Jump to content
MaxCheaters Official Group on Discord
Top.MaxCheaters.com is now OPEN – Lineage 2 Top Listing

Lineage 2 GF AI Decompiler

verbrannt
 Share

Recommended Posts

verbrannt Newbie

verbrannt

Posted
  • Author
Posted
23 minutes ago, eressea said:

Sorry for doublepost, one more question, I keep getting av_quest0 instead of myself.av_quest0, any idea what can be wrong?


Was bug in NASC generator. Fix in master.

 

46 minutes ago, eressea said:

I'm using tree structure with UTF-8 files and when I build the AI, I go through the directory structure (so I know which is the correct class order) and join all the files into one big UTF-16LE file, then I let it compile and then I split it again to directory structure

 

Sounds convenient. I'll make it default behaviour.

verbrannt Newbie

verbrannt

Posted
  • Author
Posted (edited)
14 minutes ago, eressea said:

Thanks a lot :)

 

I've found another problem, in some cases, I get results like this: 


if (5 == 100 || Rand(100) < 5) {
	if (35003 == 35001) {
		select (11) {
...

Class + data: http://download.l2shrine.com/decompiler-problem1.zip

 

 

Seems like code generated correctly, but looks suspicious. 

	...
	push_const 5			//unary->INTEGER_CONST
	push_const 100			//unary->INTEGER_CONST
	equal
	...
	push_const 35002			//unary->INTEGER_CONST
	push_const 35001			//unary->INTEGER_CONST
	equal
	...

Stack machine in my head generates same code as decompiler :)

 

BTW: Can I push your h5 data folder to my repository? Also the '@' prefixes in enums.json now not needed (because of generators to another languages), & type P_STATE renamed to PSTATE.

Edited by verbrannt
eressea Newbie

eressea

Posted
Posted

Oh, I didn't notice that, well, that's really strange... Also my compiler complains somewhere there: "Stack underflow may be occurred by unary constant" - maybe the select (11)?

verbrannt Newbie

verbrannt

Posted
  • Author
Posted
2 minutes ago, eressea said:

Oh, I didn't notice that, well, that's really strange... Also my compiler complains somewhere there: "Stack underflow may be occurred by unary constant" - maybe the select (11)?

Yeah, this warning because of constants in statements like select/if/while.
I think NCSoft uses some sort of preprocessor for AI scripts, that generates such strange code.

eressea Newbie

eressea

Posted
Posted
12 minutes ago, verbrannt said:

Yeah, this warning because of constants in statements like select/if/while.
I think NCSoft uses some sort of preprocessor for AI scripts, that generates such strange code.

 

For me it seems the ai.obj I'm trying to decompile has already been decompiled (but wrong) and recompiled... So I'll probably have to fix these manually...

 

Great job, the decompiler works nice :)

eressea Newbie

eressea

Posted
Posted (edited)

Updated h5 definitions http://download.l2shrine.com/h5.zip

 

EDIT: Found one more new problem, if I have 'star_buster_rsk.focus' in precompiled headers, it generates @star_buster_rsk.focus which fails to compile (needs to be quoted like this @"star_buster_rsk.focus")

Edited by eressea
verbrannt Newbie

verbrannt

Posted
  • Author
Posted (edited)
50 minutes ago, eressea said:

Updated h5 definitions http://download.l2shrine.com/h5.zip

 

EDIT: Found one more new problem, if I have 'star_buster_rsk.focus' in precompiled headers, it generates @star_buster_rsk.focus which fails to compile (needs to be quoted like this @"star_buster_rsk.focus")

Thx for h5 defs!

As for precompiled headers issue, I didn't knew such syntax is possible (dots in these constants). I've added simple check for dot, and if dot exists, constant will be quoted. Checkout the master branch.

 

Quote

For me it seems the ai.obj I'm trying to decompile has already been decompiled (but wrong) and recompiled... So I'll probably have to fix these manually...

 

Idk. Found similar code in the leaked GF AI. Class ssq_npc_priest for example:
  

if (2 == 2) {
    ShowSystemMessage(talker, 1273);
} else if (2 == 1) {
    ShowSystemMessage(talker, 1274);
}

select (1) {
case 1:
    ShowSystemMessage(talker, 1275);
    break;
case 2:
    ShowSystemMessage(talker, 1276);
    break;
case 3:
    ShowSystemMessage(talker, 1277);
    break;
}

 

Edited by verbrannt
eressea Newbie

eressea

Posted
Posted

Great, now it works fine :)

 

Another problem, I get this: 

Say("판정식 : " + myself.i_ai6 + " - " + FloatToInt(myself.sm.hp) + " = " + myself.i_ai6 - FloatToInt(myself.sm.hp));

which is wrong (won't compile) because there's no minus operation for strings. It should be parenthesised: 

Say("판정식 : " + myself.i_ai6 + " - " + FloatToInt(myself.sm.hp) + " = " + (myself.i_ai6 - FloatToInt(myself.sm.hp)));

It's a minor bug (it's just once in whole ai) but it's bit annoying to have to fix it manually ;)

verbrannt Newbie

verbrannt

Posted
  • Author
Posted
25 minutes ago, eressea said:

Great, now it works fine :)

 

Another problem, I get this: 


Say("판정식 : " + myself.i_ai6 + " - " + FloatToInt(myself.sm.hp) + " = " + myself.i_ai6 - FloatToInt(myself.sm.hp));

which is wrong (won't compile) because there's no minus operation for strings. It should be parenthesised: 


Say("판정식 : " + myself.i_ai6 + " - " + FloatToInt(myself.sm.hp) + " = " + (myself.i_ai6 - FloatToInt(myself.sm.hp)));

