Category: Python

fatal problem with esp-idf 5.2.3 building micropython 1.24.0 preview

whbeebe

Espressif released ESP-IDF version 5.2.3 two days ago from the date of this post. I installed a copy and then compiled the latest version of MicroPython I have on my system. When I then flashed the build onto one of my Espressif boards, an ESP32-S3 N32R8 DevKit C board, the MicroPython software crashed, performing an infinite boot loop. This is one cycle of the boot loop I recorded via Thonny plugged into the board’s micro USB port.

MPY: soft rebootMemory: 8,308,816 MB Flash: 8,388,608 MB  Platform: MicroPython 1.24.0 preview xtensa IDFv5.2.3 Unique ID: 68B6B33D7814  SSID: ESP32S3-7814 CPU Clock: 160,000,000 Hz   I2C: SoftI2C(scl=2, sda=1, freq=500000)   I2C: Devices found: ['0x3d']   I2C: SSD1306 OLED FoundA fatal error occurred. The crash dump printed below may be used to helpdetermine what caused it. If you are not already running the most recentversion of MicroPython, consider upgrading. New versions often fix bugs.To learn more about how to debug and/or report this crash visit the wikipage at: https://github.com/micropython/micropython/wiki/ESP32-debuggingMPY version : v1.19.1-2669.g17d823445.dirty on 2024-09-27IDF version : v5.2.3Machine : Generic ESP32S3 module with Octal-SPIRAM with ESP32S3Guru Meditation Error: Core  1 panic'ed (LoadProhibited). Exception was unhandled.Core  1 register dump:PC  : 0x4005544b  PS  : 0x00060830  A0  : 0x82007cc1  A1  : 0x3fcbb590  A2  : 0x00000000  A3  : 0x3c131a50  A4  : 0x00000000  A5  : 0x3fcbb5e4  A6  : 0x00000000  A7  : 0x00000001  A8  : 0x82008a70  A9  : 0x3fcbb540  A10 : 0x00000000  A11 : 0x3fca4184  A12 : 0x00000001  A13 : 0xffffffff  A14 : 0x00000000  A15 : 0x3fca41ac  SAR : 0x00000016  EXCCAUSE: 0x0000001c  EXCVADDR: 0x00000000  LBEG: 0x400570e8  LEND: 0x400570f3  LCOUNT  : 0x00000000  Backtrace: 0x40055448:0x3fcbb590 0x42007cbe:0x3fcbb5a0 0x4200837e:0x3fcbb5e0 0x4200dcbe:0x3fcbb610 0x42015279:0x3fcbb630 0x42015341:0x3fcbb650 0x40378e76:0x3fcbb670 0x4200ddc7:0x3fcbb710 0x42015279:0x3fcbb780 0x42015341:0x3fcbb7a0 0x40378e76:0x3fcbb7c0 0x4200ddc7:0x3fcbb860 0x42015279:0x3fcbb8b0 0x4201528e:0x3fcbb8d0 0x42022fef:0x3fcbb8f0 0x42023347:0x3fcbb980 0x42006922:0x3fcbb9b0

Lines 3 through 11 are output by my software. Line 14 is where the crash dump starts. All the output text my software prints comes from a single file, boot.py, which is the first file that is always executed. The next file to be executed is main.py, which is where the NeoPixel is exercised (it displays six colors then turns off). The last thing main.py does is instantiate an instance of my web server, then kick it off. I suspect that the crash has something to do with the built-in WiFi radio.

At this point I’ve simply dropped back from 5.2.3 to 5.2.2, recompiled MicroPython, re-flashed my board, and get it running happily again, just like it was doing before I tried to step up to the latest ESP-IDF release.

I have no idea why it’s crashing, and I have no desire to go looking. Since there’s nothing special in version 5.2.3 that I just have to have, I’ll stick with version 5.2.2 for building MicroPython, and version 5.3.1 for writing C++ for the Espressif chips. Be advised that ESP-IDF 5.2.3, at least for the ESP32-S3 in this circumstance, has a critical internal problem.

building python 3.13.0 rc1 on a raspberry pi 5

whbeebe

This is an interesting time for Python development. Python continues it’s upward popularity increase, while the Python language developers keep pushing the Python interpreter to be more performant with each release. In particular, with the imminent release of version 3.13.0, the developers are now disabling the global interpreter lock, or GIL, for efficient concurrency. Building Python betas and release candidates has become something of a ritual for me. With the addition of the 512 GiB NVMe SSD on my Raspberry Pi 5, I decided to build Python 3.13.0 release candidate 1 on my Raspberry Pi 5 under Ubuntu 24.04.

I documented the build process flow in an earlier post (see in the link section below). I’m going to highlight one key step in that process flow, the installation of build support libraries.

sudo apt install zlib1g-dev \ libncurses5-dev \ libgdbm-dev \ libnss3-dev \ libssl-dev \ libreadline-dev \ libffi-dev \ libsqlite3-dev \ libbz2-dev \ tk \ tk-dev \ liblzma-dev \ libgdbm-compat-dev \ libmpdec-dev

Everything succeeded except the attempt to install libmpdec-dev. You need that library to support Big Number math in Python. In every release before 3.13 that support was internal. However, starting with 3.13 and later, that support is now via libmpdec-dev. Every one of those support packages installed successfully in Linux Mint 21.3, which is based on Ubuntu 22.04. But it doesn’t work on Ubuntu 24.04 on a Raspberry Pi 5. In the end I had to locate, download and manually install the individual deb files:

sudo apt install ./libmpdec3_2.5.1-1_arm64.debsudo apt install ./libmpdec-dev_2.5.1-1_arm64.deb

Those two deb files satisfy the libmpdec-dev requirement. After that I executed the following to configure the Python 3.13 build:

../Python-3.13.0rc1/configure --enable-optimizations --disable-gil

Once built I installed according to the step documented in the linked post below.

Then I gave it a whirl. The first thing I did after installing it was to create a Python 3.13 virtual environment and then pip install luma.led-matrix. That supports silly clock. When I activated the Python 3.13 virtual environment and ran silly_clock.py, the following was emitted by Python 3.13 in the terminal window:

<frozen importlib._bootstrap>:488: RuntimeWarning: The global interpreter lock (GIL) has been enabled to load module 'PIL._imaging', which has not declared that it can run safely without the GIL. To override this behavior and keep the GIL disabled (at your own risk), run with PYTHON_GIL=0 or -Xgil=0.

That’s going to be an interesting problem going forward, as there’s a ton of legacy Python code that will more than likely trigger that RuntimeWarning.

Note

In order to successfully install luma.led-matrix in a Python virtual environment on a Raspberry Pi 5 under Ubuntu 24.04, you will need to install yet another support library, libpng-dev.

Links

building python 3.13.0 beta 1 on linux mint 21.3