What does "clip model outside world bounds" mean? (TDM 2.10)

[thebigh]

I loaded up a mission I haven't worked on for a while, dmapped it, and tried to play it. As soon as I did, my frame rates zoomed to about 0.5fps and my console was filled with error messages telling me that the clip models for player1, the guards, and all the rats were outside world bounds.

The map worked fine a month ago when I tested it last and I haven't modified it since. All I've done in the meantime is update to the latest 2.10 build.

What is going on?

[stgatilov]

According to the code, the warning means that the bounds of an entity and the bounds of the whole world's collision map (expanded by 128 units) are disjoint, which means that entity is outside playable area (dropped down into abyss, or got through wall).

No idea why it happens for you.

[nbohr1more]

I believe this is a symptom of a map that has brush leak to the void.

[thebigh]

So it seems this bug is affecting everything in the map with physics attached to it: me, rats, moveable loot, crates, etc-- so I'm assuming everything with a clip model.

I doubt if it's a leak to the void because I don't have a .lin file, and if I seal myself into my starting room by turning the visps out of it to stone all the moveables inside and outside the room still show the error.

Physics is messed up: fired arrows and thrown objects remain hovering in midair.

Luckily for me, I keep at least 20 previous versions of the map as backups. So I was able to determine approximately when the bug was introduced. It seems I hardly did anything in that session- I just turned a few models into movable buttons and levers and such. The only unusual thing about these is that they're models that I rescaled them using the old rotation matrix hack to get them to fit snugly.

Now here's something really funky. If I dmap and run a buggy later version after running a non-buggy earlier version, it's fine. Restart TDM and load that last map and it's a slide show again.
Looking at the console log, I also see the following errors. I overlooked these earlier because of all the blabber about rats. I have never seen these before and I suppose they must be related:

WARNING:Compiling shader file volumetric.fs failed:
ERROR: 0:136: 'function' : is removed in Forward Compatible context
texture2DProj
ERROR: 0:136: 'function' : is removed in Forward Compatible context
texture2DProj
ERROR: 0:136: 'texture2DProj' : function is not known
ERROR: 0:135: '=' :  cannot convert from 'const highp float' to '4-component
vector of highp float'
ERROR: 0:137: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:137: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:137: 'texture2D' : function is not known
ERROR: 0:136: '=' :  cannot convert from 'const highp float' to '4-component
vector of highp float'
ERROR: 0:147: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:147: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:147: 'texture2D' : function is not known

 

[nbohr1more]

What happens if you replace the "rotation hack" scaled models with models scaled in Dark Radiant ( native function )?

[stgatilov]

4 hours ago, thebigh said:

Looking at the console log, I also see the following errors. I overlooked these earlier because of all the blabber about rats. I have never seen these before and I suppose they must be related:

WARNING:Compiling shader file volumetric.fs failed:
ERROR: 0:136: 'function' : is removed in Forward Compatible context
texture2DProj
ERROR: 0:136: 'function' : is removed in Forward Compatible context
texture2DProj
ERROR: 0:136: 'texture2DProj' : function is not known
ERROR: 0:135: '=' :  cannot convert from 'const highp float' to '4-component
vector of highp float'
ERROR: 0:137: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:137: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:137: 'texture2D' : function is not known
ERROR: 0:136: '=' :  cannot convert from 'const highp float' to '4-component
vector of highp float'
ERROR: 0:147: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:147: 'function' : is removed in Forward Compatible context texture2D
ERROR: 0:147: 'texture2D' : function is not known

 

These are errors in compiling WIP volumetric light error.
@duzenko might be interested in seeing them, but this shader is only used if you use "volumetric" material keyword, and it only affects graphics: shaders never affect gameplay.

So, it is completely unrelated to your problem.

[stgatilov]

You can check if the map works properly on TDM 2.09.

If it does, then bisect over dev builds of 2.10 and see which dev builds is the first non-working.
After that we can looked at changelog and maybe we'll guess what is the problem.

Although I think at least some understanding about how it happens would be welcome.

By the way, when player1 is said to be outside world, where are you located? Are you somewhere in the void, or are you still standing at starting spawn location?

[thebigh]

8 hours ago, nbohr1more said:

What happens if you replace the "rotation hack" scaled models with models scaled in Dark Radiant ( native function )?

Unfortunately, not every model can be scaled using the model scaler. It won't work on book models, for instance. For these I have to use the rotation hack. Anyway, I don't think that's the issue because the problem persists even after deleting the rotation spawnarg on all those hacked models.

 

6 hours ago, stgatilov said:

You can check if the map works properly on TDM 2.09.

It works fine in 2.09a. The problem seems to have been introduced in the very latest 2.10 dev build: 16358-9588 works fine, 16368-9627 is a slide show.

 

6 hours ago, stgatilov said:

By the way, when player1 is said to be outside world, where are you located? Are you somewhere in the void, or are you still standing at starting spawn location?

I'm in valid player space. I can walk around, run, jump, and frob things as normal. And using killmonsters to remove the rats helps a little, though fps plummets again if I pick up a carryable.

[stgatilov]

Which value of "Uncapped FPS" do you use?
Does the problem happen with both values?

[thebigh]

18 minutes ago, stgatilov said:

Which value of "Uncapped FPS" do you use?
Does the problem happen with both values?

I normally play with uncapped FPS off. Turning it on (166) makes no difference.

[stgatilov]

Can the problem be reproduced on released FM?

Can I somehow get a local copy of your FM for debugging?

[thebigh]

I haven't seen the issue on any released FM, but I've only played a handful since the latest dev came out.

Since I'm a bit hesitant to release a crap unoptimized WIP into the wild, and I definitely don't want to spoiler people on the cool puzzle I'm building, I'll PM you a download link. The maps folder contains the last non-buggy version and my latest test version with visportals turned to stone and all the scaled props deleted altogether.

Watch what happens when you dmap and run the earlier map and then the later one in the same TDM session- suddenly it works!

[stgatilov]

I reproduced the error on current SVN.

The warnings are generated because world bounds are "(0,0,0) - (0,0,0)". Since most entities are far from origin, this warning is printed for them.
It also explains the weird "memory effect" that you are seeing: proper bounding box was loaded during previous map start, but bounding box is not updated updated during current map start, so the bounding box from the previous start retains. It might be totally incorrect if two maps are unrelated, but in your case it stops the warning spam.

I did not get to the root cause yet, but will do in the future.
Most likely something about loading collision models got broken. I refactored/optimized something related to collision models loading in the latest dev build, and probably did not test it well enough to notice that I broke stuff real hard.

Meanwhile, use previous dev build.

Also, if someone else sees the same problem on the latest dev build, please report it here.
I'm not yet sure if the problem happens always, or is specific to this map...
 

[thebigh]

Odds are I've committed some bit of uniquely bad mapping practice that's now interacting with the collision model generation in the latest update. My maps only look somewhat tolerable on the surface; behind the scenes it's a mess of ugly hacks, kludges, and heinous sealing geometry.

[stgatilov]

I fixed the issue in svn rev 9643.

For the previous dev. build, I fixed some weird behavior which looked like a bug, but it turned out that some code was working only thanks to that weird behavior, and hence it broke.

Here is the difference between working and broken maps:

• If .map file has "name" spawnarg set on the worldspawn, then the map loading was broken.
• If .map file does not have "name" spawnarg on worldspawn, then it map loaded fine.

Of course, some existing FMs (e.g. Window of Opportunity) have "name" spawnarg, so that was a serious issue.

The next dev build should work fine in all cases.

[thebigh]

Yep, my old WIP now works fine again.