NTai XE10.1b

[AF]

hmm that assertion is there because you should never call GetNextTask if there are no tasks left, and I cant find any calls not already encased in a HasTasks() if statement, is there any way of getting a stack trace?

[hoijui]

hpw is it with NTai AF? is the code in spring trunk your latest?

[AF]

It seems so. That code would crash the next line if there were no assertion anyway

[DJ]

it seems to be as soon as the first unit is destroyed, although that's from obsvervation, i only have the stack trace from info log which i imagine won't work with the stack trace translator for AI's. None the less:-

Code: Select all

[   9189] (0) D:\spring0_77b2\Spring\AI\Bot-libs\NTai.dll [0x6B699ACF]
[   9189] (1) D:\spring0_77b2\Spring\AI\Bot-libs\NTai.dll [0x6B7678B0]
[   9189] (2) D:\spring0_77b2\Spring\AI\Bot-libs\NTai.dll [0x6B6A6C14]
[   9189] (3) D:\spring0_77b2\Spring\AI\Bot-libs\NTai.dll [0x6B6849C1]
[   9189] (4) D:\spring0_77b2\Spring\spring.exe [0x007D26FF]
[   9189] (5) D:\spring0_77b2\Spring\spring.exe [0x0081F1D7]
[   9189] (6) D:\spring0_77b2\Spring\spring.exe [0x0084B27C]
[   9189] (7) D:\spring0_77b2\Spring\spring.exe [0x0085059D]
[   9189] (8) D:\spring0_77b2\Spring\spring.exe [0x005B9C7B]
[   9189] (9) D:\spring0_77b2\Spring\spring.exe [0x005C5FC8]
[   9189] (10) D:\spring0_77b2\Spring\spring.exe [0x005B8B6D]
[   9189] (11) D:\spring0_77b2\Spring\spring.exe [0x005B8D69]
[   9189] (12) D:\spring0_77b2\Spring\spring.exe [0x00A32C01]
[   9189] (13) D:\spring0_77b2\Spring\spring.exe [0x004010A7]
[   9189] (14) D:\spring0_77b2\Spring\spring.exe [0x00401123]
[   9189] (15) C:\WINDOWS\system32\kernel32.dll(RegisterWaitForInputIdle+0x49) [0x7C816FD7]

-- edit --

Tested a further three times, its definitely when a unit gets destroyed. in each test an assault unit was destroyed and the crash occured

i'd guess it was this remove tasks call in the unit destroyed handler causing the problem.

Code: Select all

G->RemoveHandler(currentTask);
taskManager->RemoveAllTasks();
G->RemoveHandler(taskManager);

[AF]

Indeed I don't know how to use that stack trace I need function/method names.

The fact it does it for assault units though is curious I'll have to look into it.

[DJ]

any joy finding this problem?

[AF]

remove the assertion, and instead change it to an if statement. If there are no tasks then return a shared_ptr to 0 or null and check for this in the CUnit.cpp file. This will not fix the issue but it should prevent the crash and allow NTai to carry on running. More of a hack around than a bugfix

[DJ]

had a bit of time today so i've installed a build environment and sorted this error. If anyone wants a binary just post here

Solobuild doesn't seem to work anymore which hampers NTai alot

[chadj]

If anyone has a 0.77b5 build of NTAII.dll let me know, I'd like to get my hands on it!

-Chad Johnson

chad.j.johnson@gmail.com

[Satirik]

If anyone has a 0.77b5 build of NTAII.dll let me know, I'd like to get my hands on it!

-Chad Johnson

chad.j.johnson@gmail.com

viewtopic.php?f=15&t=16503

[bibim]

Here is the BuildServ's build of NTai for 0.77b5. This binary has been built during the Spring 0.77b5 installer generation, so it should be possible to translate the stacktraces using the !translate command.

[AF]

A: I did not commit a bug fix for the issue DJ had although I describe the necessary steps. I do not have any machines with an svn client and repo set up to apply the fix at the moment.
B: Satirik, please do not post in my threads, remember, you've been warned by moderators, and I have abided by the same rule that applies to you.

[DJ]

There are a number of problems with NTai at the moment.

1) A crash bug that occurs some 25 minutes into the game or so, I've tried building a version with logging enabled but it doesn't give any information at all on the crash location. This can only really be fixed with someone who has either a working debugging environment or alot of time to insert insane amounts of logging code.

2) NTai now seems to have a problem reading its config files from any class other than Global. I've confirmed this bug but i've yet to find a reason for it, doesn't help that I suck at C++

3) Singlebuild/Solobuild no longer work, now this could easily be linked to bug 2 but considering bugs 1 and 2 this isn't the major show stopper.

Without these bugs being fixed (by someone who knows what they're doing I might add) you can pretty much consider this AI dead.

[hoijui]

i do have a working debug environment.
..well: spring set up to build in debug mode, plus i a debugger :D
if it is a simpe task, i could try ot find the location of the crashes, but i do not want to spend more then 30 mins on this.
if you give me the NTai config files plus other stuff i may need, plus some instructions, i may help.

[AF]

hmm can you describe #2 but rephrase, I'm not quite sure what you mean, though my interpretation is that the config loading class only works when called from CGlobal which seems odd. I'll have to see what I can do tonight, but I will have to setup spring in visual studio and it will take a while as I have to sort out my svn and do a whole checkout and dependencies etc etc.

#3 can easily be attributed to #2 though.

Is #1 related to the issue which you reported the other day? I have yet to get around to applying a fix for that.



Hoijui, one thought that would eb very useful is that currently AIs only have the console to route messages through. If NTai has no configs because the user mistakenly thinks they only need a DLL I want a warning box to appear that says so very clearly, but chat console messages are ignored or they get pushed out of view by more messages....

[DJ]

AF I fixed the unit destroyed crash bug before I started testing the rest, this is a different problem.

#2 you're assumption as to what I mean by this is correct, to show it in code

Code: Select all

string sb = G->Get_mod_tdf()->SGetValueMSG("AI\\attackers");

gives an empty string. There is also a subtle problem that maybe linked to this; the config names that ntai creates have changed. The ba config is now BAXX.sd7.tdf whereas it used to be BA.tdf for all versions of BA....

[AF]

hmm how does this relate to the modinfo.tdf in the BA SDZ?

[hoijui]

Hoijui, one thought that would eb very useful is that currently AIs only have the console to route messages through. If NTai has no configs because the user mistakenly thinks they only need a DLL I want a warning box to appear that says so very clearly, but chat console messages are ignored or they get pushed out of view by more messages....

in other words, you want a callback function like this:

Code: Select all

class IAICallback {
   void spwnErrorMessage(std::string title, std::string msg);
}

?

[DJ]

hmm how does this relate to the modinfo.tdf in the BA SDZ?

I don't think the modinfo is the deciding factor. I tested with BA 4.7 which is the exact same file i used in previous spring versions. Previously I got the BA.tdf file created as usual, in the latest version i get BA47.sd7.tdf

It's not a problem aside from each version needing a new config i just thought it might be related.

[AF]

Some means of notifying the user and making them aware that things are seriously wrong. My thoughts are it would likely require a lua widget bundled with spring.

So basically its using <archivename>.tdf ? On the offchance, is it using a modinfo.tdf or a modinfo.lua?