I downloaded the new install file, and start a SP game, and it says, incorrect global ai dll: \aidll\globalai\KAI.dll
Can someone help me?
[AI: KAI] Why it dont work with 0.73b1?
Moderators: hoijui, Moderators
2 KAI binaries have been on smoths forum for the last 2 weeks, and they're a testament to just how badly spring 0.73 fucked up the AIs. KAI in 0.73 is a huge dissapointment when compared to the same code in spring 0.72.
KAI 0.11 and 0.2 ned dbugging by someone who knows what they're doing to fix them, but the only people with enough knowledge todo that are krogothe tournesol and firenu.
Also krogoth, by reverting KAI 0.11 to your original copy in the svn you undid a slew of crashbugfixes tobi made.
KAI 0.11 and 0.2 ned dbugging by someone who knows what they're doing to fix them, but the only people with enough knowledge todo that are krogothe tournesol and firenu.
Also krogoth, by reverting KAI 0.11 to your original copy in the svn you undid a slew of crashbugfixes tobi made.
File bugs then instead of complaining... the first will get you a solution, the second will get you nothing.
Which functions don't work?
What is broken?
Which revision broke it? (do a bisection search on all revisions between 0.72 and 0.73)
Preferably submit smallest-possible code to reproduce the problem too.
Personally I haven't seen any problems with 0.73 and AIs that are clearly caused by the engine.
Which functions don't work?
What is broken?
Which revision broke it? (do a bisection search on all revisions between 0.72 and 0.73)
Preferably submit smallest-possible code to reproduce the problem too.
Personally I haven't seen any problems with 0.73 and AIs that are clearly caused by the engine.
changes in UnitAI, and UnitDef affected AIs, how exactly we aint sure, but for some reason code that works wit 0.72 doesnt work with 0.73 without fixes (not including some additions to IGlobalAI that arent necessary to function).
the 0.72 interface is also more fault tolerant than the 0.73 interface.
the 0.72 interface is also more fault tolerant than the 0.73 interface.
The problem is that there are so many revisions of the svn between 0.72 and 0.73 that it'd take weeks of compiling and testing to go through them all.
Meanwhile us AI devs tend to work with the current spring no the dev spring, and since most people dont tell us about degradation in performance but rather is it still building or not, its awkward for us when a new release is out and our AIs arent behaving anymore as a result.
However I see the most powerful skirmish AIs being reduced to something that barely issues a command at all proof enough.
Meanwhile us AI devs tend to work with the current spring no the dev spring, and since most people dont tell us about degradation in performance but rather is it still building or not, its awkward for us when a new release is out and our AIs arent behaving anymore as a result.
However I see the most powerful skirmish AIs being reduced to something that barely issues a command at all proof enough.
The point is that it's useless to only complain about it all the time.
You're exagerating again..., AAI works, CSAI works, only KAI doesn't because krog doesn't work on it anymore. One unmaintained AI not working != everything fucked up.However I see the most powerful skirmish AIs being reduced to something that barely issues a command at all proof enough.
err, NTai didnt work and it needed fixing and it got fixed, how I dunno exactly, and the same with AAI.
CSAI was built after the 0.72->0.73 transition.
I just said there were changes in the AI interface that werent noticed that broke AIs and they needed fixing. Most likely lots of different little things...
But your sort of missing the whole point of why i said anything to begin with and turning it into a 'there was no issue with 0.72->0.73' arguement.
The original points where that in order to fix KAI we would need krogoth tournesol or firenu todo it because nobody knows enough to get it going again, or why exactly it doesnt go at all.
However I spy in the sentence you quoted a small spelling mistake,
Thus I suggest that you calm down because you seem to be taking it to heart from where I'm seeing it, and you're overreacting.
CSAI was built after the 0.72->0.73 transition.
I just said there were changes in the AI interface that werent noticed that broke AIs and they needed fixing. Most likely lots of different little things...
But your sort of missing the whole point of why i said anything to begin with and turning it into a 'there was no issue with 0.72->0.73' arguement.
The original points where that in order to fix KAI we would need krogoth tournesol or firenu todo it because nobody knows enough to get it going again, or why exactly it doesnt go at all.
However I spy in the sentence you quoted a small spelling mistake,
Although considering there was a stage during 0.73 development where I could have put the s back on, a stage where I was so infuriated by the AI interface and the lack of progress with NTai and the svn build of spring that I almost gave up altogether.However I see the most powerful skirmish AIs being reduced to something that barely issues a command at all proof enough.
Thus I suggest that you calm down because you seem to be taking it to heart from where I'm seeing it, and you're overreacting.
No, it takes less time if you plan it a bit smart, know what you're looking for, and use scripting (if you don't mind linux bash script, look at this script which can be used to precompile as much revisions of spring as you want, without touching your computer once). I've done a bisection search using binaries generated by this script at least once to find a few sync errors introduced in a range of a few hundred revisions between 0.72 and 0.73... only took 4 hours or so (after another 4 hours of background compilation)AF wrote:The problem is that there are so many revisions of the svn between 0.72 and 0.73 that it'd take weeks of compiling and testing to go through them all.
The fact that KAI wasn't and NTAI isn't in SVN adds some time tho because you need to add code to the script to check out the matching revision from your own SVN and compile that alongside the spring binary.
As JC already said, I see AAI* working fine, which, following the same reasoning, proves that the the interface isn't broken.However I see the most powerful skirmish AIs being reduced to something that barely issues a command at all proof enough.
As an extra note, are you sure it isn't the AIs fault? Does valgrind/purify detect no memory corruption / usage of uninitialized variables at all?
Do you compile using the exact same compiler (version)? (since microsoft is known for breaking C++ ABI compat with every compiler they released...)
*current SVN @ ubuntu edgy eft
-
1v0ry_k1ng
- Posts: 4656
- Joined: 10 Mar 2006, 10:24
