Re: [Flightgear-bugs] Issue 424 in flightgear-bugs: AI objects in scenarios re-loading far too often without unloading, cmdarg().getPath() becoming corrupted.

[<fli...@go...>]

Comment #2 on issue 424 by bre...@gm...: AI objects in scenarios  
re-loading far too often without unloading, cmdarg().getPath() becoming  
corrupted.
http://code.google.com/p/flightgear-bugs/issues/detail?id=424

Upon further investigation, this happens when using main aircraft that have  
submodel weapons (machine guns, etc) that put out a high number of  
submodels, and when using AI scenarios or playing over MP.

The problem seems to happen when the main aircraft's weapon (machine gun,  
typically) has fired, and so has created numerous (dozens, at least)  
ballistic submodels.

At this point, any of the AI or MP models that are even momentarily out of  
sight, seem to be dropped from the models cache, even if they are still  
very close to the main aircraft and/or just out of it's view.

When they reappear, either because the main aircraft has turned or just  
viewing in a different direction, the model is first invisible and then  
there is a major pause in framerate as the model re-loads, and then it  
becomes visible.

This is annoying/non-ideal behavior and it seems that it would be better to  
prioritize AI and MP models above ballistic models in the model caching  
algorithm, especially since ballistic models and usually small and  
essentially invisible at a distance where the AI and MP model might be  
plainly visible.

However, as mentioned in the main bug report above, the main problem is  
that when the model re-loads to become visible again, it re-runs any Nasal  
code in the <load> section of its XML file.  Since the <unload> nasal code  
has never been called this results often in duplicate listeners, loops, and  
other similar problems.

The end result is often massive slowdown in the framerate after this  
happens a few times.