all the symbols that POSIX says they must and fewer that they can't and,
most importantly, to not require a specific ordering of headers.
ports testing by naddy@
ok millert@ deraadt@

``-upper_bound % upper_bound''.  Simplifies the code and makes it the
same on both ILP32 and LP64 architectures, and also slightly faster on
LP64 architectures by using a 32-bit remainder instead of a 64-bit
remainder.
Pointed out by Jorden Verwer on tech@
ok deraadt; no objections from djm or otto

clear out the entire requested area, not just a perfect fit.  second,
use mquery to check for room to avoid getting an address we don't like
and having to send it back.

so that it can't get a signal while still running on the parent thread's
stack.  Also, pass in sizeof(struct __tfork) to provide forward compat
when more members are added.  This is an ABI change, so switch syscall
numbers and bump lib majors this time.
ok deraadt@ matthew@

descriptor starvation, we can crank the default user file descriptor
limits.

in order to span the the entire cache.  second, on free use the same offset
to put things in the cache instead of always starting at zero.
ok otto

- everything that has prereq will also do includes, add it late, so that
gnu stuff (the only part with prereq) happens late.
- new RUN_MAKE snippet, to be able to have the makefile magically decide
between Makefile.bsd-wrapper and Makefile.
So there's no longer any need to remember complicated rules: add it to
RDIRS if it just wants make includes, add it to PRDIRS if it wants prereq too,
and that's all.
okay deraadt@, miod@

overallocating and then releasing unneeded memory pages.
ok otto

ok gilles@

from the most recent version, use the implementation namespace for variables
in inline functions, and add __restrict as specified.
ok millert@

Bulk build test by naddy.

on descriptors, this can be achieved thanks to the new getdtablecount()
system call. application may provide a reserve count to ensure that the
recvmsg() call is not called when they don't have enough descriptors to
work properly.
change the API so that transient errors that can be retried immediately
are retried within the function right away, whereas transient errors for
which the application may want to take action will set errno to EAGAIN.
ok deraadt@ and henning@

preventing them being locally allocated as dynamic ports.
If AMT is configured with the same IP address as the host (and, in
the case of redirection/https ports, if the relevant feature is
enabled), it will steal packets sent to its ports.
ok mikeb@. kettenis@ thinks it's probably a good idea.

Update SYNOPSIS for setkey() to show it's in <stdlib.h>
ok jmc@, millert@

and valloc() are not in the current version, while posix_memalign() mkstemp(),
and mkdtemp() are, and setstate()'s argument has lost a bogus 'const'.
ok millert@ jmc@ espie@ kettenis@; ports build testing by naddy@

These are useful if you're behind a bad nat with short timeouts as often
found in airport lounges and hotels. If the keepalives fail because the
network goes away (moving out of wireless reception, swapping cables etc)
your connections are more likely to drop so this is a double-edged sword.
ok henning@

from sys/unistd.h to unistd.h.
From Brad; ok millert and me

daemon is not active (happens often to me, as I forget which box has
which daemons).
okay ajacoutot@

provide IP22 and IP28 kernels, as well as the ECOFF bootblock.
No objection from deraadt@

as NULL has. This will (hopefully) simplify ports where the user
passes the device string.

ok guenther, millert (and me); bulk build test by naddy

and have been stubbed on openbsd.org anyway ok millert@ robert@ otto@

Okay otto@

* strptime(3) was introduced in XPG 4.0
* clock_gettime(3) and friends were introduced in 1003.1b-1993
* asctime_r(3) and friends were introduced in 1003.1c-1995
and conditionally provide prototypes and associated types accordingly.
This makes our <time.h> standards compliant except for some functions that are
still missing.
ok guenther@

friends in <time.h>.  The kernel needs access to the associated #defines, but
can't get them from <time.h>, so introduce a new header <sys/_time.h> and
include that from <time.h> for userland and <sys/time.h> for the kernel.
ok matthew@, guenther@, millert@

ok jmatthew@

default, similar to the v6 entries.
While here, add a filter for 100.64.0.0/10, which is now reserved by RFC 6598
OK henning@, sthen@

/dev/ptm.  Users are strongly encouraged to upgrade to a more recent
release if they haven't already.
ok deraadt

ok deraadt

(yeah, the md parts are NOT, not yet).
okay miod@, deraadt@

ok miod@

effective result.  its use is NOT discouraged -- it is not
common, but when you need it, there is nothing else that will do.

suggested by deraadt@

to fix and there are several alternatives that don't tell quite so many
dangerous lies.
enthusiastic agreement from a bunch of people

ok deraadt@