Add support for getrandom() or equivalent system calls and use them by default

Reviewed-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
GH: #5910

Fix building linux-armv4 with --strict-warnings

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6026)

Ensure the thread keys are always allocated in the same order

Fixes: #5899
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5911)

openssl/err.h: remove duplicate OSSL_STOREerr()

Two definitions in lines 127 and 136, introduced in 71a5516dcc8.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6029)

openssl/ssl.h: restore some renamed public SSL_CTRL defines

Fixes #6022

In commit de4d764e3271, the following SSL_CTRL #define's where renamed

    SSL_CTRL_GET_CURVES        ->  SSL_CTRL_GET_GROUPS
    SSL_CTRL_SET_CURVES        ->  SSL_CTRL_SET_GROUPS
    SSL_CTRL_SET_CURVES_LIST   ->  SSL_CTRL_SET_GROUPS_LIST
    SSL_CTRL_GET_SHARED_CURVE  ->  SSL_CTRL_GET_SHARED_GROUP

The corresponding function-like macros (e.g, SSL_get1_curves(ctx, s)) were
renamed, too, and compatibility #define's were added. This was overlooked for
the above constants. Since the constants are part of the public interface,
they must not be removed for a minor release.

As a consequence the Qt5 configure check (and the build) fails.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6023)

Add a test for SSL_pending()

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6020)

Fix SSL_pending() for DTLS

DTLS was not correctly returning the number of pending bytes left in
a call to SSL_pending(). This makes the detection of truncated packets
almost impossible.

Fixes #5478

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6020)

Test the state of SSL_in_init() from the info_callback

Check that in a handshake done event SSL_in_init() is 0 (see #4574)

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6019)

Make sure SSL_in_init() returns 0 at SSL_CB_HANDSHAKE_DONE

In 1.1.0 and before calling SSL_in_init() from the info_callback
at SSL_CB_HANDSHAKE_DONE would return 0. This commit fixes it so
that it does again for 1.1.1. This broke Node.

Fixes #4574

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6019)

Document supported digest functions

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6024)

Enable all implemented digests

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6025)

Clear buffer in PEM_write_bio

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5814)

Reflect special `DEFAULT` behavior in ciphers(1)

Actual behavior of DEFAULT is different than currently described.
Rather than actinf as cipher string, DEFAULT cannot be combined using
logical operators, etc.

Fixes #5420.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5428)

Don't distribute team internal config targets

Configurations/90-team.conf isn't for public consumption, so we rename
it to 90-team.norelease.conf and make sure 'make dist' and 'make tar'
don't include it in the tarball.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5836)

correct spelling errors detected by Debian lintian

CLA: trivial

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5801)

Add a test for a NULL X509_STORE in X509_STORE_CTX_init

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6001)

Don't crash if there are no trusted certs

The X509_STORE_CTX_init() docs explicitly allow a NULL parameter for the
X509_STORE. Therefore we shouldn't crash if we subsequently call
X509_verify_cert() and no X509_STORE has been set.

Fixes #2462

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6001)

Fix ocsp app exit code

If we run the ocsp command line app and the responder returns a
non-successful status code then the app should exit with a failure code.

Based on an original patch by Tatsuhiro Tsujikawa.

Fixes #2387

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5998)

Fix no-ec

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5997)

Correct an ommission in the EVP_DigestSignInit docs

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5996)

Add missing index_index() when reloading OCSP responder

Also, future-proof index_index() return codes by requiring success
to return a positive value.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Clarify the configuration module in config.pod

Similar to 0652e8a7 ("Clarify default section in config.pod",
2018-04-12), reword a sentence to make it easier to parse.

Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5794)

apps/s_socket.c: fix memory sanitizer problem in ACCEPT printout.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5994)

TLSProxy/Proxy.pm: preclude output intermix.

s_server -rev emits info output on stderr, i.e. unbufferred, which
risks intermixing with output from TLSProxy itself on non-line
boundaries, which in turn is confusing to TAP parser.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5975)

recipes/70-test_ssl{cbcpadding,extension,records}: make it work w/fragmentation.

This fixes only those tests that were failing when network data was
fragmented. Remaining ones might succeed for "wrong reasons". Bunch
of tests have to fail to be considered successful and when data is
fragmented they might fail for reasons other than originally intended.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5975)

TLSProxy/Record.pm: add is_fatal_alert method.

(resolve uninitialized variable warning and harmonize output).

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5975)

TLSProxy/Proxy.pm: refine NewSessionTicket detection.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5975)

TLSProxy/Message.pm: refine end-of-conversation detection logic.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5975)

poly1305/asm/poly1305-armv4.pl: remove unintentional relocation.

