Author Topic: Turned Poison into Disease through AI, but...  (Read 15012 times)

Armorvil

  • *
  • Posts: 621
  • Working on : FFVII Total Grudge
    • View Profile
Re: Turned Poison into Disease through AI, but...
« Reply #25 on: 2011-05-23 21:39:54 »
Quote
This works for Battle Texts, but I don't know if it works inside scripts like that. The way you reference Character names (Barret, Cloud) inside the AI text is different.

Ah yes. Cloud is {EA0000h} in the AI text, and Barret is {EA0001h}, etc... Actually, it looks a lot like the FF7-text format, but all my different tests didn't net me any results regarding monster names. If this script becomes manageable in the future, I guess a simple "OVERKILL" message will do...

Quote
Multiply by 3, divide by 2. Boom, 1.5

lol, basic maths. I should have realized one can do more than one operation ^^;

Quote
I believe there's a Var:4XXX battle address for this. I don't recall what it is though as my notes for that section are incomplete I know that there are a lot of values in there that aren't used in AI, but the battle engine itself might handle.

Ah, this gives me hope - though I don't hold my breath.

Quote
Hmm. I don't know if you can find that info out. You can find out the amount of HP lost since last turn, but that's not the same thing. I'll have to investigate...

If those investigations will require you to in-game test each one of the 4xxx battles addresses not referenced into wallmarket.dat, I wish you good luck. Looking at them, there are 4041-4047, 4049, 4051-4057, 4059, 4061-4067, 4069, 4071-4077, 4079, 4081-4087, 4089 (I start to see a pattern :P), 4091-4097, ... (not gonna list them all, actually O_O). We're talking about hundreds - maybe thousands of battle addresses ! I may be crazy like that, but even I couldn't test them all - and I suggest you don't either, if you value your sanity. Still, perhaps many of those can already be ruled out ?

And about Oil : wouldn't it be possible to increase the level and/or the magic stat of the caster, instead ?... ...After all, levels can go up to 255 - might as well put this to good use.

And it's too bad about elemental weapons :( I asked in another thread if the AI could check for the presence of a piece of equipment, but I guess this means it can't.

Quote
The only other alternative is to find a way of setting a weakFire flag via AI. I can't really help at the moment because my machine is out of commission, but to find the weakFire flag, you would need to

- take Akari's memory map of the NTSC PSX build's battle engine
- run an emulator and compare memory maps for monsters with weakFire and without
- use a process of elimination to see where the weakFire might be stored
- look at the addresses in the memory map the AI can, uh, address, and see if some of the unknown addresses tally with the location of your weakFire variable
- set weakFire on when the target was Blind

Yeah, but weakFire is just a basic x2 damage. So, oiling an already fire-weak monster wouldn't reap any more benefits.

EDIT:

About Overkill, how about this :

In General Counter :

Code: [Select]
If HP < MaxHP
If target:self is NOT dead
Set {VAR:E0} to 1

In Death Counter :

Code: [Select]
If {VAR:E0} is 0
If attack ID is NOT 18 (=Death)
If attack ID is NOT 3D (=Odin)
Display message: "OVERKILLED!!"
Set Exp to 2x
Set Gil to 2x

What do you think ? Alternatively, I thought about using the in-game timer, and if you finished a battle quickly enough, you'd get added bonuses. FFVI can check a timer in its AI (some enemies can call backup, after a while), but I'm not sure FFVII can. And, third idea : couldn't we simply make use of the "Greatest elemental damage" battle address ? Since almost all attacks are elemental thanks to the physical elements, if one put, in the Death Counter script :

Code: [Select]
- Check for self + max HP
- Put this value in memory
- Check if the greatest elemental damage taken is equal or greater than the value we just stored in memory
- If not, jump to the end of the script
- If true, display message: "OVERKILL!"
- Check for AP
- Check for XP
- Check for Gil
- multiply those by 3
- divide them by 2
...

