nm it was unit learnign data.
can someone take the NTai source, start a game of XTA 8.1 with lindirs config on brazillian battlefields as arm, and play the game. I can crash NTai repeatedly on this map but I cant debug it so I cant find where its crashing.
Move the NTai commander around and press stop on it, build lots of builders and so on, keep pressing stop on idle buidlers untill it crashes or they do stuff, if they're idle mvoe them into open areas.......
NTai XE10.1b
Moderators: hoijui, Moderators
It always crashes for me after an AI builds its second mex. However, on a different computer where I plugged the downloaded config into a config file generated by an older version of NTai, it doesn't crash. For team 2 log looks like this:
[00:57] < Frame: 1739 ><1>function :: COrderRouter::CleanUpOrders()
[00:57] < Frame: 1739 ><1>function :: COrderRouter::IssueOrders()
[00:57] < Frame: 1739 ><1>function :: CActions::Update
[00:57] < Frame: 1739 ><1>function :: Assigner::Update()
[00:57] < Frame: 1739 ><1>function :: CManufacturer::Update
[00:57] < Frame: 1739 ><1>function :: CUnit::RecieveMessage
[00:57] < Frame: 1739 ><1>function :: CUnit::GetAge
[00:57] < Frame: 1739 ><1>function :: CKeywordConstructionTask::Init
[00:57] < Frame: 1739 >CKeywordConstructionTask::Init
Team 1
[00:57] < Frame: 1739 ><0>function :: CActions::Update
[00:57] < Frame: 1739 ><0>function :: Assigner::Update()
[00:57] < Frame: 1739 ><0>function :: CManufacturer::Update
[00:57] < Frame: 1739 ><0>function :: CUnit::RecieveMessage
[00:57] < Frame: 1739 ><0>function :: CUnit::GetAge
[00:57] < Frame: 1739 ><0>function :: Chaser::Update
[00:57] < Frame: 1739 ><0>function :: Chaser::Update :: DONE
[00:57] < Frame: 1739 ><0>function :: Global::Update :: done
The logs look the same for every time it's crashed.
[00:57] < Frame: 1739 ><1>function :: COrderRouter::CleanUpOrders()
[00:57] < Frame: 1739 ><1>function :: COrderRouter::IssueOrders()
[00:57] < Frame: 1739 ><1>function :: CActions::Update
[00:57] < Frame: 1739 ><1>function :: Assigner::Update()
[00:57] < Frame: 1739 ><1>function :: CManufacturer::Update
[00:57] < Frame: 1739 ><1>function :: CUnit::RecieveMessage
[00:57] < Frame: 1739 ><1>function :: CUnit::GetAge
[00:57] < Frame: 1739 ><1>function :: CKeywordConstructionTask::Init
[00:57] < Frame: 1739 >CKeywordConstructionTask::Init
Team 1
[00:57] < Frame: 1739 ><0>function :: CActions::Update
[00:57] < Frame: 1739 ><0>function :: Assigner::Update()
[00:57] < Frame: 1739 ><0>function :: CManufacturer::Update
[00:57] < Frame: 1739 ><0>function :: CUnit::RecieveMessage
[00:57] < Frame: 1739 ><0>function :: CUnit::GetAge
[00:57] < Frame: 1739 ><0>function :: Chaser::Update
[00:57] < Frame: 1739 ><0>function :: Chaser::Update :: DONE
[00:57] < Frame: 1739 ><0>function :: Global::Update :: done
The logs look the same for every time it's crashed.
My own investigation into this hass defied explanation. I cant determine the cause using logging, I need a debugger, which is unfortunate as codeblocks+gdb never executes spring.exe thus never debugging NTai, and Visual studio 2005 builds NTai with different sized UnitDef classes, making the VS2005 build of NTai unsusable.
I can get simple stacktraces that give addresses when spring crashes via infolog.txt but they're just addresses and I dont know how to resolve them into anything meaningful.
I can get simple stacktraces that give addresses when spring crashes via infolog.txt but they're just addresses and I dont know how to resolve them into anything meaningful.
*doublepost sorry*
Although I did manage to solve a few issues with build placement and other things, not that all issues have dissapeared, I have debunked the myth that factories never build, and fixed a factory unit building bug that made them use the full build placement algorithm (doesnt affect factories building units, but it does affect them once the blocking map around the factories full, and it lags).
I also found out slopemaps arent used at all for building placement, but maxheight difference is used instead.
Although with my changes the circumstances for the crash happening increase greatly.
Although I did manage to solve a few issues with build placement and other things, not that all issues have dissapeared, I have debunked the myth that factories never build, and fixed a factory unit building bug that made them use the full build placement algorithm (doesnt affect factories building units, but it does affect them once the blocking map around the factories full, and it lags).
