Are there any issues with combining simple shapes to create more complex shapes - where the mesh is not “hermitically” sealed - or should the mesh not have intersecting shapes.
In the the image on the left, the object is constructed from four intersecting meshes - on the right the object is constructed as a single mesh.
Would either work? Are there known issues with light bleeding etc. Obviously the left object was easier to create and has a slightly smaller file size.
Z fighting is probably the biggest issue in most/all engines, though I haven’t seen that issue for a while.
Light bleed is also something to watch for, as you stated, but with the bias setting I think this can be reduced/eliminated.
Please share results from any tests you perform.
Note that you’ll get the effect of light bleeding from edges if the texels (the lightmap’s pixels) cross over the boundary edge, or you don’t have enough texel resolution in your lightmaps and the lightmaps are exposed to open air or another lit area (thus getting light contribution on half or very-hard-to-get-precision exposed texels from another lightsource or ambient light).
For a modular design, it’s worth having walls that aren’t paper thin. You can get around this with blocking geometry which has no lightmap contribution (it’s lightmap static, but has no lightmap scale so isn’t assigned to a lightmap), which you hide or destroy after baking or at runtime…give it it’s own tag for easy selection.
Increasing lightmap resolution will often help though it’s not always necessary as a first resort.