NTai XE10.1b
Moderators: hoijui, Moderators
Re: NTai XE9.8 experimental for 0.76b1
there's a version to be found here http://www.darkstars.co.uk/2008/ntai-and-076b1/
The crash bug I reported has never been resolved unfortunately so you may still have problems
The crash bug I reported has never been resolved unfortunately so you may still have problems
Re: NTai XE9.8 experimental for 0.76b1
That crash bug was fixed iirc, however due tot he confusion at the tiem I have no idea if ti was the same crashbug or what version ti was in.
After the boost threading bugs where fixed I attempted to recreate your issues and I never could despite what you said.
And no, as I keep saying, I do not have the ability at the moment to compile NTai, spring or any AI under mingw32, I do not even have it installed.
If you want a visual studio 2005 build however that can be arranged but there is nothing releasable at the moment, and I haven't the time or the will to mess around with mingw32 again.
After the boost threading bugs where fixed I attempted to recreate your issues and I never could despite what you said.
And no, as I keep saying, I do not have the ability at the moment to compile NTai, spring or any AI under mingw32, I do not even have it installed.
If you want a visual studio 2005 build however that can be arranged but there is nothing releasable at the moment, and I haven't the time or the will to mess around with mingw32 again.
-
- Posts: 10
- Joined: 02 Nov 2007, 12:49
Re: NTai XE9.8 experimental for 0.76b1
What mods does this AI support?
Re: NTai XE9.8 experimental for 0.76b1
At the moment there is no release for public consumption, and I am unable to compile a binary that will work with a released spring.
So I cannot answer that question.
So I cannot answer that question.
-
- Posts: 10
- Joined: 02 Nov 2007, 12:49
Re: NTai XE9.8 experimental for 0.76b1
Thanks for the info - best of luck with your AI.
-
- Posts: 34
- Joined: 29 May 2008, 21:06
Re: NTai XE9.8 experimental for 0.76b1
NTai looks great, and I want to use it on my Mac, so I was using XCode to compile a dylib of NTai. Unfortunately boost won't compile correctly. The error produced is:
Other than that NTai seems to compile correctly. Please help!
Code: Select all
/Users/<user>/Desktop/Source-Code/spring/trunk/AI/Global/NTai/AI/NTai/resources/boost/test/utils/wrap_stringstream.hpp:66: error: no match for 'operator<<' in '(+ targ)->boost::basic_wrap_stringstream<CharT>::stream [with CharT = char]() << t'
/Developer/SDKs/MacOSX10.4u.sdk/usr/include/c++/4.0.0/bits/ostream.tcc:67: note: candidates are: <snip>
Re: NTai XE9.8 experimental for 0.76b1
I would look into the version your using, of boost. That and you dont need to compile all of boost for NTai and spring either.
There's also an issue of configuration files once you've gotten NTai compiled for which there's no quick fix at the moment.
There's also an issue of configuration files once you've gotten NTai compiled for which there's no quick fix at the moment.
Re: NTai XE9.8 experimental for 0.76b1
This was a reply to someone who sent me a stack trace from one of the NTais I released just after spring 0.76b1 was released.More to the point you have an NTai that works with the current release.
This means:
I appreciate the feedback, but sadly at the moment while all those bugs are fixed and I have a stable NTai, I cant give you a copy as I can only compile using Visual Studio, and for it to work it needs compiling using mingw32.
- You're using one of the 3 or 4 ancient versions of NTai that i released just after spring 0.76b1 was released
These versions have known crash bugs involving threading which are fixed in the svn- You compiled it yourself
You should use a debugger and make sure you have the latest svn, cant use stack traces spat out by spring
Alternatively I believe NTai is still being built by buildbot. You could install the latest svn installer and download the NTai.dll built by buildbot, and then drop in the configuration files etc from the NTai download.
May I stress once again, there are no stable NTai builds that have been compiled for 0.76b1. All the bugs in these builds have already been fixed in the svn, and I do not have the means to build a spring compatible release containing those fixes.
I apologize for the inconvenience this incurs people but theres not a lot I can do about it with the time restraints I have and my other projects. NTai development for the moment remains in the Visual Studio realm, and Id be happy to work on that for academic purposes as I have done in the last months since Christmas.
But the hassle of keeping multiple svn revisions on my machine, and installing mingw32 and codeblocks and making sure they stay installed and working, as well as having projects and codebases for both visual studio and codeblocks, and the huge gap in support for debugging libraries under windows in mingw32 with gdb and codeblocks, and the issues with Vista caused by shoddy command line usage in mingw32 internals, Id rather just work on other stuff that will give me far more gains for far less effort.
It gives me a headache just thinking about it.
-
- Posts: 34
- Joined: 29 May 2008, 21:06
Re: NTai XE9.8 experimental for 0.76b1
Well, if I manage to build NTai for mac, then I will be able to use the XCode debugger pretty easily to figure out what is wrong. I also have access to a vista computer with visual studio on it, and it wouldn't be too hard to install mingw32 on that either.
Re: NTai XE9.8 experimental for 0.76b1
If you do get it working please could you drop me a PM and I'll send you the config I had over. Hopefully you might be able to prove that the crash bug I had is fixed or find the cause.
Re: NTai XE9.8 experimental for 0.76b1
Upload your config somewhere public instead.
After you reported that issue I tested and I found code that explained it all and I fixed it, I even reran the test and found a few more issues and fixed those, then ran tonnes of test and couldn't reproduce the error.
