Offscreen Renderer need "two passes" to render correctly

[rhesus-ltd]

Hi there

I'm using:

pythonocc-core            7.4.1            py37h2700f40_0    conda-forge

on Ubuntu 18.04.2 LTS

I see a strage behaviour I'm not sure if this is a bug or a mismatch in my setup.

I have the following code in a Jupyter-Notebook:
(A:)

from OCC.Display.WebGl.jupyter_renderer import JupyterRenderer
from OCC.Extend.DataExchange import read_step_file_with_names_colors
from OCC.Core.Graphic3d import Graphic3d_BufferType
from OCC.Display.OCCViewer import Viewer3d

from PIL import Image
from PIL import ImageOps

my_renderer = JupyterRenderer()

W = 196
H = 128

filename = 'CAPC-0805-T0.95-BN.stp'
shapes_labels_colors = read_step_file_with_names_colors(filename)

and use the following to render an image of the step file:
(B:)

offscreen_renderer = Viewer3d()
offscreen_renderer.Create()
offscreen_renderer.SetSize(W, H)
offscreen_renderer.SetModeShaded()

for shp in shapes_labels_colors:
    label, c = shapes_labels_colors[shp]
    offscreen_renderer.DisplayColoredShape(shp, color=c, update=True)  

data_raw = offscreen_renderer.GetImageData(W, H, Graphic3d_BufferType.Graphic3d_BT_RGB)
img = Image.frombytes('RGB', (W, H), data_raw)
img = img.rotate(180, expand=True)

Now if I look at the image it looks like this (regardless of jupyter-Renderer or written to disk using Pillow:

The magic happens when I run the Block B a second time. Then the image becomes:

This can be reproduced outside of jupyter as well.

But: If I use View.Dump to write a PNG file the file is correct the first time.

Thanks and feel free to close if this is not the right place to mention such an issue.

• addedd better example and step file
  CAPC-0805-T0.95-BN.stp.zip

[tpaviot]

I'm not aware of this issue, thank you for providing the STEP file I'll have a look.