On Mon, 9 May 2005, Cyrus Harmon wrote:
> memory and still do a gc. One interesting tidbit is that it's only the gc
> that causes this to die. What happens on allocation prior to gc? Is this
> memory overwriting the dynamic-1? Should we be failing earlier in such cases?
On non-x86's GC involves copying dynamic-pace-0 to dynamic-space-1 or vice
versa it, so it sounds like only either one is stomping on something.
(Eg. imagine the bad area in the middle of dspace-1: you can safely fill
dspace-0, as the bad-addresses will not be touched before GC, but then
Now that I think of it, someone (Thomas Burdic maybe?) iirc posted
a patch with an extended address range that apparently worked for him
on OS X some while back.
-- Nikodemus Schemer: "Buddha is small, clean, and serious."
Lispnik: "Buddha is big, has hairy armpits, and laughs."