Rasbian Trixie

[jlpoolen]

I downloaded Rasbian's Trixie, installed and tried to run this project. Within the default Python environment:

 sudo apt-get install python3-opencv

I clone this project and tried to run PySpectrometer2-Picam2-v1.0.py, but I ran into an incompatibility with the NumPy installed on the system.

jlpoole@spectro:/usr/local/src/PySpectrometer2/src $ python PySpectrometer2-Picam2-v1.0.py
[0:54:53.039411102] [3073]  INFO Camera camera_manager.cpp:330 libcamera v0.5.2+99-bfd68f78
[0:54:53.067831324] [3076]  INFO IPAProxy ipa_proxy.cpp:180 Using tuning file /usr/share/libcamera/ipa/rpi/vc4/imx477.json
[0:54:53.073905326] [3076]  INFO Camera camera_manager.cpp:220 Adding camera '/base/soc/i2c0mux/i2c@1/imx477@1a' for pipeline handler rpi/vc4
[0:54:53.073969306] [3076]  INFO RPI vc4.cpp:440 Registered camera /base/soc/i2c0mux/i2c@1/imx477@1a to Unicam device /dev/media2 and ISP device /dev/media0
[0:54:53.074003380] [3076]  INFO RPI pipeline_base.cpp:1107 Using configuration file '/usr/share/libcamera/pipeline/rpi/vc4/rpi_apps.yaml'
[0:54:53.081826170] [3073]  INFO Camera camera.cpp:1215 configuring streams: (0) 800x600-RGB888/SMPTE170M/Rec709/None/Full (1) 2028x1520-SBGGR12_CSI2P/RAW
[0:54:53.082366773] [3076]  INFO RPI vc4.cpp:615 Sensor: /base/soc/i2c0mux/i2c@1/imx477@1a - Selected sensor format: 2028x1520-SBGGR12_1X12/RAW - Selected unicam format: 2028x1520-pBCC/RAW
Loading calibration data...
Loading of Calibration data failed (missing caldata.txt or corrupted data!
Loading placeholder data...
You MUST perform a Calibration to use this software!


Calculating second order polynomial...
           2
3.125e-05 x + 0.4375 x + 380
Generating Wavelength Data!


Done! Note that calibration with only 3 wavelengths will not be accurate!
Traceback (most recent call last):
  File "/usr/local/src/PySpectrometer2/src/PySpectrometer2-Picam2-v1.0.py", line 261, in <module>
    intensity = savitzky_golay(intensity,17,savpoly)
  File "/usr/local/src/PySpectrometer2/src/specFunctions.py", line 138, in savitzky_golay
    b = np.mat([[k**i for i in order_range] for k in range(-half_window, half_window+1)])
        ^^^^^^
  File "/usr/lib/python3/dist-packages/numpy/__init__.py", line 400, in __getattr__
    raise AttributeError(
    ...<3 lines>...
    )
AttributeError: `np.mat` was removed in the NumPy 2.0 release. Use `np.asmatrix` instead.
jlpoole@spectro:/usr/local/src/PySpectrometer2/src $

So I created my own Python environment that shares the system's modules, notably opencv, and then I downgraded to numpy:

 python3 -m venv --system-site-packages ~/pyspec-venv
 source ~/pyspec-venv/bin/activate
 pip install --upgrade pip
 pip install "numpy<2.0"

I also was able to work remotely through an SSH shell and have the PySectrometer2 appear on my workstation. In Trixie, ssh is de-activated by default. So:

 sudo systemctl enable ssh
 sudo systemctl start ssh

I logged into my Raspberry Pi4B with:

 ssh -Y [username]@[server name]

[jlpoolen]

Further testing is showing many things do not work. I'm working with ChatGPT and asked it to provide a summary of what it and I are facing. Here is the summary:

Here’s a concise, technically accurate summary you can use to update your GitHub issue — it explains both what’s broken and why, and outlines the migration path we’ve taken to modernize PySpectrometer2 for Debian Trixie / Bookworm+ environments.
Feel free to quote or adapt this directly for the issue description.

---

Summary: Why PySpectrometer2’s GUI breaks under Debian Trixie

1. Difference in Windowing Systems

• Bullseye (the previous RPi OS) used an X11 session by default.
  OpenCV’s cv2.imshow() relied on X11’s “HighGUI” backend (GTK2/GTK3) for window creation, resize, and event handling.

• Trixie (current RPi OS) defaults to Wayland, with OpenCV now compiled against Qt5 instead of GTK.
  Under Wayland, OpenCV’s HighGUI layer no longer controls the drawable area directly; Wayland enforces strict surface sizes and the “autosize lock” appears.
  The result:

  • Windows open, but the image canvas remains fixed (cannot enlarge).
  • The title-bar X close button doesn’t always send a destroy event.
  • getWindowImageRect() returns constant values, so resize logic breaks.
  • Fullscreen toggles (WND_PROP_FULLSCREEN) no longer behave as intended.

In short, the OpenCV HighGUI layer isn’t a full GUI toolkit anymore under Wayland—it’s a thin compatibility shim meant for quick demos.

2. NumPy ≥ 2.0 Compatibility

• Trixie ships NumPy 2.x, which removed some legacy matrix APIs and changed certain default types.

• The original specFunctions.py used np.mat and operations that implicitly returned matrices. NumPy 2.0 removed that alias, causing crashes like:

  AttributeError: module 'numpy' has no attribute 'mat'
  

• We patched the code to use plain np.array operations and ensured all return types are standard ndarrays.
  (This fix is complete and backward-compatible with NumPy 1.x.)

3. Path Forward — Transition to a Native Qt Display

To restore reliable GUI behavior, we’re migrating the window layer to Qt 5, which both OpenCV and Raspberry Pi OS Trixie already include.

What stays the same:

• All existing image-processing and spectral-analysis code (cropping, Savitzky–Golay smoothing, peak finding, wavelength calibration).
• All OpenCV drawing (cv2.line, cv2.putText, etc.) on NumPy arrays.

What changes:

• We remove all HighGUI calls (namedWindow, imshow, waitKey, etc.).
• The main script (app_qt.py) wraps the processed frame in a QWidget using PyQt5’s QPainter.
• The camera feed (Picamera2 RGB888) is composited via NumPy and displayed as a QImage.
• Resize, fullscreen, and keyboard shortcuts (f, r, space, q, etc.) are handled natively by Qt and now work both on the Pi console and over SSH/X-forwarding.

4. Current Status

• The Qt window opens, resizes, and closes correctly on both console and remote sessions.
• The core image pipeline runs but is being refactored into a build_composite_frame() function (pure NumPy + OpenCV) callable from the Qt layer.
• Once integrated, the program will display the live camera spectrum exactly as before, but with modern, Wayland-safe window management.

5. Takeaway

Trixie didn’t “break” PySpectrometer2’s math—it broke its window manager assumptions.
Migrating from OpenCV HighGUI to PyQt fixes those limitations, ensures future compatibility with Wayland, and simplifies integration with newer Picamera2 and NumPy 2.x libraries.

[gomills]

Hi. Yes the original code uses deprecated NumPy expressions. I recommend you not to work with a virtual environment since due to OS-level interactions, virtual environments cannot be used. Also, if you're having problems you could try out my refactor, it's in the pull section. I haven't been able to try it with real hardware, just mocking testing.