Wouldn't it work ?
« Last Edit: 2011-05-23 23:37:16 by Armorvil »

Jenova's Witness

  • Right Wing Safety Squads
  • *
  • Posts: 471
  • I ♥ SCIENCE
    • View Profile
.
« Reply #26 on: 2011-05-23 22:44:40 »
.
« Last Edit: 2015-11-16 10:51:15 by Jenova's Witness »

NFITC1

  • *
  • Posts: 2944
  • I just don't know what went wrong.
    • View Profile
    • WM/PrC Blog
Re: Turned Poison into Disease through AI, but...
« Reply #27 on: 2011-05-24 04:59:06 »
Ah, this gives me hope - though I don't hold my breath.

Good idea. There's likely nothing like that, but you never know. :)

If those investigations will require you to in-game test each one of the 4xxx battles addresses not referenced into wallmarket.dat, I wish you good luck. Looking at them, there are 4041-4047, 4049, 4051-4057, 4059, 4061-4067, 4069, 4071-4077, 4079, 4081-4087, 4089 (I start to see a pattern :P), 4091-4097, ... (not gonna list them all, actually O_O). We're talking about hundreds - maybe thousands of battle addresses ! I may be crazy like that, but even I couldn't test them all - and I suggest you don't either, if you value your sanity. Still, perhaps many of those can already be ruled out ?

No, that's not how they work. Think of those addresses as pointers to BITS, not bytes. Every 8 is a different byte, and we're talking in Hexadecimal. So between 4008 and 4010 there are 8 values (4009, 400A, 400B, etc). My notes say that the highest 4XXX value is 42E0 which contains the total experience gained in that battle (technically, it goes to as far as 433F before it leaks into another actor's data, but it doesn't look like 4300 and beyond are used). Almost every value in between is used for something.

Bosola

  • Fire hazard!
  • *
  • Posts: 1745
    • View Profile
    • My YouTube Channel
Re: Turned Poison into Disease through AI, but...
« Reply #28 on: 2011-05-24 08:41:24 »
Can these addresses leak into another actor's data? Which data exactly (it might well be data I want to manipulate)?

I need to get to know that memory map better. I'd like to see how data for different actors is positioned.

I wonder which function assigns pointers to 'relative' addresses like 0x4001 etc. I also wonder if we might hijack it.

Armorvil

  • *
  • Posts: 621
  • Working on : FFVII Total Grudge
    • View Profile
Re: Turned Poison into Disease through AI, but...
« Reply #29 on: 2011-05-24 09:01:12 »
Quote
No, that's not how they work. Think of those addresses as pointers to BITS, not bytes. Every 8 is a different byte, and we're talking in Hexadecimal. So between 4008 and 4010 there are 8 values (4009, 400A, 400B, etc).

Why do you feel the need to tell me that, because I know this and- */me re-reads his previous post* ...OK, wow, I took the addresses as if they were decimals, I must have been really tired  ;D

So, what do you guys think about the three ideas I posted when I edited my previous post ? I'm especially interested in the GreatestElementalDamage battle address (VAR:4058).
« Last Edit: 2011-05-24 09:05:47 by Armorvil »

NFITC1

  • *
  • Posts: 2944
  • I just don't know what went wrong.
    • View Profile
    • WM/PrC Blog
Re: Turned Poison into Disease through AI, but...
« Reply #30 on: 2011-05-24 12:12:16 »
As far as I can tell, 4058 is only modified when a 96 command is executed. I believe it's blank otherwise. You'll have to read my notes in WM's or PrC's readme for more details.

Armorvil

  • *
  • Posts: 621
  • Working on : FFVII Total Grudge
    • View Profile
Re: Turned Poison into Disease through AI, but...
« Reply #31 on: 2011-05-24 18:25:17 »
Thank you. Will do.

