NTai XE10.1b
Moderators: hoijui, Moderators
NTai XE10.1b
Last edited by AF on 04 Mar 2010, 19:18, edited 61 times in total.
Testing
I am sorry to report that I Just tryed out your recent Ai verson using XTA mod on the following maps of Evergeenhaven, Foothills and Mars. From the start of each game I continually got delayed sync and no sync responces and made it unplayable.
It is an engine bug, I have had this bug playing offlien with no AI's runnign at all when against the random spawn script.
You can stop the hogging by manually taking control of the unit causing the problem and moving it, or destroying whatevers stopping it.
And there is nothing in the AI code that makes a freaker want to walk into the wall next to it. The engine should give a UnitMoveFailed() event call to the AI in this case, but it doesnt, it should also wait till the unti has left the factory to call UnitIdle like it usually does rather than issue it as soon as it gets built.
I will look into ti but ti will mean introducing a lot more code into the scouter and factor classes, and if I do manage a workaround then I'll do a 0.29a release, with fixed unit capture code and maybe a few other random improvements.
But ym code is not at fault. This code was in place before the problem started in th first place, it wasnt till 0.65 that I ever saw this problem, and I made error reports about it and nothing was done.
Try it again and press f3, and check if the factory is being blocked by a tree, stopping untis from exiting.
You can stop the hogging by manually taking control of the unit causing the problem and moving it, or destroying whatevers stopping it.
And there is nothing in the AI code that makes a freaker want to walk into the wall next to it. The engine should give a UnitMoveFailed() event call to the AI in this case, but it doesnt, it should also wait till the unti has left the factory to call UnitIdle like it usually does rather than issue it as soon as it gets built.
I will look into ti but ti will mean introducing a lot more code into the scouter and factor classes, and if I do manage a workaround then I'll do a 0.29a release, with fixed unit capture code and maybe a few other random improvements.
But ym code is not at fault. This code was in place before the problem started in th first place, it wasnt till 0.65 that I ever saw this problem, and I made error reports about it and nothing was done.
Try it again and press f3, and check if the factory is being blocked by a tree, stopping untis from exiting.
well I've had this problem and 80% of the time ti was on brazillian battlefields and the AI ahd built a factory in the middle of a forest adn a freaker got trapped.
Does KAI have the same problem? Eitherway I'll work on optimizing several routines and release a better version shortly, perhaps with nanoblobz support if I can backwards port from 0.3, or forwards port 0.29 routines to the test build code.....
Does KAI have the same problem? Eitherway I'll work on optimizing several routines and release a better version shortly, perhaps with nanoblobz support if I can backwards port from 0.3, or forwards port 0.29 routines to the test build code.....
-
- Posts: 46
- Joined: 25 Jul 2005, 20:07
Could it be that NTAI sends too much Commands into the networking stream? Such a flood of network packets can cause the sockets to drop packets I think, it certainly has some negative effect.
IIRC the loopback network buffer has a limited size
In any case it is pointless to keep saying that it is an engine bug, when the other AIs don't have this problem. That's just putting your head in the sand.
I'm not saying the engine is bugfree, but NTAI must be doing something from the other AIs causing this behavior, so you're the one who has to start looking into what causes this.
IIRC the loopback network buffer has a limited size
In any case it is pointless to keep saying that it is an engine bug, when the other AIs don't have this problem. That's just putting your head in the sand.
I'm not saying the engine is bugfree, but NTAI must be doing something from the other AIs causing this behavior, so you're the one who has to start looking into what causes this.
Yes when the commander is attacked the d-gun bug will not occur.
As for that zaphod, I notice that when a freaker gets stuck (sometimes not in a factory at all), it generates these problems yet there is no unitIdle or UnitMovefailed calls according to chat console in verbose debug mode, so no extra commands are executed (there isnt code pertaining to this present in update()).
As for that zaphod, I notice that when a freaker gets stuck (sometimes not in a factory at all), it generates these problems yet there is no unitIdle or UnitMovefailed calls according to chat console in verbose debug mode, so no extra commands are executed (there isnt code pertaining to this present in update()).
QUOTE:start of each game I continually got delayed sync and no sync responces and made it unplayable.
I got the errors at the start of each game. No attacking was going on ,no dguning, no bots stuck in factories, as I had none built. On each of the maps I was only able to build 2 solars and 2 mexes, I was playing 2 ntals against me. I get no sync errors yet with any of the other current AI's
I got the errors at the start of each game. No attacking was going on ,no dguning, no bots stuck in factories, as I had none built. On each of the maps I was only able to build 2 solars and 2 mexes, I was playing 2 ntals against me. I get no sync errors yet with any of the other current AI's
i deffinetly get this errors of network stream.. will look for screshots.Zaphod wrote:Could it be that NTAI sends too much Commands into the networking stream? Such a flood of network packets can cause the sockets to drop packets I think, it certainly has some negative effect.
IIRC the loopback network buffer has a limited size
In any case it is pointless to keep saying that it is an engine bug, when the other AIs don't have this problem. That's just putting your head in the sand.
I'm not saying the engine is bugfree, but NTAI must be doing something from the other AIs causing this behavior, so you're the one who has to start looking into what causes this.
and i did report that this build .29 does not build const bots!
maybe it happened just once, will test again and see what happens.
hm... ive seen it building radar tower.. question. when it gets destroyed, does it rebuild it?. cheers.
I wasnt aware of the possibility of the network buffer being voerran by too many commands.
To correct this I shall release 0.29a shortly, which is incapable of issuing mroe than 6 commands per frame, any mroe and ti caches them and executes them in the next frame.
As usual, I release fix, I add an improvement to make up.
The build tree has been tweaked again to take into account games where the AI has been given lots of resources,
Con units reclaiming when attacked wont restart and waste time if another shot is fired anymore.
A few threat matrix improvements, aswell as an initial threat matrix adjustement so attack units are more likely to attakc an enemy startpos than an outpost.
The first Attack group NTAI builds will raid mex spots rather than attack the largest threatening sector.
To correct this I shall release 0.29a shortly, which is incapable of issuing mroe than 6 commands per frame, any mroe and ti caches them and executes them in the next frame.
As usual, I release fix, I add an improvement to make up.
The build tree has been tweaked again to take into account games where the AI has been given lots of resources,
Con units reclaiming when attacked wont restart and waste time if another shot is fired anymore.
A few threat matrix improvements, aswell as an initial threat matrix adjustement so attack units are more likely to attakc an enemy startpos than an outpost.
The first Attack group NTAI builds will raid mex spots rather than attack the largest threatening sector.
that sums up to roughly 180 commands per second, what the hell are you doing?AF wrote: To correct this I shall release 0.29a shortly, which is incapable of issuing mroe than 6 commands per frame, any mroe and ti caches them and executes them in the next frame.
i'd estimate aai doesnt issue more than 5-10 per sec