Failed KO animation

[SneaksieDave]

Have considered doing that, but I was a bit wary to do a full mod-wide rollback (tons of changes) or try just doing individual files (error prone, plus I don't know which files specifically would be needed). Especially trusting the whole thing to an already-established-as-buggy SVN tool.

 

Maybe.

[Tels]

Have considered doing that, but I was a bit wary to do a full mod-wide rollback (tons of changes) or try just doing individual files (error prone, plus I don't know which files specifically would be needed). Especially trusting the whole thing to an already-established-as-buggy SVN tool.

 

Maybe.

 

If you have enough diskspace (and time you can make a full copy of your darkmod folder, then do a rollback to a week ago, then go onwards maybe daily or so until you narrowed it down. Takes ages, tho...

[Springheel]

I'm not convinced that anything has actually changed, for me. I've KO'd people in the combat map before, a few weeks ago, and I'm not feeling that anything is noticeably different.

[SneaksieDave]

Well, I could be imagining it if while for weeks as I worked on the stealth area, going through an increasing number of individual stealth sections over and over while trying to tweak the gameplay, not to mention any other separate testing (I use my own local temp7.map frequently which always includes taking out one guard facing away), never having any problem KO'ing anyone, in fact it seemed too easy as I swung from what felt like 6 feet away and still took them down, without a single failure, every single time...

 

...and then suddenly, some day in the last week, I became much less skilled at it. Night and day.

 

Not sure how else to put it. To me it seems far more likely something changed with all of the recent changes to attached items and rigging/anims.

 

Anyway, I digress unless/until someone mentions a likely cause (see other thread). For the record, I think the current method is too difficult and people will complain again as they did in the past.

[Springheel]

I wasn't claiming that you were mistaken. I'm just saying I haven't noticed the same thing on my system, which might suggest that some local file is at work (either on your system or mine).

[SneaksieDave]

I know, nothing personal taken. I did wonder if it was something local here or on the end of someone developing the AI (yourself or Ishtvan or maybe even the animators?), but I haven't reverted anything recently, so if it was a local modified file, it should remain as such. I dunno, will think on it more, and we'll see I guess.

 

I do think it's too hard though.

[Mortem Desino]

What I've done before instead of a totally clean checkout is

 

1. Delete select file(s) that are the possible culprit.

2. When SVN updating, the original unmodified file will be restored to your darkmod.

 

If that doesn't work

1. Delete select directory(ies) containing possible culprit(s).

2. SVN update will redownload the directory(ies) and contents from scratch.

 

Best of luck to you. I'll also be looking into this, since I was playing Outpost and simply couldn't KO the stupid priest (among other AI).

[Noisycricket]

I replaced the short failed_KO.m5danim with a longer one, as well as added the .ma file to model_src.

[Noisycricket]

This is in code now, a new PK4 will be uploaded soon. You can test it on the builder guards in test/walk (you need to spawn an atdm:weapon_blackjack beforehand, of course).

 

I get unknown command when i try to create the blackjack, is there a space i should be entering?

[greebo]

This is the command:

 

spawn atdm:weapon_blackjack

[Noisycricket]

Thank you

[Noisycricket]

When i go to blackjack an AI, i get an error "can't find function legs_pain in ai_darkmod_base. I don't need to see it though, I'll see it eventually.

 

Squill, are you going to re-create your failed_ko?

[squill]

Squill, are you going to re-create your failed_ko?

 

i should be able to finish that rough version today.

[Ishtvan]

I was on travel for a few days, just got back. Those KO cones Dave posted are misplaced. They should be at the center of the head, they look vertically low. If anything about the head model or head attachment joint changed, that might do it. I believe the head center offset is a spawnarg relative to the head model origin, or the head joint if it doens't have a separate head model. NOt sure exactly what the spawnarg is though, must sleep for now.

[greebo]

When i go to blackjack an AI, i get an error "can't find function legs_pain in ai_darkmod_base. I don't need to see it though, I'll see it eventually.

Could be my fault - I probably forgot to commit the script update on ai_darkmod_base.script.

[Springheel]

If anything about the head model or head attachment joint changed, that might do it.

 

The head attachment point is much lower than it used to be--it's "spine2" now, rather than "head". The origin of the head entity is thus much lower on the new heads.

[SneaksieDave]

Interesting. I'll try KO'ing lower, I guess, and see if that does it.

[Ishtvan]

Okay, we'll have to modify the offset spawnarg to be higher then, to put the point back in the center of their head where it's supposed to be. One issue is that right now it's a spawnarg on the AI, rather than the head (I think).

 

Ideally we would re-code it to read the offset from the head entity instead of the AI. However, I think we can still get away with the current setup, if AI are only going to have one type of head attached to them, old vs. new? That would make sense if the attachment joint is either going to be the old joint or the new joint. Then we can just put different head center offsets in ai_humanoid vs ai_humanoid_newskel.

[Springheel]

I think we can still get away with the current setup, if AI are only going to have one type of head attached to them, old vs. new? That would make sense if the attachment joint is either going to be the old joint or the new joint.

 

Yes, that's right.

 

Then we can just put different head center offsets in ai_humanoid vs ai_humanoid_newskel.

 

That seems like the best way to do it...that way old skeleton characters like the builder priest can still be KO'd effectively.

[SneaksieDave]

1. Not sure if anything has changed on this yet, but I just played through stealth (notarget) and failed to KO almost every AI on the first (sometimes second or third) try. Only after I saved and reloaded, mentally adjusting my strike point (to what felt like a shoulder strike) did I start improving my ratio (slightly). So perhaps the offset guess above is correct. No doubt in mind of the change.

 

2. Related, I also tried striking them in the face for KOs too, but that wasn't working at all (I think I had a fluke once or twice out of dozens). Aren't we allowing for fully un-alert blow to the face KOs (as long as no helm/face guard)?

[greebo]

I changed the offset so that it is located a bit more upwards for the newskel AI. Is this about right?

 

 

However I don't seem to get a green cone drawn on the screen - is this normal?

[Ishtvan]

The green cone is representing the vertical angle, the red horizontal (or vice versa, I forget). If you only see one color, it's probably because the horizontal and vertical angles are equal. Although, when I last set it, I didn't think they were equal. Maybe I'm thinking of the fov though, maybe the ko was equal...

[SneaksieDave]

So we are or aren't supporting a zero alert face blow (blind alley sneak)? I thought we were, without helmet or alert. I'm not sure how it works though, if the cone is the only KO zone at all, or just in terms of alertness.

[Ishtvan]

You can do that if you want, it's all up to what spawnargs you set. I forget what we left the default at right now. It probably needs more attention before release. I don't think combattants can be KO'd to the face at zero alert. You could test this though by setting yourself notarget and trying to hit a Builder in the face.

[Springheel]

I think the default was that guards without helmets can be KO'd to the face when nonalert, but then only from behind when alert.

 

However, I don't think our KO behaviour is accurate right now, as per: http://forums.thedarkmod.com/index.php?showt...mp;#entry190392