Funky Chicken Clone

[branko]

Man, that is some dance

(Sorry if reported ... I looked and looked but couldn't find it).

Shooting a prone clone makes the ragdoll effect last for about 30-loops, which is long enough (as opposed to 3 or so loops for a standing clone).

But, stand a clone on top of the vehicle spawn at an FB (TIP: type .supertroop to run quickly out there). Shoot the clone (so that it falls down onto the veh spawn roof).

Either climb onto the tent, or better yet, wander into the vehicle spawn ... and look up.

Man, that is some crazy dancing. And it goes on for hours and hours ... well 15 minutes plus; the hilarity had worn off by that point, so I despawned.

Now try it with about 8 clones up there ... That's one wild disco!

And speaking of supertroop. If I leave it ON when I despawn, and then try to respawn I CTD - No error messages or pauses, just a straight CTD. And no error log. Reproduceable. Not that it matters much ... just saying ...

[GOPHUR]

I'm more interested in the CTD. I never gett that and I've spawned with supertroop a hindred or more times today. Tell me more.

[GOPHUR]

Tell me how to do this dance a little better.

[branko]

Specs

Intel Core2 Extreme X9650 at 3GHz (4 cores)

9800GT x 2 in SLi Mode - SLi enabled, Physics disabled - Total Approx Memory 2286MB

Creative SB X-fi Sound Card

4GB Ram

Win 7 64-bit

Crash (log file entry in post)

I did a lot of testing yesterday on cover/concealment. I CTD twice and had a complete machine lock up once (in about an hour of play). At all times the crashes were with Supertroop ON.

HOWEVER - the night before I had updated my NVIDIA drivers (I'm sure they stealth release the blimmin' things!) - I was also able to despawn a fair few times with Supertroop ON and no CTD.

Only one log entry has been written/saved, and is it as follows:

ww2_log.txt

File version: 1, 31, 0, 18

Date and time: 1/8/2010 19:13:18

WWIIOL aborted via improper means

(The weird thing being, I am almost certain I did not CTD on that login ...)

From an UNRELATED CRASH (as in not the same date as the above ww2_log) I have the following DUMP file:

errinfo.dmp

1

16843009

39679

39679

0

0

3221225477

1342253803

$Id: wwiiClientMessages.cpp 1374 2009-05-21 17:08:55Z dchristopher@CRS $

Summary:

My NVIDIA update has enabled me to spawn/despawn with SuperTroop on (Sorry, I thought I was up to date) but I am still getting a few CTD (maybe 20% of the time) and I did get one hard lock up.

All errors are on despawn (I have never CTD/Crashed while running about in game).

Ignore it for the moment, Gophur, and will try to isolate the problem a bit more.

Funky Chicken

To repeat (and I can, everytime):

1. I spawned Brit Rifle at Schilde S CP and turned on SuperTroop

2. I ran out to the AXIS FB SW of town (51.12.41N - 4.37.56E)

3. I stood an infantry clone on the roof of the vehicle spawn tent

4. I walked over to the inf spawn, turned and shot the clone.

5. It took "0 at 295 seconds" before the clone stopped jitterbugging

6. I turned SuperTroop OFF, stood another clone on the tent roof - Went back down to ground and shot it

7. Same result, but this time "0 at 276 seconds".

8. I stood another clone on the tent and added a fourth Prone Clone

9. Climbed off the tent and shot each of them

10. "0 at 274 seconds" (the one that was standing) and "0 at 316 seconds" before the 3rd and fourth clones stopped their dance.

Note: Although all of these clones were not prone (3 were full stand) when shot, all prone clones (regardles of surface) take longer to process the physics (on my client) than a clone shot in any other state (on average, roughly 4 / 6 / 28 "hops" for standing, crouching and prone).

UPDATE

Having finished the "funky chicken" test above, I despawned, planning to do the test at another FB.

I got the "Out of Memory" Error:

(ww2_log):

Application exited: out of memory on d/1/9 2010:14:29

(No dmp file created)

I will restart the client and test the funky chicken at a different FB. I will update this post when done.

Funky Chicken Update:

I spawned at Arlon S AB and ran out to the W FB (49.41.53N - 5.52.4E)

Spawned one standing and one prone clone on top of vehicle spawn

"0 at 278 seconds" and "0 at 296 seconds" were the results.

So far, I am unable to find other surfaces that force the physics process to loop for so long (so I am sure it is surface related)

A further EDIT:

Church in Arlon (49.41.20N - 5.48.46E)

I stood a clone up on the wooden surface near the pulpit took 3 seconds to stop processing

I setup a prone clone towards the rear of that surface took "28 seconds" to stop processing

I setup another prone clone right at the front of the surface (over hanging the stone floor) - When shot it fell off the wooden pulpit surface onto the stone church floor - Took "2 seconds" to stop processing.

I can repeat with almost identical times

So a change in surface for a dead clone affects/stops the amount of processing required for the rag-doll physics. E.G. Tried same as above using concrete "wharf" in front of warehouse and cobbled road surface. A prone clone staying on the concrete wharf took "27 seconds" to stop dancing, one that fell down onto the cobbles took "6 seconds"

Edited January 9, 2010 by branko

[AHWULF]

There might be a leak in the clones, I was messing around with clones in Willemstad Castle and discovered that about 400MB or so had been used up.