Branch to global symbol results in reference to PLT, and when compiling
for THUMB-2 - in a R_ARM_THM_JUMP19 relocation. Some linkers don't
support this relocation (ld.gold), while others can end up truncating
the relocation to fit (ld.bfd).

Convert this branch through PLT into a direct branch that the assembler
can resolve locally.

See https://github.com/android-ndk/ndk/issues/337 for background.

The current workaround is to disable poly1305 optimization assembly,
which is not optimal and can be reverted after this patch:
https://github.com/freedesktop/gstreamer-cerbero/commit/beab607d2b1ff23c41b7e01aa9c64be5e247d1e6

CLA: trivial

Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5949)

Style: ssl.h

fix some indents, and restrict to 80 cols some lines.

Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4466)

Update EVP_DigestSignInit() docs

Explicitly state which digests can be used with which algorithms.

Fixes #5854

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5992)

test: Remove redundant SSL_CTX_set_max_early_data

Client can only send early data if the PSK allows for it, the
max_early_data_size field can only be configured for the server side.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5702)

Add support for logging early exporter secret

This will be necessary to enable Wireshark to decrypt QUIC 0-RTT data.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5702)

Add test for CLIENT_EARLY_TRAFFIC_SECRET key logging

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5702)

Add support for logging TLS 1.3 exporter secret

NSS 3.34 and boringssl have support for "EXPORTER_SECRET"
(https://bugzilla.mozilla.org/show_bug.cgi?id=1287711) which is needed
for QUIC 1-RTT decryption support in Wireshark.

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5702)

BIGNUM signed add/sub routines refactory

Old code replaced in favor of a clearer implementation.
Performances are not penalized.

Updated the copyright end date to 2018.

Reviewed-by: David Benjamin <davidben@google.com>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5963)

Fix a memory leak in an error path

Found by Coverity.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5970)

Check the return from EVP_PKEY_get0_DH()

Fixes #5934

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5983)

Extend the SSL_set_bio() tests

The SSL_set_bio() tests only did standalone testing without being in the
context of an actual connection. We extend this to do additional tests
following a successful or failed connection attempt. This would have
caught the issue fixed in the previous commit.

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5966)

Fix assertion failure in SSL_set_bio()

If SSL_set_bio() is called with a NULL wbio after a failed connection then
this can trigger an assertion failure. This should be valid behaviour and
the assertion is in fact invalid and can simply be removed.

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5966)

Update fingerprints.txt

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5987)

Update the info callback documentation for TLSv1.3

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5874)

Add a test for the info callback

Make sure the info callback gets called in all the places we expect it to.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5874)

Make sure info callback knows about all handshake start events

The first session ticket sent by the server is actually tacked onto the
end of the first handshake from a state machine perspective. However in
reality this is a post-handshake message, and should be preceeded by a
handshake start event from an info callback perspective.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5874)

Call the info callback on all handshake done events

Fixes #5721

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5874)

Ignore the status_request extension in a resumption handshake

We cannot provide a certificate status on a resumption so we should
ignore this extension in that case.

Fixes #1662

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/5896)

SSL_CTX_set_tlsext_ticket_key_cb.pod: fix error check of RAND_bytes() call

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5977)

p5_scrypt.c: fix error check of RAND_bytes() call

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/5977)

DRBG: fix coverity issues

- drbg_lib.c: Silence coverity warning: the comment preceding the
  RAND_DRBG_instantiate() call explicitely states that the error
  is ignored and explains the reason why.

- drbgtest: Add checks for the return values of RAND_bytes() and
  RAND_priv_bytes() to run_multi_thread_test().

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5976)

apps/s_socket.c: print only dynamically allocated port in do_server.

For formal backward compatibility print original "ACCEPT" message for
fixed port and "ACCEPT host:port" for dynamically allocated.

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5956)

Add a config option to disable automatic config loading

./config no-autoload-config

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5959)

Prepare for 1.1.1-pre6-dev

Reviewed-by: Matt Caswell <matt@openssl.org>

Prepare for 1.1.1-pre5 release

Reviewed-by: Matt Caswell <matt@openssl.org>

Update copyright year

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5990)

OpenSSL 1.1.1-pre5: update CHANGES with recent user visible changes

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5989)

Revert "Add OPENSSL_VERSION_AT_LEAST"

Fixes #5961

This reverts commit 3c5a61dd0f9d9a9eac098419bcaf47d1c296ca81.

