[sldev] Re: Reviewing/Fixing the Particle System

[Argent Stonecutter]

Problem #1, in LLViewerPartGroup:


	mRegionp = gWorldPointer->getRegionFromPosAgent(center_agent);
	llassert_always(center_agent.isFinite());
	
	if (!mRegionp)
	{
		//llwarns << "No region at position, using agent region!" << llendl;
		mRegionp = gAgent.getRegion();
	}

This seems to be there to allow particles to be generated by objects  
off-sim. This is useful: you can put user cloud generators off in the  
void this way. It's not a problem, itself, but I thought "what if  
this was changed to check if the particles were out of draw  
distance". So I wondered if there was a mechanism to remove particle  
sources that are out of draw distance.

Trying to track this down, it looks like the only place that the  
particle sources get cleaned up is when a region object is destroyed.

The region object is destroyed when the client disconnects from the  
simulator, in removeRegion(host).

When does the client disconnect from a simulator?

Can a client be connected to a simulator that is out of draw distance?

In addition, there's checks in the code to see if the region that is  
being examined is connected to a defunct host, which implies that  
there can be regions associated with hosts that have been restarted  
in the world list... and if so, these regions would never be removed.

My suggestion is to try some checks on the region, and see if you can  
find evidence of a "region leak", or particle systems associated with  
regions that are out of draw distance.