compiles asking for too much memory understood

[Stephen Weeks]

> Ok,  I  found  the  reason for the problems with MLton using too much memory.
> The problem is caused by the fact that GC_loadWorld()  sets  the  maxHeapSize
> parameter in the GC_state to what ever it finds in the heap.  

Good find!  Are we all in agreement that GC_saveWorld should not save
maxHeapSize, and that instead maxHeapSize should be set from GC_loadWorld
exactly as it is in GC_init?

> Any way this clearly has to be fixed.  I vaguely recall some reason  why  you
> (Stephen)  had some worry about trying to load the heap into less memory than
> when you saved it.  Do you remember what the worry was?

Nope.  I looked through all my old mail and couldn't find anything.

> My notion is to change things so that the normal heap parameters are  obeyed,
> and if the saved heap doesn't fit, fail right away.

Makes sense.