Spring RTS Engine

Also, i noticed crashes when running it on AA 1.35s.
It seems a bit map dependend, but it crashes on "clearaditch" about 20 minutes into the game...

But one thing i really love is that this is the first ai i ever saw to make sensible defense placements.

On desert dunes, an ai starting in the corner, being attacked from the south in the game before, build right in th beginning defenses towards the south. After an attack from the west, it responded by building up defenses there, too.

Btw: does the non-existant expansion also fall under the "no econoimic model"? As it seems the ai never tries to build on mexes more then 1 or 2 mapunits from the startpos, and generally doesnt try to spread across the map.

I just played Absolute Annihilation 1.35 game with 3 AAI in map "Eurogreen". All AI's had built lots of metal extractors and then stopped doing anything. One of those AI's built 1 solar and 21 mexes. LOL
I can send you log from this game if you want.

I also tried to play AAI in mod The Lost Legacy 1.04 and it always crashed at the beginning. I tried various maps (Omega, Mars, Small Divide) but it crashed when the game starts.


Btw sorry for my English, i hope you understand what I mean.

IMSabbel wrote: Btw: does the non-existant expansion also fall under the "no econoimic model"? As it seems the ai never tries to build on mexes more then 1 or 2 mapunits from the startpos, and generally doesnt try to spread across the map.

yep thats it :)

you can increase MAX_BASE_SIZE in the mod cfg-file (iirc 5 by default) but this will have some negative side effects:
aai will place its defences only at the borders of its base and it will use its commanders to build things all over its base, so the greater the base the more likely the comm. walks into enemy lines...

@rafal99:
thx for letting me know, i'll download this map and check it...
i'd appreciate if you mail me the log files (you'll find my mail-add in the readme)

Will future versions support a base hirarchy?
The current "base" is quite nice for the central production/economy area, but it would be nice when the ai would also create smaller "outpost" bases, with priorities shifted to defense, while filling the area inbetween with mexes...

Playing versus 3 AAI bots I suffered waves of simultaneous 3-way team assaults, both with ground units and air support.

Very impressed - even with the diversion of resources to attack 0,0 bug - at the job the AI's did. They kept me bottled up on a hill, on the Plains and Passes map.

Unfortunately Spring crashed as another massive triple assault wave was closing in...

this ai is already pretty ok to, but its true he should first improve its stability a bit more because it isnt too stable yet.
I can rarely play 30 min without a crash

strangly, 0.30 NEVER crashed on me, even if i let it run for hours (at least if it managed to parse the metalmap).

So something that changed to the latest version has to be at fault...

i think its a problem with the introduction of aircrafts:
they can fly "out of the map" causing aai to address sectors that doesnt exist -> memory access fault


im working on a fixed version, expect a small bugfix release soon

my fixed version ran stable for nearly two hours (2 vs 2 on comet catcher) before finally spring crashed. none of the crashes i observed that evening occurred in aai's code (debugger reported crashes in other dlls) so i assume it's spring that crashes the game

Well done!

When will you be releasing this new version?

look here
http://taspring.clan-sy.com/phpbb/viewt ... 1226#51226

submarine wrote:my fixed version ran stable for nearly two hours (2 vs 2 on comet catcher) before finally spring crashed. none of the crashes i observed that evening occurred in aai's code (debugger reported crashes in other dlls) so i assume it's spring that crashes the game

However only some ai's seem to be pushing spring crashes , and lets not forget that not all the spring code has checks and therefore an ai misscoding can generate engine crashes (ive seen it myself on my ai).

agreed, but its very hard to find bugs caused by sending wrong data to the interface. until now most aai crashes have been related to wrong adresses (array out of bound exceptions, deleted pointers were another class sharing this pointer had not been informed and so on)

but for me ta spring crashed in human vs human as well (especially when major battles take place)

Tobi (tvo on berlios) recently discovered that std::sort actually corrupts data if a >= is used instead of > (or was it the other way around )... so the next release might actually be somewhat stabler in heavy battles (this bug involved the CProjectileHandler)

submarine wrote:agreed, but its very hard to find bugs caused by sending wrong data to the interface. until now most aai crashes have been related to wrong adresses (array out of bound exceptions, deleted pointers were another class sharing this pointer had not been informed and so on)

but for me ta spring crashed in human vs human as well (especially when major battles take place)

In human vs human ive played hundreds of hours and have noticed very few crashes lately, and with your previous ai versions i did notice quite some engine crashes.
Try making some mad testing in VS or another compiler running the spring there so that u can get those crashes and see what func's are receiving wrong values . (ive done quite some times and its very easy to figure the problme that way)

Chocapic wrote: Try making some mad testing in VS or another compiler running the spring there so that u can get those crashes and see what func's are receiving wrong values . (ive done quite some times and its very easy to figure the problme that way)

lol what do you think i'm trying to do?

of course im using vs.net debugger but the debugger tells me the crashes happen somerwhere in spring not in aai - at least the last ones, there are probably more aai bugs i havent come across yet.

vs.net can only offer me a disassembly of the lines where it crashed but this is not very useful for me:)

submarine wrote:

Chocapic wrote: Try making some mad testing in VS or another compiler running the spring there so that u can get those crashes and see what func's are receiving wrong values . (ive done quite some times and its very easy to figure the problme that way)

lol what do you think i'm trying to do?

of course im using vs.net debugger but the debugger tells me the crashes happen somerwhere in spring not in aai - at least the last ones, there are probably more aai bugs i havent come across yet.

vs.net can only offer me a disassembly of the lines where it crashed but this is not very useful for me:)

Most of the times that it crashes in spring for me its in some spring funcs or some c++ libs where i can atleast have an idea where it was as i have access to the callstack and so!
Im not saying its allways straight forward to see where the bug is but most of the times it isnt very hard , and i thought you werent do that.
But if you are thats the way

looking at the call stack tells me there were no aai-functions called before the crash. zaphod mentioned theres a bug in the projectile handler i think this could probably be the reason, since all non-aai crashes occured when major battles took place

2 errors I've noted, both AI related possibly but Spring caused essentially.

Selecting units the AI has given commands to to change to high trajectory:
If you tell the itnerface to change all AK's traj setting to high traj only, it works! No visible effect, hwoever if you then select the AK the game crashes. Go take cotnrol fo NTAI 0.29a and try to select a thud, adn you'll se eit happening but it doesnt affect you if you just pay against it.

Printscreen:
Sometimes it crashes instantly on pressing f12, but onyl if I'm runngin an AI for some reason.

Units pushing other units:
A unti pushes another unit out of the way if ti is blocking a path, it even pushes untis that arent finished building, and whatsmroe ti pushes them off the map, causing crashes when they try to execute commands ro printscreens, (especially prominent on a game of NTAI test build 5 nanoblobz v0.1, where I had mistakenly made con bots progress northwards building and they got in a jam at the top and started pushing units off map.