Bug in malloc under MSC v5.1?

[Bruce Gotwols]

Under Microsoft-C in the large memory model, I have observed what appears to be
a bug in the malloc function.  If I allocate an array that extends to the end of
the physical 64k page, then I would expect the next call to malloc to return a 
pointer which starts at the beginning of the next physical 64k page (but 
because of housekeeping it will be a few bytes higher in memory than this).
This is exactly what happens when the array size is below about 50000 bytes.  I 
can go along merrily "mallocing" until I run out of memory.  However, for an 
array size somewhat greater than 50000 (say 52000) the next pointer returned 
will not start anywhere near the beginning of the next page.  In fact it may
even start halfway into the page.  I cannot see any reason for this behavior
since there is nothing I am aware of that should push the pointer higher in 
memory so that it won't be contiguous to the last block of memory malloced.

Has anyone seen this behavior before, and is it a bug, or is there a rational
explanation?  This is an extremely important question on PC/AT's where direct
memory access requires that an array not cross a physical 64k page boundary.
			Thanks,  Bruce
--
--
Bruce L. Gotwols
Johns Hopkins University, Applied Physics Lab., Laurel MD 20723
Internet:  gotwols at warper.jhuapl.edu   (128.244.176.48)