Capx: dl.dropbox.com/u/5731948/FallThruBug.capx
Demo: dl.dropbox.com/u/5731948/FallThruBug/index.html
Conditions tested on:
Windows 7 Ultimate (latest)
Chrome (latest)
Radeon 7690M
Also tested on several other systems/browsers with similar results but this is my main machine
Setup:
This bug is a bit tough to see if you're not sensitive to differences smaller than 5 frames and therefore that makes it a bit hard to explain. Nonetheless, in the demo provided, there are 3 columns of fall thru platforms (FT for short). For convenience, I've labeled the FTs in this screenshot
<img src="http://snag.gy/pCcdr.jpg" border="0" />
Bug:
Ok, so when you actually land on an FT with another FT close above it (i.e. 2b and 2c), it usually takes at least 2-3 frames longer to land than it should. To test this you can idle at 2a for 2 seconds and then hit the down key to fall to 2b. Then try the same thing at 3a. It will take about 21 frames to drop from 2a and 22 frames to drop from 3a. These times are really close to eachother considering 3a-3b is nearly twice as far apart as 2a-2b. If you look closely, you can even see the Player hang for a bit above 2b (which doesn't happen above 3b).
*NOTE: if you don't idle for 2 seconds at 2s, you might see the Player fall through as quickly as 19 frames. This is much closer to what I'd expect. Also, droping from 2b-2c can be as fast as 15 frames!
You can test this furthermore by jumping from the floor to to 2b. It doesn't land immediately and you can notice a 2-3 frame "bounce" before Player lands. Whereas if you jump on 3b you notice no such bounce.
This issue may seem trivial but it's very important for a twitch platformer of which I'm creating. Speed runs are differentiated by hundredths of seconds and each time this happens it adds another couple hundredths. And placing platforms close together is important to my design. Please, if nothing else, take this bug seriously for the amount of detail I've put into it.