I also found out slopemaps arent used at all for building placement, but maxheight difference is used instead.
Although with my changes the circumstances for the crash happening increase greatly.
-
1v0ry_k1ng
- Posts: 4656
- Joined: 10 Mar 2006, 10:24
my attempt to use the newest NTAI and lindirs build tree with xta 8.1 resulted in this log and the commander not moving. what am I doing wrong?:: NTAI XE9.5 Log File
:: Programmed and maintained by AF/T.Nowell
:: Copyright (C) 2004-7 Tom Nowell/AF
:: Released under the GPL 2.0 Liscence
:: Game started: 6.2.2007 18:27:19
:: XTA v8.1
:: XTA Version 8.1
:: First instance of NTAI
:: [RoX]1v0ry_k1ng
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: spring_engine
:: AI DLL's in game
:: AI/Bot-libs/NTai.dll : AI :: 1
[-] < Frame: 0 >logging started
KAI Metal Class by Krogothe
Metal Spots loaded from file
Metal Spots Found 60
:: Found 60 Metal Spots
[-] < Frame: 0 >Order Router constructed
[-] < Frame: 0 >DTHandler constructed
[-] < Frame: 0 >RadarHandler constructed
[-] < Frame: 0 >Planning constructed
[-] < Frame: 0 >Assigner constructed
[-] < Frame: 0 >Scouter constructed
[-] < Frame: 0 >Economy constructed
[-] < Frame: 0 >Manufacturer constructed
[-] < Frame: 0 >BuildingPlacer constructed
[-] < Frame: 0 >Chaser constructed
[-] < Frame: 0 >Initialisising
[-] < Frame: 0 >Mod TDF loaded
[-] < Frame: 0 >Mod TDF loaded
[-] < Frame: 0 >values filled
[-] < Frame: 0 >Value ai\solobuild missing in file buffer
[-] < Frame: 0 >Value ai\alwaysantistall missing in file buffer
[-] < Frame: 0 >Value ai\singlebuild missing in file buffer
[-] < Frame: 0 >Arrays filled
[-] < Frame: 0 >Loading unit data
[-] < Frame: 0 >error in Global InitAI, cannot continue
[-] < Frame: 0 >boost::spirit::parser_error
delete the learning file, you have an old version which has syntax errors which is conflicting with the new vesion.
In the mean time nvidia have compeltely obliterated all ability to run spring on my PC. I installed the newer 101.41 drivers and got the GL_ARB_Texture_env_combine extension is missing error. The problem is I've downgraded to the 100.65 drivers again yet the error wont go away. I cannot debug NTai, I cannot run NTai to log its output, I cant even start spring, and nobody seems at all interested in running NTai through a debugger.
In the mean time nvidia have compeltely obliterated all ability to run spring on my PC. I installed the newer 101.41 drivers and got the GL_ARB_Texture_env_combine extension is missing error. The problem is I've downgraded to the 100.65 drivers again yet the error wont go away. I cannot debug NTai, I cannot run NTai to log its output, I cant even start spring, and nobody seems at all interested in running NTai through a debugger.
Also be aware that as I stated, lindirs config is incomplete. He did not include the msot important file in the entire config, and as such you'll need NTai to generate a blank config, then copy paste XTA_BT2.tdf from lindirs archive into /NTai/configs/
I'd like it if somebody uploaded an unborked version of lindirs config.
And all players (IvoryKing?) who want to downlaod NTai, and lindirs config and play happy XTA games and play away the hours, you're totally out fo luck. You'll experience random hard crashes, and the NTai commander will barely do anything its told. My own personal copy was starting to behave itself save for the horrible crash bug that has yet to be fixed, and the terrible nvidia driver attrocity that followed.
I'd like it if somebody uploaded an unborked version of lindirs config.
And all players (IvoryKing?) who want to downlaod NTai, and lindirs config and play happy XTA games and play away the hours, you're totally out fo luck. You'll experience random hard crashes, and the NTai commander will barely do anything its told. My own personal copy was starting to behave itself save for the horrible crash bug that has yet to be fixed, and the terrible nvidia driver attrocity that followed.
kloot, you obviously didnt test NTai while debugging at all and ran it with a mod that had no config, as I have uncovered a crashbug which you must have come across if you had used a config.
I spent ages rebuilding my old PC and getting ti goign and ran NTai through the debugger, and found that it was passing a null pointer along during universal build routines involving b_rule and its variants. Still running my old PC is a nightmare the monitor has a giant blurry patch where its not been used for so long that I cant read, and the mouse double clicks when you single click, and the keyboard keys stick easily, and there's no stand so I have to sit on the tower with the mouse above my head on my other desk.
