MeshSet produces corrupt mesh on first run unless dummy mesh is created and saved (works on rerun)

[JonasTriki]

When using pymeshlab.MeshSet to load and merge multiple OBJ meshes, the first run in a fresh Python process produces a corrupt output mesh (missing geometry, visuals, etc). If I rerun the code (with n_runs=2), the mesh is saved correctly—even without the dummy mesh workaround. Creating and saving a dummy mesh before the main processing also avoids the issue.

Minimal reproducible example:

import os
import pymeshlab

mesh_data_dir = "data/mesh_extracted" # Dir that contains .obj, .mtl and .jpg files per object. See attached `mesh_raw` ZIP and unextract to `data/mesh_extracted`
processed_mesh_path = "data/processed_mesh.obj"

# Get a list of all .obj files
mesh_filepaths = []
for root, dirs, files in os.walk(mesh_data_dir):
    for file in files:
        if file.lower().endswith((".obj")):
            mesh_filepaths.append(os.path.join(root, file))
print(f"Found {len(mesh_filepaths)} mesh files.")

n_runs = 1  # Change to 2 to 'fix' the issue, or uncomment the ms_dummy lines below

# Create and save a dummy mesh to initialize PyMeshLab properly
# ms_dummy = pymeshlab.MeshSet()
# ms_dummy.create_cube()
# ms_dummy.save_current_mesh("data/empty.obj")
# del ms_dummy

for run_idx in range(n_runs):
    print(f"Run {run_idx + 1}/{n_runs}")
    ms = pymeshlab.MeshSet(verbose=True)
    for mesh_filepath in mesh_filepaths:
        ms.load_new_mesh(mesh_filepath)
    ms.generate_by_merging_visible_meshes()
    ms.save_current_mesh(processed_mesh_path)

Symptoms:

• On first run after kernel restart, the output mesh is corrupt (missing geometry, invalid file, etc).
• On rerun (or with n_runs=2), the mesh is saved correctly.
• Creating and saving a dummy mesh before main processing also avoids the issue.
• Even if all input OBJ files load fine individually, the problem still occurs on first run.
• Workaround: Uncommenting the dummy meshset code before the main loop makes the first run work as expected.

What I tried:

• Using ms.clear() before loading meshes (does not help).
• Re-instantiating MeshSet (does not help).
• Running in a fresh kernel (problem persists on first run).

Environment:

• PyMeshLab version: 2025.7 (also tested 2023.12.post3)
• Python version: 3.12.5
• OS: macOS 15.6.1 (24G90)
• Installation method: pip

Expected behavior: MeshSet should work correctly on first run without needing a dummy meshset workaround or rerunning the code.

Actual behavior: MeshSet produces a corrupt mesh on first run unless a dummy mesh is created and saved first, or the code is rerun.

Additional info:

• No similar issues found in the public issues/discussions.
• I have uploaded raw data files, as well as example ZIPs with bugged and working output OBJ-files: https://we.tl/t-hxckXJIzSH. If the link does not work anymore, please comment and I will create a new link.