The further testing ran by students in germany which involved running games of NTai over and over again at high speed in batch runs suggests your bug was corrected. They did not report this issue and they were using the latest builds of NTai and spring in the svn.
Even the code in question was totally removed and rewritten using a totally different design.
If I remember correctly the issue involved construction plans which were shared between NTai instances and between threads. It involved my usage of mutexes and a tonne of dodgy usage of pointers. This would lead to crashes and invalidated iterators. The entire plan system was changed around, the bPlan structures in question were removed completely, the plan system was simplified and re-factored.
Now if you can reliably reproduce the error with the latest svn and prove without a doubt that it is actually svn code and not backdated code, in such a way that I myself can reproduce the error, or at least understand the error or know where the code crashes then I can do something about it. But until then I have no clues as to where it crashes, and I have no proof that it crashes save for your word which at the moment is just "it crashes" with little explanation, and I cant do anything with that aside from spend a lot of time scanning the code base for anomalies which I don't have time for.
Regarding further changes to NTai, after discussion with some of the students after the academic involvement, I did several small refactors which greatly simplified things regarding unit task management, mainly the use of a separate class to track and provide the CUnit object with tasks.
The intention of this was that academics could create new TaskFactory objects that did not read the task lists and implemented their own rules and logic. This would allow them near total control of the CUnit object and the task management by being able to introduce their own tasks as well as having full control over unit behaviour.
After you reported that issue I tested and I found code that explained it all and I fixed it, I even reran the test and found a few more issues and fixed those, then ran tonnes of test and couldn't reproduce the error.
The further testing ran by students in germany which involved running games of NTai over and over again at high speed in batch runs suggests your bug was corrected. They did not report this issue and they were using the latest builds of NTai and spring in the svn.
Even the code in question was totally removed and rewritten using a totally different design.
If I remember correctly the issue involved construction plans which were shared between NTai instances and between threads. It involved my usage of mutexes and a tonne of dodgy usage of pointers. This would lead to crashes and invalidated iterators. The entire plan system was changed around, the bPlan structures in question were removed completely, the plan system was simplified and re-factored.
Now if you can reliably reproduce the error with the latest svn and prove without a doubt that it is actually svn code and not backdated code, in such a way that I myself can reproduce the error, or at least understand the error or know where the code crashes then I can do something about it. But until then I have no clues as to where it crashes, and I have no proof that it crashes save for your word which at the moment is just "it crashes" with little explanation, and I cant do anything with that aside from spend a lot of time scanning the code base for anomalies which I don't have time for.
Regarding further changes to NTai, after discussion with some of the students after the academic involvement, I did several small refactors which greatly simplified things regarding unit task management, mainly the use of a separate class to track and provide the CUnit object with tasks.
The intention of this was that academics could create new TaskFactory objects that did not read the task lists and implemented their own rules and logic. This would allow them near total control of the CUnit object and the task management by being able to introduce their own tasks as well as having full control over unit behaviour.
Re: NTai XE9.8 experimental for 0.76b1
Heya, the most recent SVN NTai seems to have problems in BA when a wind generator is attacked. I noticed this first when I put KAIK vs NTai, and also NTai vs NTai --- every time a wind generator is near death, spring freezes.
I tried this myself, and when I tried to shoot a wind generator with my comm, spring froze. Can anybody reproduce this?
I tried this myself, and when I tried to shoot a wind generator with my comm, spring froze. Can anybody reproduce this?
Re: NTai XE9.8 experimental for 0.76b1
I cant think of anything that would cause such a slowdown on unit death. If it were a unit finishing its orders or a unit that's just finished being built perhaps I could understand that, but Im inclined to think its spring at fault or possibly BA, especially since its a wind generator being destroyed not an active unit with a tasklist, or a scouter, or an attacker.
More information is needed. Does this happen in KAIK vs KAIK too? if so then you can definately discount both NTai and KAIK.
More information is needed. Does this happen in KAIK vs KAIK too? if so then you can definately discount both NTai and KAIK.
Re: NTai XE9.8 experimental for 0.76b1
No, this only happens when it's NTai vs anything else.AF wrote: More information is needed. Does this happen in KAIK vs KAIK too? if so then you can definately discount both NTai and KAIK.
It doesn't happen to any other AI.
Re: NTai XE9.8 experimental for 0.76b1
Can you verify this happens in a game where there's is only one instance of NTai and no other AIs, and no lua widgets turned on?
Re: NTai XE9.8 experimental for 0.76b1
is there a version of ntai for this release? I'm hoping there was a version from the buildbot as I know you don't use ming
Re: NTai XE9.8 experimental for 0.76b1
Go grab the buildbot version and upload it somewhere before it disapears or the c++ AI interface changes again. Ive not tested it or downloaded it myself
Re: NTai XE9.8 experimental for 0.76b1
id love to grab the buildbot version but i don't know where it is!
Re: NTai XE9.8 experimental for 0.76b1
Im not sure either =/
Re: NTai XE9.8 experimental for 0.76b1
Tested using hoijui's build and i got the following crash bug
Assertion Failed
AI/Global/Ntai/Units/CConfigTaskManager.cpp
Line 56
Expression HasTasks()
having said that the crash bug occurred about 10 minutes in at which point it had looked really strong, attacking is working really well again.
Assertion Failed
AI/Global/Ntai/Units/CConfigTaskManager.cpp
Line 56
Expression HasTasks()
having said that the crash bug occurred about 10 minutes in at which point it had looked really strong, attacking is working really well again.