I'll post RC3 shortly
I spent ages rebuilding my old PC and getting ti goign and ran NTai through the debugger, and found that it was passing a null pointer along during universal build routines involving b_rule and its variants. Still running my old PC is a nightmare the monitor has a giant blurry patch where its not been used for so long that I cant read, and the mouse double clicks when you single click, and the keyboard keys stick easily, and there's no stand so I have to sit on the tower with the mouse above my head on my other desk.
I'll post RC3 shortly
http://spring.unknown-files.net/file/24 ... t_version/
XE9.55
please test as much as possible and identify as many issues as possible and list them nicely, then summarize.
be aware that the XTA arm peewee is used for debugging purposes and if your not using XTA it will spam errors saying it doesnt exist. Ignore these.
Also for logging purposes try not to run 2 or more NTais, because then you have the added confusion of figuring out which NTai crashed and which didnt.
XE9.55
please test as much as possible and identify as many issues as possible and list them nicely, then summarize.
be aware that the XTA arm peewee is used for debugging purposes and if your not using XTA it will spam errors saying it doesnt exist. Ignore these.
Also for logging purposes try not to run 2 or more NTais, because then you have the added confusion of figuring out which NTai crashed and which didnt.
1) You were talking about a crashbug at startup which I pointed out and fixed, and that was the extent of my testing NTAI (ie. making sure it compiled and loaded OK sans config since I didn't have one).AF wrote:kloot, you obviously didnt test NTai while debugging at all and ran it with a mod that had no config, as I have uncovered a crashbug which you must have come across if you had used a config.
2) I don't really appreciate being told what I did and didn't do in my not-quite infinite free time, thanks.
That is your original message in reference to the crashbug.Just compiled it, doesn't crash at all and nicely writes stuff to infolog.txt and AI/NTai.
And in all honesty, I was right. You didnt debug NTai with a config, you just started it, exited after a few seconds, checked it had logged hey presto no crashes here, you even said so in the last post you made. And its an easy conclusion to make when you finally find the crashbug and see what it is and the fact that unelss you didnt have a config loaded, the crash was unavoidable and it would have shown up one way or another.
If you had told me that when you did it you could have saved me a lot of time, and a lot of confusion, because what you said is a vague "I debugged NTai, it ran fine, its outputting to the logfiles all good and proper". And there are XTA and EE configs on UF. You muddled up my trail of thought completely and left me clueless as to where to look for the bug. Imagine if you couldnt run spring or debug KAI, and I said it ran fine with a debugger and everybody else said it crashed instantly, yet I was adamant it ran fine? Would you not be peeved at all if it turned out all I had done was compile and load the game without playing the AI for any length of time with any test data in order to get a result? Its like experimenting with lab mice by prodding pictures of them and recording the result.You were talking about a crashbug at startup which I pointed out and fixed, and that was the extent of my testing NTAI
OK tested the 9.55 version BA 4.7 Altored Divide. Good news is the b_radar, b_defence now work again. The bad news is that now b_mex doesn't build metal extractors on metal - it just builds them anywhere it feels like, this makes testing the rest quite a problem... When i built mex's for it i still experienced the same problem of vehicle labs ceasing production, also the commander and construction vehicles stopped building after a while.
NTai XE9.56
changes:
- Some changes were made to the keyword system that might fix the b_power etc and streamlined the b_rule type keywords
- Extra behaviour added to the dgun code to make a dgunning unit runaway if damaged from afar by more than one third of its health, otherwise itll move into firing range and attack. This should prevent 2 or 3 rocket bots sitting outside the commanders firing range and attacking while NTais commander sits still.
- Debug symbols removed and speed optimizations reintroduced
- Changed a few parts to use C++ std::string operations rather than sprintf() and C style char[] buffers
- Other minor changes
Logging is still set at highly detailed so beware of gigantic log files reaching into the gigabytes for long games. This will be removed in XE9.57 if no major bugs are found in this build.
changes:
- Some changes were made to the keyword system that might fix the b_power etc and streamlined the b_rule type keywords
- Extra behaviour added to the dgun code to make a dgunning unit runaway if damaged from afar by more than one third of its health, otherwise itll move into firing range and attack. This should prevent 2 or 3 rocket bots sitting outside the commanders firing range and attacking while NTais commander sits still.
- Debug symbols removed and speed optimizations reintroduced
- Changed a few parts to use C++ std::string operations rather than sprintf() and C style char[] buffers
- Other minor changes
Logging is still set at highly detailed so beware of gigantic log files reaching into the gigabytes for long games. This will be removed in XE9.57 if no major bugs are found in this build.
