Spring RTS Engine

For Testing Purposes Only

Test Build 4!

This is a development build of NTAI and is not meant to be the next version of NTAI as it is incomplete.

Most of the bugs in the previous versions of NTAI where core problems which ended up being rewritten and so on. To eradicate all these bugs and come up with an AI base that is much more expandable and flexible I have scrapped the lot and kept very little intact and recoded. Currently only the build system is actually present, and the debugging tools (streaks, cursor, logging). I want to make sure it is working 100% before I continue. I do have exams and coursework to work on so I wont be doing as much as I could be doing other the next 3 weeks, but I need this code tested thoroughly.

Features of this build that are used:

Entirely new task system for building
Integration of Krogothes mex class
Universal mod support
Eradication of all known bugs including the out of virtual memory bug
Plaintext logging turned on by default, it will now log to AIlog.txt
Numerous fixes to logging class (such as when NTAI deliberatly crashed by typing .crash, the html tags where messed, all gone now)

Features that have yet to be added (coded but not enabled):

Tasks that check themselves and plan for contingencies (with extra unit behaviours) to eradicate idle unit stalls
Idle unit behaviours (such as repairing nearby units, reclaiming forests of trees, idle attack units approaching enemy units that arent protected)
Better building selection (NTAI isnt very good at picking it's buildings, and improvised learning code has bene commented out completely)

I need this testing on as many mods as possible on as many maps as possible, and all the logfiles sent to me, preferably by email or pm but not by posting them in this thread.

People who continually get errors on NTAI 0.28.10+ try running this and tell me if they still occur?

And I repeat this is not going to give you a good fight, it isnt meant to be a new release of NTAI it is for testing purposes only, so no AAI or OTAI performs better than this, or this AI sucks it didnt even attack me, it is not that sort of release

Download here Test Build 4
PS: there's no antistall code in here so if the commander builds something then does absolutely nothing take control manually and type in the command ".verbose" then move it around or press stop and see if it drives back into action. If the array of messages apeparing in the chat console mentions the word exception, tell me immediatly!!!!!!!!! And provide a log god dammit!

Remember this version is mod independant ti will work just as well on XTA core as AA-F Arm as TA:FF Core as TA:K Veruna or AATA etc

".verbose" - displays log messages in console when logging to file
".crash" Forcibly crashes Spring[/b]

NTAI Test Build 4

cookie for jellyman and weaver for their tests, this version ahs a change to give more logging info aswell as an extra check which might solve the problem.

Test Build 4

A minor fix and lots of logging additions to track an error (see bits int he log where it's like [00:43][00:43][00: [00:43]? now I get to find out what's generating them)

No more test runs of test builds 1+2 though, I've changed the log file header to suit. And could somebody take a pic of it running under some strange mod such as the TA:kingdoms port?

That include infolog.txt aswell as ailog.txt and tell me how it crashed (ai exception message? The usual spring crash?)

thx again to all those who've sent in logs so far ^_^

Whats the point in releasing a "test build" that reproducably crashes on any map at startup?
Run that thing in your debugger and just trace the crash, would be much more helpful than getting tons of log that look like this:

[00:07] Exception, Update
[00:07] Exception, Update
[00:07] Exception, Update
[00:08] Exception, Update
[00:08] Exception, Update
[00:08] Exception, Update
[0

.

Oh, and a map where it doesnt crash on startup is wideopencombat. But _any_ kind of chat command will cause a hard crash (.lines, .cheats, you call it....)

Insufficient information. i ahve no information thre about what happened upto the crash also, noen fo the numerous logfiles I have recieved had eceptions int he Update() routine, primarily because there is no update routine for there to be an exception in.

The point is that i want this 100% stable 100% robust. The problems are as follows:

I have limited time avilable to doe xtended test plays

For all purposes it runs fine on my computer with no probs but I knew it crashed masse's comptuer and Noizes too soemtimes when ti ran fien on ym computer.

My debugger is broken and I dont know what's wrong with it.

What may be reproducable on oen system isnt reproducable on others. I cannot reproduce the errors people are showing me but they can reproduce them, I thus want to investigate it

I dont have every map and mod available and others may eba ble to test ti out on those mods. I would onyl test it out on a handful of maps and XTA core, TLL if you're lucky because my situation doesnt afford me a great wad of download bandwidth anytime anywhere.

As fo yet i dont think it's too safe to run the .lines commands etc, .crash does what ti says perfectly, as does .verbose. The streaks and cursor will ahve to eb recoded anyway.

And let's not forget that no two computers are the same, even if they've only been operating for a few days, and are the exact same specifications.

whats the diference of these builds say to ntai0.28.10 ?

NTAI 0.28.10+'s purpose is to provide a skirmish AI.

The dev builds purpose are to test stability in new code, so that the next official NTAI version works perfectly on as many mods/maps/computers as is possible. Afterrall is the last thing you want is to see an error prompt? While I can test and fix on my machine till it works fine, the problem comes when the perfectly working code I have only works perfectly on my machine, so I need it testing on other machines, on maps I dont have with mods I dont have, in time I dont have.

When it's time for the next proper NTAI release you'll be seeing NTAI 0.3, it'll be based on these dev builds, but these dev builds are nowhere near a completed AI. The current one is just enough to test out, it doesnt however ahve attacking, scouting, and so on.

Yay the AI now works on my computer. Only one besides JCAI to do so. Tested on AA and SWTA. Construction units still tend to be quite prone to becoming permanently idle. Current build choices are bad, but I think you are yet to work on that aspect.

I playtested on Xect v Mynn (outdated though), and foudn one ro two mistakes I'd made, one of whcih si backed up by results Targon sent me of NTAI Build 4 running under TA:FF, I made the mistake of not checking if the unit could fly thus flying units where treated as immobile structures. Also the task lists where itnerpreted back to front for some reason. I'll add some code to try and do the new anti-unit stall stuff before I release build 5 and I'll add the unit behaviours (repair nearby + attack nearby+reclaim nearby). Otherwise build 4 is stable, and I have seen not one shred of evidence to say otherwise.

Also current build choices are appalling, codewise it lists all the units it can build that do the required job (e.g. all the factorys it can build), then picks the first one in the list.

After testing in for several hours in my mod.

It will build ocassionally stalling of course.. It never attacked.

Also it only build a few of the buildings in the first tech tree never moving up to second level. It also only built 2 units out of the basic factory. The construcion unit and the zaku120mm that was it. So all there was were level 1 buildings, power plant, mechineist, turret and pad. It did not build the second factory nor did it when I built the second factory utlize the construction unit there in.

Again, it did not attack.. it just sat there building construction zakus and zakuII(the machinegun version)

There is only construction code. Also any call to build a factory simply returns the first factory it finds whcih seems to be the lvl 1 factory because ti's the first in the build menu.

When I finish Antistall code in build 5 I'll do that and you should see an improvement.

no problem just trying to give you feedback.

I'd have released test build 5 but a combination fo unit behaviour and AI behaviours are making testing nanoblobz impossible past building a single sheep.

Test Build 5

This version however ahs scouting and atatcking agents forward proted from NTAI 0.29b, with a few extra fixes and tweaks.

Nanoblobz/TA:K Taros support, antistalling measures, anti overflow command caching, core commander hardcoded build tree added and more...

And please, post reports on darkstars.co.uk and not pm/post here, I can onyl access darkstars in shcool and not ehre, so I'll have ample tiem to go voer them adn reply and respond quickly, or take suggestions, and maybe even have the time to spell better ^^

Ok, after 3 tests, the AI still stalls out standing there looking for an overlord to press stop...

Also, I have not gotten far enough to see how it performs because it crashes.