NFITC1

  • *
  • Posts: 2944
  • I just don't know what went wrong.
    • View Profile
    • WM/PrC Blog
Re: Turned Poison into Disease through AI, but...
« Reply #32 on: 2011-05-25 16:23:34 »
Can these addresses leak into another actor's data? Which data exactly (it might well be data I want to manipulate)?

Addresses can leak into other actors' data I don't see any restrictions in the AI code to cap the address it will access. Here are some of the findings that I've made. Each 4XXX structure is 104 bytes in memory. Normally, these will go up to 0x433F then another actor's data will start. Actor 0 (I believe this is the top slot of player characters) would be the only one to access everyone else's data. Starting at 0x4340, Actor (X) would be able to access Actor (X+1)'s active statuses. Adding another 340h to that and Actor (X) can access Actor (X+2)'s data, etc. Technically, I believe there are 10 actors (three playable characters, up to six enemies, and one "battle actor" that would contain formation-based scripts). For Actor (0) to access Actor (9)'s data, it would start at 0x5D40 and should be no greater than 0x607F. This applies to ANY 4XXX Battle variable. This doesn't apply to 2XXX addresses, because they are technically temporary to which ever character's turn it is.

So it IS possible for actor 9 to perform some DRASTIC memory leaks. I don't really know what's beyond the battle addresses, but Actor 9 can access quite a bit of it. Just shy of 6K outside of it as a matter of fact.

I need to get to know that memory map better. I'd like to see how data for different actors is positioned.

So follow this if you want to access other actors' data. Just add these amounts to the address of what you want to access:

Code: [Select]
Actor (X)   Actor (X+1)   Actor (X+2)   Actor (X+3)   Actor (X+4)   Actor (X+5)   Actor (X+6)   Actor (X+7)   Actor (X+8)   Actor (X+9)
        0           340           680           9C0           D00          1040          1380          16C0          1A00          1D40

I'm not certain about which actor is which, but I think it goes: PartyCharTop, PartyCharMid, PartyCharBottom, Battle, Enemy0, Enemy1, Enemy2, Enemy3, Enemy4, Enemy5

I wonder which function assigns pointers to 'relative' addresses like 0x4001 etc. I also wonder if we might hijack it.

There aren't really any pointers. It's all pretty hard-coded static address lookups in every location it gets read from/written to. You can't really hijack it because they're all referenced all over the place. I suppose you could hijack lookups to those addresses though, but that's a lot of hooks.

Bosola

  • Fire hazard!
  • *
  • Posts: 1745
    • View Profile
    • My YouTube Channel
Re: Turned Poison into Disease through AI, but...
« Reply #33 on: 2011-07-02 00:39:50 »
I've been looking into this tonight. As far as I can tell, I just can't get a Counter:PreTurn script "Make Self.NullRestore = 1" or "Make Party.NullRestore = 1" to stop healing. I'm not holding any counter materia, so I'm developing a suspicion that the element control flags don't work outside the initialization scripts.

Will continue researching this.

Bosola

  • Fire hazard!
  • *
  • Posts: 1745
    • View Profile
    • My YouTube Channel
Re: Turned Poison into Disease through AI, but...
« Reply #34 on: 2012-01-01 19:49:50 »
*bump*

I finally got around to trying this out - I've uploaded a Youtube video here: http://youtu.be/hxhSvh5IHJI

In this case, I created a pre-attack script on one character:

Code: [Select]
12 2060
10 42A9
80
02 2060
00 4003
80
90
73

Then I just linked my characters to the other's scripts and was on my way.

For some reason, trying to make the script set allActors.nullRestore = allActors.poison (using the address 20E0 rather than 20E0 in the script above) didn't work well - inflicting / curing poison on one character would apply / remove nullRestore on everyone. I'm not sure why this is - though I'd appreciate some pointers.

In any case, I just made my script set self.nullRestore = self.poison, and it worked fine for the player characters. Any ideas how I can make it work OK on enemies without having to give them all a similar script? Maybe if I take an activeMask and iterate through all the formation numbers?