Rising door bounces an imp to the ceiling (Doom2, MAP01)

[simc]

I'm sorry to bring this matter up again. I wrote earlier a similar topic about demons and it was judged as unfixable behavior.
http://forum.zdoom.org/viewtopic.php?f=7&t=51975

But I now see that in Doom2 MAP01 an imp next to a door suddenly jumps up when the door opens and other imps can walk under it. After that the imp falls down and continues normally. Because this does happen when the lift is still down it can't be the upwards going player pushing the imp up. This is *not right*.


This anomaly does occure with these ZDoom versions:
zdoom-2.9pre-948-gebca170
zdoom-2.9pre-928-g841a579
zdoom-2.9pre-854-g1722902
zdoom-2.9pre-506-g1648a71
zdoom-2.9pre-415-g346fae8
zdoom-2.9pre-362-g9b23b41
zdoom-2.9pre-270-g60c2a55 (Feb 26th 2016 22:04)

But I can't get it happen with:
zdoom-2.9pre-154-gb9037ef (Feb 16th 2016 13:46)
zdoom-2.9pre-95-ge2de258
zdoom-2.8.1 (Feb 21st 2016 20:14)

So there must have been a fairly recent change to cause it.

To test by your own:
1) Get in the main hall of the Doom2 MAP01 in UV (save the game for faster reruns)
2) Switch the lift down.
3) Before the lift goes up hit the back wall to open the door.
4) Quickly back off so that imps wake up during the opening of the door and when the lift is still down.

If you go slightly to the *left* while backing off the imp goes up with the door almost every time. Here's also a demo file (20 s) recorded with zdoom-2.9pre-928-g841a579. Run ZDoom with Doom2 and type into console 'playdemo [path]demo_ceilingimp.lmp'.

[Nash]

Where's the demo file?

[Graf Zahl]

Problem is, the way this code was written the current outcome is precisely what it's supposed to be. That doesn't mean that I don't consider the entire algorithm pretty much broken. But I simply cannot find the difference that made it not trigger in older versions.

EDIT: Found it, it just was in a completely different place than what I would have expected.

[simc]

The demo file. It works also with the latest zdoom-2.9pre-948-gebca170.

[Graf Zahl]

Oh, just a final note: If you manage to get some older 2.0.x versions working again, you'll see this effect all over the place. The only reason it got less frequent was several attempts to patch this glitch. I doubt it's completely fixable but this should be addressed in a more stable and predictable manner. Much of what happens here depends on the order in which actors are stored.

[NeuralStunner]

Oh man, nice to see the "picked up an imp stack" bug addressed again.