[dekard]

in regards to the funky chicken dance, there's issues where it's colliding with floors or objects that are too thin. In so it falls through gets caught and comes back up, falls through, etc.. etc.. we maybe able to eliminate this with a thicker collider but there are other problems in doing that as well.

[branko]

I couldn't really care less about the clones (I guessed it was surface related) I was more concerned about once it goes live.

Watching 20-odd prone ei wriggling around after they die, for 30 seconds or so, will get old very fast.

Is it possible to turn off ragdoll depending on the surface the inf dies on? (I assume not).

Shame, as the ragdoll effect is good, but this will become a real nuisance once it's live (assuming it doesn't affect just clones?).

And there is something weird about the effect with prone inf ... It always takes many more hops (seconds) to process the ragdoll effect on a prone inf than on standing or crouching inf (regardless of surface type).

Leaky Clones

AHWulf: That may well explain my CTD and memory errors if there is. Whilst I wouldn't swear to it, I'm pretty sure all my crashes have been when clones were present (and almost always multiple clones).

[GOPHUR]

Dekard is going to tag that for no ragdoll which means they'll just fall through. Kinda like on fire escape railings.

[branko]

Good work around, Gophur/Dek.

Suggest some effort is put into finding surfaces with thin colliders and same fix done if not too bad from a prettiness point of view (like the wharfs in front of warehouses would not be too pretty with a drop through effect).

Possible targets:

* - Warehouse Wharfs (tested, looping is about 30-secs) - Not a good drop through target?

* - Church pulpet area (tested, looping is about 30-secs) - Not a good drop through target as, I assume, part of body could stick out through church walls? If not then would be OK.

* - Warehouse mezzanine floors (not tested, I guess will be as same above) - Drop through probably OK, esp. if bodies cannot collide through building walls

If I come across anymore, on my wanderings, I'll update this post.

Cheers

[gtanner]

FB tents (Vehicle and infantry) produce a loop of around 240 seconds, might have to add them to the list.

[GOPHUR]

Let's keep building the list. I think we might have to find a better solution if it gets too big.

[branko]

Then someone is going to have to come up with a number for X in the statement "X number of loops is too many", Gophur.

The 30-loop is not too bad, it is kind of like the final death throws from a non-clean kill (gut shot etc).

Question: After - say - loop 10, is it possible to affect (in code) what animation gets displayed?

A little bit of mild twitching is not too distracting (and like I say, a little bit real-world explainable), it is when the twitching is bad, or goes on for such a long time, that it becomes unworkable.

I will go test some surfaces that appear "thin" and update this post (at some point today).

[branko]

NOTE: If infantry in the live game will not behave as clones currently behave (when killed) then none of this is an issue!!

I am no longer sure that the "funky chicken affect" can/should be solved by "fixing colliders"?

A think collider may make things worse (like a tent roof) but I do not think it is the root cause of the problem?

THE BUILDING ROOF TEST

1. Climb/float onto a shallow house roof (test is too hard on a steep sloped roof!) - There are some "white" houses to the West of the church in central Schilde that are good

2. Setup a STANDING clone - close to the edge of the roof - and shoot it.

Result: It will almost certainly fall off the roof, and the physics loop will stop processing very quickly (3 or 4 seconds)

3. Setup a PRONE clone on the APEX of the roof, lying parallel to the apex (you are attempting to kill it but not get it to fall off the roof!)

Result: It will (usually) do one of two things:

a) It will flop over (on its back) and the physics will stop in under 10-secs

It will remain on its front (it is a bit challenging to get it to do this!) - In this state it will get the jitters - Physics loop will take 25+ seconds to complete.

ON THE GROUND TEST

1. Setup a PRONE clone on a large flat piece of ground (road, grass ... it don't matter)

2. Shoot it

Result: It will get the jitters and take 25+ seconds to stop jittering

3. Setup a STANDING clone in the same place

4. Shoot it

Result: The physics will stop after 2 or 3 seconds

CONCLUSIONS

The amount of jittering (if any) varies by the amount of INITIAL MOMENTUM of the corpse. The less INITIAL momentum the worse the jitters will be.

For example: Even a prone clone that tumbles off a roof will process correctly after the fall (a couple of seconds). (Try this on a very steep roof)

SUGGESTION

If possible, turn off physics processing at very low levels of momentum. A prone clone very rarely has any momentum, and so just the head slump and slight "body curl" would be enough to properly simulate death. Yes, this may spoil the "flops on to back" physics that can occur (e.g. on a roof apex) but - cool as it looks - it is better to lose that than have a 30-second jittering corpse (I think)

A Cool Death

Shoot a clone on the church stairs ... depending on the amount of INITIAL MOMENTUM you may well get a fantastic effect ... Or you will get a clone with the jitters. Its initial state (standing/prone) will almost certainly affect the coolness-factor.

Cheers

Edited January 19, 2010 by branko