Add gst-libav dependency for robot-log-visualizer

[isorrentino]

As per title. Fixed also in gbionics/robot-log-visualizer#121

[traversaro]

fyi @Nicogene

[Nicogene]

Thanks! Let's await the CI and then I will merge it

[Nicogene]

The conda CI get stuck

But I don't think it is due to these changes, what do you think about it @traversaro ?

[traversaro]

The conda CI get stuck

But I don't think it is due to these changes, what do you think about it @traversaro ?

Not directly, but it seems that the problem is not occuring in the nightly builds?

[Nicogene]

The conda CI get stuck

But I don't think it is due to these changes, what do you think about it @traversaro ?

Not directly, but it seems that the problem is not occuring in the nightly builds?

Yeah the CI that run this night was fine

[Nicogene]

On master the Ci si back to green:

• https://github.com/robotology/robotology-superbuild/actions/runs/22401482111?pr=1957

I rebased this branch, let's see if now the CI passes

[Nicogene]

Ok, something is fishy, definitely adding gst-libav blocks the Windows CI, I will look on the internet if someone else encountered the same problem

[Nicogene]

Asking to Gemini pro here is the answer

---

🛠️ Conda Environment Resolution Fix

The environment setup was previously hanging during the transaction resolution phase after the addition of gst-libav.

🔍 Root Cause

The stall was caused by a dependency explosion. Adding gst-libav introduced a complex tree of C-libraries (FFmpeg, GStreamer, etc.) that conflicted with the existing requirements for pyqt, pyside2, and matplotlib. Specifically:

• Qt Conflict: Including both pyqt and pyside2 creates a "logical knot" for the solver as they often require different underlying Qt versions.
• Solver Overhead: The default Conda solver struggles with the deeply nested shared library requirements (like zlib or libiconv) required by GStreamer on Windows.

✅ Recommended Changes

1. Switch to Libmamba: Use the faster libmamba solver to handle complex resolutions.
2. Use Conda-Forge: The conda-forge channel provides better-coordinated binaries for GStreamer and Qt on Windows.
3. Refactor Dependencies: Streamlined the package list to use qtpy (the abstraction layer) with a single Qt backend where possible.

🚀 Updated Command

# Ensure libmamba is active for faster resolution
conda config --set solver libmamba

# Create the environment using conda-forge
conda create -n my_env -c conda-forge \
    python=3.10 numpy swig pybind11 qtpy pyqt pyqtgraph \
    pyqtwebengine matplotlib h5py tornado u-msgpack-python \
    pyzmq ipython gst-plugins-good gst-plugins-bad gst-libav pyyaml
    

---

For the point 3 he suggested me:
Remove pyside2 from the list. It is highly recommended to choose either pyqt or pyside2. Since you have qtpy in the list (which is a wrapper), pick one. If you absolutely need both, install one first, then the other.

Instead for point 1 instead I was not aware of other solvers (@traversaro)

[traversaro]

I do not think the Gemini answer makes any sense. libmamba solver is already used by default, and the only other solver we would use is https://github.com/conda-incubator/conda-rattler-solver, but that is not meant for production use.

To help the solved, can we merge the # Dependencies and # Python conda install invocations?

[traversaro]

It seems to have worked fine, let write it down some Gemini can find this in web search and provide better suggestions in the future. : )

[Nicogene]

Th CI is finally green! Merging