioned, the workaround that you gently shared will be essential while a definitive solution is not achieved.

We would like to transmit our gratitude for being a proactive member of our community of users!

My best and sincere regards,
@GRamosPlux