TUL device dies after not being used for a long time

[emogenet]

Hi,

I own a busware.de NCN5130 + ESP32-based TUL device.

I have flashed it with the USB-serial emulation firmware to try and get it to behave like the old ATMEL-based TUL device.

After flashing from Chrome, the device says:

SP-ROM:esp32c3-api1-20210207
Build:Feb 7 2021
rst:0x15 (USB_UART_CHIP_RESET),boot:0x8 (SPI_FAST_FLASH_BOOT)
Saved PC:0x40384ee6
SPIWP:0xee
mode:DIO, clock div:1
load:0x3fcd5810,len:0xf4
load:0x403cc710,len:0x894
load:0x403ce710,len:0x2298
SHA-256 comparison failed:
Calculated: 5ffda6de84dc3dc4cb3a897082d65f27058273d706f9c8c1149e1fa69db0878f
Expected: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
Attempting to boot anyway...
entry 0x403cc710
TUL-KSGEY - init succeed - running: v1.4+42 - 2024-02-01 11:29:25 @ 160 MHz

(there is a checksum error, but I assume this is a harmless bug)

Right after flashing, the device works fine.
I use the device to monitor and log activity on the KNX bus, and I can read data fine.

However, if I leave the device plugged on USB on one side, and the KNX bus on the other, but I don't read from it for a long time ... the device stops working.

And, it stops working for good: usbreset does not fix it, power-cycling it does not fix it.
The only thing that fixes it is re-installing the firmware with an erase.

Also, I've noticed that when I do not read from it for a while, and then connect to it to read from it, it dumps a huge chunk of data, as if it'd been quietly buffering it.

Could it be that leaving the device running without reading from it, it buffers things up, and then runs out of space and starts overwriting key stuff in permanent storage ?

Also, a pointer to the code specific to the USB-serial emulation mode would be great so I can try to understand what's going on.

Thanks!

[emogenet]

So, an update:

I have installed command-line platformio, cloned this repo and recompiled the Simple/USBSerial firmware for my TUL.

The code, as cloned from the repo does not compile:

- I had to #define which transceiver to use (not sure why it's not in the platformio.ini file as a -D define)
- I had to add AsyncTCP as a lib_deps in the platformio.ini file to get it to compile (and let platformio decide which version to use out of the gazillion versions available, what a mess). Not sure why AsyncTCP is needed for something that just copies from serial to serial.
- I also had to create the target build directory for the final firmware to actually build properly.

The resulting firmware .factory.bin is much bigger (800-ish K) than the one found in the firmware directory (550k or so).

However, the very good news is that this newly compiled firmware does not seem to have the bug anymore: it's been running flawlessly for 12+ hours, dumping what it sees on my KNX bus.

The bad news is we don't know what the bug was, why the new binary works, and if it's fixed at all.