Currently, the glider is fully simulated to determine the path for its shadow. On the other hand, all other skills merely run an approximation - in some cases this is sufficient for all cases (eg. Bomber), in some cases fairly detailed but not perfect (eg. Basher; which sometimes comes up incorrect and doesn't account for up/down, but overall gives a good approximation), and in some it can be quite lacking (eg. Builder; which shows the path of the bridge but doesn't give any information on rebounding off blockers, where the bridge will actually stop, etc).
Would I be right in assuming people would prefer these to also show accurately-simulated previews?
Doing full simulations for all skills is already on my to-do list.
Alright. I was going to look at it myself, but if you're already planning it, I'll leave it to you. :)
Are you expecting to do it soon, or might it be a while? (To determine whether or not to wait for this before releasing V1.48n.)
Simulations are added for all shadow types now.
But there are still a few inaccuracies:
- Copied bricks from builders turning around at blockers/force fields are not visible in shadows. However all further bricks that are placed after turning around are at the correct position.
- Bashers stop slightly too early in front of steel. This is because the stop-working-checks in normal game-play use a simulation and simulating lemming within a simulation doesn't work too well.
- If bashers or miners move more than one pixel up or down, there will be a small gap in the shadow pixels.
- If bashers or miners turn around at blockers, there will be a few unnecessary shadow pixels placed in the middle of the tunnel.
There is still the problem how these shadows should be colored (http://www.lemmingsforums.net/index.php?topic=2892.0), to ensure good visibility in all styles.
This is still a HUGE improvement over before. :) Previously, many common cases would give misleading shadows; based on your descriptions and quick testing, now it's only fairly uncommon edge cases and/or those that can easily be filled in by a human prediction rather than requiring trial-and-error (or in-depth physics knowledge).