S, all the fluid that is going up is going up on the left side, from 7s. The fluid on the right is being filled *from above* by the river, never from the bottom. There is no water popping up illegally, that water is coming from above. Left sides always fill from below after the bottom has filled with 7s. The camera view goes down at first, and I only scroll back up when the lowest floor is completely filled with 7s. Then the right side fills from above until it is 7s, and at that point the pressure model kicks in, and fills that floor on the left side from below, through 7s. Now, your edit changes focus halfway through to fluid *not* flowing up the other side until the water at floor -1 is up to 7s. That is a different issue than filling up from squares that aren't 7. It is true that the model does not attempt to equalize if there is a height disparity of only one z-level. This is a limitation that comes from processor speed, given the way I'm doing it, but it doesn't have anything to do with illegal fills from below.
edit: If the issue is that there is any water on the right side floor -1 at all while the bottom is still filling up, that's because the water doesn't flow infinitely fast. The upper level of a river can still have deep water even if it ends in a waterfall.
Staircases and ramps are just treated as holes. People complain about the fluid speed a lot -- this is due to both the processor load if I made it go faster, and the random nature of its movement. I'm fine with the speed the way it is. Water always falls down when it can.
The 3-D view is not currently recorded by the movies, but it contains less information than the ASCII frame so it's not a problem. Just another thing on to-do.
[ April 10, 2007: Message edited by: Toady One ]