insertion sort (when the number of elements is < 7).

Previously they would be swapped a byte at a time when sizeof(int)
!= sizeof(long).  Idea from FreeBSD.

exceeds 2 lg N and add a reference to the introsort paper.

when the recursion depth reaches 2*lg(n + 1).  This avoids quicksort's
quadratic behavior for pathological input without appreciably
changing the average run time.

side of the array being partitioned to save on stack space.  Greater
savings can be gained by choosing recursion for the smaller side
of the partition and eliminating recursion for the larger side.
This also results in a small but measurable performance gain.
OK otto@ schwarze@

From "fenderq" on freenode via tj@

- document posix_memalign() does not play nice with reacallocarray(3) and
freezero(3)

size if canaries are enabled. In that case we have the exact requested
size of the allocation.  But we can at least check the given size
against the chunk size if C is not enabled. Plus add some braces
so my brain doesn't have to scan for dangling else problems when I
see this code.

and ok jeremy@

misaligned memory accesses with armv7 ramdisk -Os bsd.rd ping
ok florian millert

version uses the two-way string matching algorithm and is faster
than the old implementation.  With this change, ports that check
for strstr having linear complexity time strstr will no longer
replace the libc strstr with a private version.
OK deraadt@ espie@

no longer has access to the content of a memmory object. It does
this by either clearing (if the object memory remains cached) or
by calling munmap(2). ok millert@, deraadt@, guenther@

ok deraadt@

detected wrt recallocarray()

ok jmc@ deraadt@

from David CARLIER

never occur

- use internal meta-data to do more consistency checking (especially with
option C)
- use cheap free if possible
ok deraadt@

make the behaviour -> use case connection.
help from jmc and jsing

help and ok from tom@ and deraadt@

with the added feature that released memory is cleared. Much input from various
developers. ok deraadt@ tom@

minek van on misc@. Thanks.
ok tb@, tj@
(and jmc@ is ok with the diff but can't ok the technical content).

a long sentence (from tj@).

free() error path.
ok otto@

a page.  This is not required by any standard and other malloc
implementation do not document (or implement) this. ok deraadt@

half a page and a page. ok jmatthew@ tb@

allocation to the size of the new allocation (instead of the requested size).
2. Previously realloc takes the easy way and always reallocates if C is
active. This commit fixes by carefully updating the recorded requested
size in all cases, and writing the canary bytes in the proper location
after reallocating.
3. Introduce defines to test if MALLOC_MOVE should be done and to
compute the new value.

This is a remnant from the original 4.4BSD code that had 'a' as
void * in the function args.  No binary change.  OK bluhm@

issue reported by scott cheloha
ok otto

been the default for ages, and I see no valid reason to be able to
disable it. ok natano@

quickly. In both cases it does not make sense to set hints on them.
So remove that option, which is just a remainder of old times when
malloc used to hold on to pages. ok stefan@

- redundant cast is redundant