🚀 go-pugleaf

In article <1992Sep25.175643.6752@pixel.kodak.com>, dj@ekcolor.ssd.kodak.com (Dave Jones) writes:
. Matthew Fisher (fisherm@infonode.ingr.com) wrote:
. :      I have a 2meg array that I have alloced on the global heap.  I'm 
> : accessing this array with huge pointers.  The array is made up of structures
. : that are 12 bytes long.  My pointer manipulation seems to work fine.
. : However when one of the structures lay across a 64k boundary my assignment
. 
. Since someone asked for a net post, here goes.
. 
. There is no such thing as a 64K boundary.  What there is, is a situation
. where the 16-bit offset of a 32-bit pointer wraps around.  Huge pointer
. arithmetic is supposed to be proof against this because the segment
. pointer is adjusted in pointer arithmetic to minimize the offset.  In
. fact the offset in a "normalized" huge pointer should be no bigger than
. 15 decimal.  The segment pointer is adjusted to the 16-byte segment
. boundary closest to the result address.
. .....
Uh,  WRONG!   In protected mode, you can't do this (the "segment" is a
selector that points to a 64K segment.  MS specifically sez that HUGE
array elements MUST be a power of 2.  So, either pad these to 16 bytes each,
or use a link list of <64K blocks seperately allocated (kinda brutal).

	-jrp