NTai XE10.1b - Page 86

NTai XE10.1b

Here is where ideas can be collected for the skirmish AI in development

Moderators: hoijui, Moderators

imbaczek
Posts: 3629
Joined: 22 Aug 2006, 16:19

Re: NTai XE9.8 experimental for 0.76b1

Post by imbaczek »

AF: git gui 'just works' after the initial setup which consists of about 5-6 shell commands which you copy&paste from a tutorial. don't spread FUD, please. i taught my girlfriend how to use it for backing up a bigger piece of writing.
User avatar
hoijui
Former Engine Dev
Posts: 4344
Joined: 22 Sep 2007, 09:51

Re: NTai XE9.8 experimental for 0.76b1

Post by hoijui »

welll.. to me.. git gui appears a bit buggy under linux even. for example, i select 3 files to stage, press Ctrl+T, and it stages only 2 of the 3, and subsequent Ctrl+T presses do nothing. only if i select an other file, select that file again and then Ctrl-T, it gets staged.
and there are other things like this. though.. for me it is ok so far, still prefer it over SVN (though i had a nicer GUI there).
User avatar
DJ
Posts: 355
Joined: 17 Jan 2007, 13:26

Re: NTai XE9.8 experimental for 0.76b1

Post by DJ »

There's definitely a problem reading the attackers tag (and i assume therefore other tags). tested using BA

Here is the test:

build a config from scratch to ensure no errors, simplest config possible to generate an attacker. basically:-

1) commander builds a few resources and a vehicle plant
2) vehicle told to build b_assault
3) only attacker set in config is armflash

the vehicle lab will build a different attacking unit as NTai resorts to using it's auto detect attackers algorithm (which only seems to detect ranged weapons or ones that stock pile weapons).

however one positive is I ran it for a long time and it didn't crash! Good times...

beginning of log is shown below:-

Code: Select all

[-]|15:50:59| < Frame: 0 >AI interface says this mod has this many units! :: 368
[-]|15:50:59| < Frame: 0 >Unitdef loader constructed
[-]|15:50:59| < Frame: 0 >Order Router constructed
[-]|15:50:59| < Frame: 0 >DTHandler constructed
[-]|15:50:59| < Frame: 0 >RadarHandler constructed
[-]|15:50:59| < Frame: 0 >Planning constructed
[-]|15:50:59| < Frame: 0 >Economy constructed
[-]|15:50:59| < Frame: 0 >Manufacturer constructed
[-]|15:50:59| < Frame: 0 >BuildingPlacer constructed
[-]|15:50:59| < Frame: 0 >Chaser constructed
[-]|15:50:59| < Frame: 0 >Initialisising
[-]|15:50:59| < Frame: 0 >error in buffer
[-]|15:50:59| < Frame: 0 >Mod TDF loaded
[-]|15:50:59| < Frame: 0 >Junk in buffer :;
[-]|15:50:59| < Frame: 0 >error in buffer
[-]|15:50:59| < Frame: 0 >Mod TDF loaded
[-]|15:50:59| < Frame: 0 >values filled
[-]|15:50:59| < Frame: 0 >Arrays filled

[-]|15:50:59| < Frame: 0 >Loading unit data
[-]|15:50:59| < Frame: 0 >error in buffer
[-]|15:50:59| < Frame: 0 >Unit data loaded
 :: Altored_Divide.smf
 :: BA641.sd7
 :: map size 16 x 16
[-]|15:50:59| < Frame: 0 >Planner Init'd
[-]|15:50:59| < Frame: 0 >Manufacturer Init'd
[-]|15:50:59| < Frame: 0 >BuildingPlacement Init'd
[-]|15:50:59| < Frame: 0 >Chaser Init'd
[-]|15:51:0| < Frame: 0 >loading contents of  tasklist :: com :: filling tasklist with #6 items
[-]|15:51:0| < Frame: 0 >CUnitConstructionTask::CUnitConstructionTask object created | params: building :: armvp using builder::armcom
[-]|15:51:0| < Frame: 0 >loaded contents of  tasklist :: com :: loaded tasklist at 6 items
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

I didn't say there were no clients that 'just work' I said I don't know of any, two totally different things. I'll take a look at git-gui.

Ill look into attackers too, it might well be indicative of an issue with config reading since that code has been changed a few times recently.

I may also add a few more console commands to make it easier to tell thigns like wether its using an algorithm or the config rather than speculation through observation.

