Re: GSIMapUnlinkNodeFromMap doesn't scale

[Richard Frith-Macdonald]

On Tuesday, May 28, 2002, at 02:44 AM, James Knight wrote:

> I suspect the reason someone put that in however many years ago was in 
> a misguided attempt to guarantee ordering of elements. (IE: new 
> elements always appear first when enumerating). This behavior is not 
> necessary - if the user wants to guarantee ordering, he should keep a 
> seperate array.

I think you are probably right ... and when I recoded the original 
functions as the GSI macros and inline functions,
I just assumed this was a necessary feature.

I have applied your patches, with a minor fix to a couple of the calls 
to the macros to clear nodes for gc.
I've also tidied up the base library code to end enumerations properly, 
and made the functions that do that
clear the enumerator so that, on a GC system, memory can be released 
after the enumeration (though I'm sure
that hash/map tables need quite a bit of work to get GC memory handling 
right).