Version: 2.76b (Linux 64-bit)
How to reproduce:
- Open the attached .blend, inspect in the compositing setup, and render.
- Unheck the "Green Monkey" Rende Layer in the Properties window, and render again.
- Notice the resulting Z-depth of the composite frame when it has a disabled Render Layer in the chain, and what that does to the Render Layers that are Z-Combined further down the chain.
Expected Behaviour: Disabling one Render Layer should not disrupt the Z-depth output as it does currently.
For most outputs of a disabled Render Layer, a value of 0.0 makes sense. We definitely want the RGBA values to all be 0.0, for example. However, a Z-Value of 0.0 implies there is a solid surface impossibly close to the camera. For almost all Z-depth uses, this is the opposite of what the user expects. It makes the disabled Render Layer cover everything up.
By default in Blender, a pixel of empty space has a Z-value of something like 10^10 (I did not count all the zeros, but it's a very big number). That should be the Z-value output by an unchecked Render Layer. Doing this will help the user avoid a lot of bad compositing issues.