It's a minor bug (it's just once in whole ai) but it's bit annoying to have to fix it manually ;)

 

Fixed.

eressea Newbie

eressea

Posted
Posted (edited)

Thanks :)

 

Found another problem that needs to be handled in order to work with H5+ AI correctly. NCsoft apparently moved FString stuff to client side, so now there's 

		{1120131; -111092; 232173; -3448; 0; 0 }

instead of 

		{"Discarded Guardian (lv20)"; -111092; 232173; -3448; 0; 0 }

in the ai.obj. It should decompile to 

		{1120131; -111092; 232173; -3448; 0; 0};

and not to 

		{"1120131"; -111092; 232173; -3448; 0; 0};

when NASCVersion is 60+

 

EDIT: Already fixed in my h5 compiler branch

 

EDIT2: Also I've removed fstring lookup on 3 different places so it probably won't be just TelPosList but maybe another 1-2 things...

 

Edited by eressea
verbrannt Newbie

verbrannt

Posted
  • Author
Posted
21 minutes ago, eressea said:

Thanks :)

 

Found another problem that needs to be handled in order to work with H5+ AI correctly. NCsoft apparently moved FString stuff to client side, so now there's 


		{1120131; -111092; 232173; -3448; 0; 0 }

instead of 


		{"Discarded Guardian (lv20)"; -111092; 232173; -3448; 0; 0 }

in the ai.obj. It should decompile to 


		{1120131; -111092; 232173; -3448; 0; 0};

and not to 


		{"1120131"; -111092; 232173; -3448; 0; 0};

when NASCVersion is 60+

 

EDIT: Already fixed in my h5 compiler branch

 

EDIT2: Also I've removed fstring lookup on 3 different places so it probably won't be just TelPosList but maybe another 1-2 things...

 

 

Can you make a PR or patch?

eressea Newbie

eressea

Posted
Posted
Just now, verbrannt said:

 

Can you make a PR or patch?

 

I'm not sure how propagate NASC version to parser.php, if it was there, the fix would be easy:

  

        if (is_numeric($row[0])) {
            $row[0] = '"' . $this->data->getString($row[0]) . '"';
        }

should be called only if $nascVersion >= 60

eressea Newbie

eressea

Posted
Posted

Also, when speaking about NASC versions, the right data directory could be chosen automatically according to the version:
gf = 2, freya = 56, h5 = 60, gd = 73

verbrannt Newbie

verbrannt

Posted
  • Author
Posted
5 minutes ago, eressea said:

 

I'm not sure how propagate NASC version to parser.php, if it was there, the fix would be easy:

  


        if (is_numeric($row[0])) {
            $row[0] = '"' . $this->data->getString($row[0]) . '"';
        }

should be called only if $nascVersion >= 60

 

So FString lookup should be removed from all places, no need to check NASC version. Anyway getString method not working now.

 

3 minutes ago, eressea said:

Also, when speaking about NASC versions, the right data directory could be chosen automatically according to the version:
gf = 2, freya = 56, h5 = 60, gd = 73

 

Yeah, I thought about it.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share
Go to topic listing


  • Posts

    • l2viserion
      L2VISERION CHAPTER IV - 16.05.2026

      By l2viserion · Posted

      And Discord: https://discord.gg/3aYqWNqb
    • l2viserion
      [L2] VISERION

      By l2viserion · Posted

      Ofc: https://discord.gg/3aYqWNqb
    • NoJobRob
      Skins for H5

      By NoJobRob · Posted

      You can find some H5 skins shared in old L2 modding Discords, but most of the higher‑quality ones are either paid or come bundled with full client edits. I usually mix in commissioned work and whatever I can patch myself. On a side note, I fund a lot of these commissions by selling off game items through instant sell cs2 skins, which has been a quick way for me to get some cash for projects.
    • Elfo
      Outerlude - A Modern Interlude aCis fork

      By Elfo · Posted

      There is no need for gRPC in this case, even tho originally it was gRPC based but since we don't need it to be bi-directional, we switched to simple http requests for the web calls and SSEs for the data streamed from the server. There are distributed locks in place to precent race conditions between actions that can happen between multiple web instances and the server.   Local models can also be slow depending on the model, and most external models can actually be faster than local ones if you use Flash 2.5 or something along those lines. I am running on 512GB of Unified Memory on my Mac Studio M3 Ultra so the speed of the local model for a small model is pretty good but I tested it with Gemini too and it works equally as fast and in some cases faster. The way it works is that I'm using pgvector (one of the benefits of moving to Postgres) to search the data and see what the player can see etc and there is some batching of the next few actions for 2-4 seconds for the user until the next LLM request fires. The batching also includes branching on logic so if they for example fall under some HP they will move to kiting instead of attacking or maybe they heal etc.   Everything is authed and permission-based. The server and the backend of the frontend have secure communication between them, either with a symmetric key (not recommended for production) or a certificate (the recommended way), so there is no worry. It's all tied to the account's access level, etc., so nobody can make an action that they normally wouldn't be allowed to do. Even the MCP is token-based, and there are prompt injection protections in place. The MCP is audited, and every mutation needs confirmation. The admin area is only accessible to the admin account anyway so normal users can't access it.  
    • Frank
      Outerlude - A Modern Interlude aCis fork

      By Frank · Posted

      Amazing job! Gratz!

  • Topics

×
×
  • Create New...

Important Information

This community uses essential cookies to function properly. Non-essential cookies and third-party services are used only with your consent. Read our Privacy Policy and We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue..