[KinoSearch] Orderly global destruction

[Marvin Humphrey]

On Feb 24, 2008, at 9:20 PM, Nathan Kurz wrote:

> I guess my instinct would be to apply the patch for the slight benefit
> in directness and performance, but I agree that the benefit is
> minimal.   Still, I think it's a benefit.

OK, fair enough -- I'm swayed.  It was tough to explain the "destroy  
the object in order to save it" code flow.  This is more coherent.

The XS isn't any harder to understand than the other XS functions --  
even the documented ones, you wind up having to look up all the time.   
GV_AMupdate etc. are now documented in some expanded comments, and the  
full rationale is preserved in these mailing list archives.

Committed as r3064.

Marvin Humphrey
Rectangular Research
http://www.rectangular.com/