[adrian-andrzjewski]

Fix

Point light shadow lookups could self-shadow receivers at grazing angles. Apply a small receiver-distance-scaled bias, capped at 0.01 world units, to shadow lookup positions in both point-light post-filter shaders. The receiver's original world position remains unchanged for filter calculations.

Validation

• Focused :jme3-core:test shadow tests: 5 tests passed across FilterPostProcessingTest and ShadowCompareModeFallbackTest.
• OpenGL 4.5 visual reproduction on an RTX 2060 SUPER: the before image shows the point-light shadow acne stripes; the after image confirms they are cleared.

Closes #2866

[jaime-jmebot]

Hi Adrian, thanks for putting together this fix and including both the focused test results and a visual reproduction. The change keeps the reconstructed receiver position intact for the filter calculations, while consistently using the small, capped lightward offset for the point-shadow lookup coordinates in both shader variants. That makes the bias apply to the comparison without perturbing the other calculations, and the zero-distance guard avoids normalizing a degenerate light-to-receiver vector.

I reviewed the paired shader changes and their surrounding point-light lookup flow. I don’t see a correctness issue with this approach, and it’s good to have the multisample and non-multisample paths kept in sync. The reported tests and visual check provide useful coverage for a rendering artifact like this. Looks good to me—thanks for addressing the acne report!