Of course a debugger would be far better at all this as you could see directly in memory the value of the variable controlling it
User avatar
Gota
Posts: 7151
Joined: 11 Jan 2008, 16:55

Re: NTai XE9.8 experimental for 0.76b1

Post by Gota »

Does NTai work with the latest spring version?
User avatar
1v0ry_k1ng
Posts: 4656
Joined: 10 Mar 2006, 10:24

Re: NTai XE9.8 experimental for 0.76b1

Post by 1v0ry_k1ng »

if it does, here is the full xta config

http://spring.jobjol.nl/show_file.php?id=1710

hf
User avatar
1v0ry_k1ng
Posts: 4656
Joined: 10 Mar 2006, 10:24

Re: NTai XE9.8 experimental for 0.76b1

Post by 1v0ry_k1ng »

got a newest working version and toolkit link? Im feelin' that configging mood
User avatar
Gota
Posts: 7151
Joined: 11 Jan 2008, 16:55

Re: NTai XE9.8 experimental for 0.76b1

Post by Gota »

Me 2.
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

No modifications to NTai were made since the last 'AI Pack' that I'm aware of. Same with toolkit.
User avatar
Gota
Posts: 7151
Joined: 11 Jan 2008, 16:55

Re: NTai XE9.8 experimental for 0.76b1

Post by Gota »

But do all the features work?
i don't wanna spend hours configuration it just to find out it doesn't work well with current spring.
The only reason im questioning this is cause of the many spring releases lately and cause of reading some posts where people say AIs are somehow fooked with current spring.
User avatar
1v0ry_k1ng
Posts: 4656
Joined: 10 Mar 2006, 10:24

Re: NTai XE9.8 experimental for 0.76b1

Post by 1v0ry_k1ng »

link to latest of each?
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

With regards to config issues, in the config file there is a tag which controls the splash message, this should instantly tell you if its loaded a config or not because it'll show your custom version.

I would also try adding another unit to a list and avoiding having lists that have only a single item. Not sure what the reason for that is or if it was a bug I fixed ages ago but it sounds about right anyway.

On top of that, a debugger would tell you the answer you seek. The algorithm is controlled by a simple boolean variable. I would suggest you get a working debugger regardless of which AI or program you use. Even VS2008 express would suffice, especially now we're in a C API.
User avatar
DJ
Posts: 355
Joined: 17 Jan 2007, 13:26

Re: NTai XE9.8 experimental for 0.76b1

Post by DJ »

I tested it, its broken. I gave you a pretty damned accurate bug report and you haven't fixed the problem so why bring this up in a thread about another AI?

If you aren't maintaining this anymore then people will understand, you have real life to contend with and many other projects. What I don't like is you dropping the name of your AI in threads about other AI's when clearly this AI is no longer supported. You will just cause confusion for new comers who want an AI that works.

Just to clarify for any people new to spring, I've been following this AI for 2 years now and there has never been a working version released in this time (As Ivory King, zwzsg or anyone else who has tried to build a config will verify).
User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Re: NTai XE9.8 experimental for 0.76b1

Post by Argh »

Just to clarify for any people new to spring, I've been following this AI for 2 years now and there has never been a working version released in this time (As Ivory King, zwzsg or anyone else who has tried to build a config will verify).
It's worked, off and on, but never has worked well enough for me to feel like spending more time on a deployment, and as of the last time that I paid attention, it had a crippling CPU bug that reared its head once you got past early game, which made it totally worthless.

Meh, AF, you should either fix it or just declare that it's dead and that your work's keeping you too busy these days. I wanted to use this for P.U.R.E., since it's about the only one that might be able to do anything right, beyond "playing" at a really simple level. As things stand, RAI is the only AI that can play P.U.R.E. ATM without crashing, despite much work to try and keep things compatible.
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

I haven't got around to doing another round of meddling with it, but last time I tried it worked, and people in other threads said it worked, and I mentioned it in that thread because what the guy wanted to do with AAI was the exact concept that drove NTai development for nearly 2 years.
User avatar
1v0ry_k1ng
Posts: 4656
Joined: 10 Mar 2006, 10:24

Re: NTai XE9.8 experimental for 0.76b1

Post by 1v0ry_k1ng »

Let me paraphrase: shit is dead, goto KAI
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

hardly, my development has shifted from low level continuous work to big bursts every few months now.
User avatar
1v0ry_k1ng
Posts: 4656
Joined: 10 Mar 2006, 10:24

Re: NTai XE9.8 experimental for 0.76b1

Post by 1v0ry_k1ng »

Just to clarify for any people new to spring, I've been following this AI for 2 years now and there has never been a working version released in this time
shit is deeead
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

Far from it, I have been diggign and thrown up some itneresting things.

Firstly IK, your configs would never load no matter what I did to NTai because somebody commented out the tags NTai was using in XTA 9.44+ so NTai could never find the config anyway.

Ontop of that my TDF Parser class is showing odd behaviour in that it thinks theres only 1 section in the modinfo.tdf file and its the entire file on a single line with no line breaks 0_o

I corrected an error with mismatched forward and back slashes and the new data folder, and made a refactor that changed the way units are stored to not involve those stupid boost:shared_ptr objects I spammed a year ago. This should probably give a large speed boost at the cost of building an array of 10,000 CUnit* pointers. Since most will be null pointers, the speed up gained is worth the sacrifice, nevermind the simplification of about 60 if statements. Instead of passing around boost::shared_ptr<CUnit> I shall now pass around int unit_id!

Anyways, after a weekend fo fiddling I'm now able to build and debug NTai and test it in visual studio 2008 using a buildbot install of spring.
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: NTai XE9.8 experimental for 0.76b1

Post by AF »

nm my parser works, its jsut beign handed the map modinfo.tdf not the mod archives version.....
Post Reply

Return to “AI”