Submitted by: David North

Submitted by: Sander Temme

PR: 1663

/*
 * This release balances code size and performance. In particular key
 * schedule setup is fully unrolled, because doing so *significantly*
 * reduces amount of instructions per setup round and code increase is
 * justifiable. In block functions on the other hand only inner loops
 * are unrolled, as full unroll gives only nominal performance boost,
 * while code size grows 4 or 7 times. Also, unlike previous versions
 * this one "encourages" compiler to keep intermediate variables in
 * registers, which should give better "all round" results, in other
 * words reasonable performance even with not so modern compilers.
 */

C++ style comments.
Signed/unsigned warning in apps.c
Missing targets in jpake/Makefile

with the appropriate parameters which calls OBJ_bsearch(). A compiler will
typically inline this.

This avoids the need for cmp_xxx variables and fixes unchecked const issues
with CHECKED_PTR_OF()

avoid inlines...)

to s_client application.
PR: #1761
Submitted by: David Woodhouse <dwmw2@infradead.org>

Duplicate const.
Use of ; outside function.

Submitted by: Alex Chen <alex_chen@filemaker.com>

condition check.

reimplement BN_nist_mod_521.

but has been omitted from HEAD (0.9.9), see commit
  http://cvs.openssl.org/chngview?cn=16627
by appro.

(Firstly... ommitted)

Secondly, it wasn't even _dropping_ the offending packets, in the
non-blocking case. It was just returning garbage instead.
PR: #1752
Submitted by: David Woodhouse <dwmw2@infradead.org>

offending record queued as 'pending'. The DTLS code doesn't expect this,
and we end up hitting an OPENSSL_assert() in do_dtls1_write().

The simple fix is just _not_ to leave it queued. In DTLS, dropping
packets is perfectly acceptable -- and even preferable. If we wanted a
service with retries and guaranteed delivery, we'd be using TCP.
PR: #1703
Submitted by: David Woodhouse <dwmw2@infradead.org>