1024cla2.wad MAP29 monster teleport fail (NOT pointonline)

[Player701]

MAP29 of Claustrophibia 1024 2 exhibits buggy behavior when played in GZDoom. Two UV-only revenants from the initial monster wave fail to spawn in.

To reproduce this bug, use the following command-line parameters:

Code: Select all

gzdoom -iwad doom2 -file 1024cla2.wad -skill 4 -warp 29

Upon entering the map, use the IDDT cheat twice to reveal everything and everyone. Find this area:



and wait. In a short while, most of the monsters will teleport out of there - all but two of them, which ultimately prevents 100% kills on this map.

Note that this issue is not fixed by enabling compat_pointonline.

Tested in GZDoom 4.5.0 and g4.6pre-22-ga6a02a2a6. In PrBoom-Plus 2.5.1.4, every monster teleports out normally.

[Player701]

Here's an isolated test map, as well as some more info on what goes on in there.

The revenants are placed on a scrolling floor, and their path is initially blocked by some barriers. For some reason, when the easternmost barrier lowers, the revenant there is somehow able to pass through the teleporter trigger without activating it. This is most definitely a bug. It does not happen in PrBoom-Plus.

There is another trigger line further ahead, which is supposed to teleport the monster to the western wall so that it can attempt to pass through the teleporter again. However, by the time it reaches the second line, the revenant to the west hasn't teleported out yet. Since it is still blocking the way, the second trigger line fails as well, and the revenant to the east gets stuck.

If I tie the barriers to a manually-activated switch, the bug cannot be reproduced reliably. Sometimes the second revenant can teleport out, and sometimes it cannot. But if I use the same mechanism as the one that's used in Claustrophobia 2 MAP29, the bug happens all the time. In my test map, I've also reduced the time needed to wait before the bug occurs.

[Graf Zahl]

From a quick glance that setup looks broken by design, falling apart at the slightest deviation of the expected outcome. I think the entire problem here is that this needs compat_boomscroll to work because ZDoom's averaging scrollers do not move the monsters fast enough.

[Player701]

Unfortunately, compat_boomscroll doesn't seem to fix this bug for me.