Sunlight appear from under buildings

[Dave Nelson]

I noticed this in V1512 but it may have been introduced earlier.

Rotating camera results in sunlight appearing on the ground on the shadow side of large static object.

[Wayne Campbell (wayneinbc)]

I am not sure what you are showing here. Please point out the problem.

[Dave Nelson (lake-muskoka)]

In the attached-to-the-bug-report image (Shadows3.jpg) there is a triangle of sunlight emanating from the joint formed by the building on the left and ground... apex at the building, spreading out towards the cars.

Were I to rotate the camera slightly to the left the triangle will change with the angle at the apex reducing until the light disappears. Rotate to the right and that angle increases. There should be no light at all; It's 5:15 pm; the view is looking NNW and the wall where the light is emanating from is the eastern facing wall. It should be casting a full shadow w/o regard to camera rotation.

[Wayne Campbell (wayneinbc)]

Ahhh.. the brown panel on the left is the vertical face of a building - I thought it was a terrain texture.

I think this may be related to Bug #1098600. And from my experience is more likely to occur if the scene objects are composed of large mesh triangles - ie > 100m in any dimension. I had some of this on the L&PS and was able to improve the shadows by breaking up such meshes into smaller sections. We'll keep the report open however as ideally it should be fixed. Meantime you might be able to try the above mentioned technique.

[Dave Nelson (lake-muskoka)]

I chopped up the building in many layers, each ~4m tall. The problem remains. Here's a screenshot that takes in the sun, the building, and cars that should be entirely in the shade. Note light passing between the two boxcars.

[Dave Nelson (lake-muskoka)]

And one more, having dropped the camera lower so it too is in the full shadow of the building... light still passing between cars.

From another angle it appeared there was an concave arc on the illuminated area on the lit edge closest to the building. I've seen that arc elsewhere as an improperly located shadow on roofs... makes me wonder if somehow one of the shadow maps has been inadvertently inverted.

[Dave Nelson (lake-muskoka)]

Here's that arc as a shadow that I mentioned from a screenshot I took ~7 weeks ago. The arc I saw this morning was very similar in shape & size to this shadow arc.

[disc (disc86543)]

I see shadowing problems too, isn't it the same as yours?

See my attached video.

[Wayne Campbell (wayneinbc)]

Kerko, you already reported this as Bug 109860. And it a fix candidate was submitted at V 1508.

If you are still having a problem, please add to your original report.

[disc (disc86543)]

That wasn't exactly like this, but... ok. BTW im using v 1517

[James Ross (twpol)]

Some shadow works have been done in X.1597 and X.1600, could you see if this situation has improved?

[Dave Nelson (lake-muskoka)]

Looks good now. Thanks.