The macros OPENSSL_MAKE_VERSION() and OPENSSL_VERSION_AT_LEAST() contain
errors and don't work as designed. Apart from that, their introduction
should be held back until a decision has been mad about the future
versioning scheme.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5968)

Remove mandatory generated files on VMS too

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5958)

Remove mandatory generated files on windows too

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5958)

RSA key generation: ensure BN_mod_inverse and BN_mod_exp_mont both get called with BN_FLG_CONSTTIME flag set.

Based on an original patch by Billy Brumley

CVE-2018-0737

Reviewed-by: Rich Salz <rsalz@openssl.org>

Remove mandatory generated files too

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5951)

Fix cygwin make dependencies

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5951)

TLSProxy/Proxy.pm: straighten inner loop termination logic.

Original condition was susceptible to race condition...

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5933)

TLSProxy/Proxy.pm: bind s_server to loopback interface.

Bind even test/ssltest_old.c to loopback interface. This allows to avoid
unnecessary alerts from Windows and Mac OS X firewalls.

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5933)

openssl#5668: corrections after compiling with -qinfo=all:als.

The ongoing discussion about casting or not in PR #5626 had me compiling
again with above mentioned flags. Indeed the compiler had to say something
about it and I did these changes to silence it again.

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5943)

make update

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5948)

DRBG: implement a get_nonce() callback

Fixes #5849

In pull request #5503 a fallback was added which adds a random nonce of
security_strength/2 bits if no nonce callback is provided. This change raised
the entropy requirements form 256 to 384 bit, which can cause problems on some
platforms (e.g. VMS, see issue #5849).

The requirements for the nonce are given in section 8.6.7 of NIST SP 800-90Ar1:

  A nonce may be required in the construction of a seed during instantiation
  in order to provide a security cushion to block certain attacks.
  The nonce shall be either:

  a) A value with at least (security_strength/2) bits of entropy, or

  b) A value that is expected to repeat no more often than a
     (security_strength/2)-bit random string would be expected to repeat.

  Each nonce shall be unique to the cryptographic module in which instantiation
  is performed, but need not be secret. When used, the nonce shall be considered
  to be a critical security parameter.

This commit implements a nonce of type b) in order to lower the entropy
requirements during instantiation back to 256 bits.

The formulation "shall be unique to the cryptographic module" above implies
that the nonce needs to be unique among (with high probability) among all
DRBG instances in "space" and "time". We try to achieve this goal by creating a
nonce of the following form

    nonce = app-specific-data || high-resolution-utc-timestamp || counter

Where || denotes concatenation. The application specific data can be something
like the process or group id of the application. A utc timestamp is used because
it increases monotonically, provided the system time is synchronized. This approach
may not be perfect yet for a FIPS evaluation, but it should be good enough for the
moment.

This commit also harmonizes the implementation of the get_nonce() and the
get_additional_data() callbacks and moves the platform specific parts from
rand_lib.c into rand_unix.c, rand_win.c, and rand_vms.c.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5920)

Rework partial packet handling once more

Address the concern that commit c53c2fec raised differently.

The original direction of the traffic is encoded in bit 0
of the flight number.

Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5923)

test/recipes/test_genrsa.t : don't fail because of size limit changes

There is a test to check that 'genrsa' doesn't accept absurdly low
number of bits.  Apart from that, this test is designed to check the
working functionality of 'openssl genrsa', so instead of having a hard
coded lower limit on the size key, let's figure out what it is.

Partially fixes #5751

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/5754)

(cherry picked from commit ec46830f8a4ce62c0c8ee7677b1eb8e53ee16df1)

Split the scrypt and RSA-PSS into man3 and man7 pages

The scrypt and RSA-PSS documents were a mixture of section 3 and
section 7 material.  With pre-1.1.1 OpenSSL, this is understandable,
since we had a different directory layout.  With 1.1.1, we've moved to
the typical man-page directory layout, and the documents need to be
updated accordingly.

Also, the scrypt document contained a description of
EVP_PKEY_CTX_set1_pbe_pass(), which is a generic function rather than
an scrypt specific function, and therefore should be documented
separately.

Fixes #5802

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5942)

Add a test for SRP

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5925)

Add support for the SRP base64 alphabet

Historically we used to implement standalone base64 code for SRP. This
was replaced by commit 3d3f21aa with the standard base64 processing code.

However, the SRP base64 code was designed to be compatible with other SRP
libraries (notably libsrp, but also others) that use a variant of standard
base64. Specifically a different alphabet is used and no padding '='
characters are used. Instead 0 padding is added to the front of the string.
By changing to standard base64 we change the behaviour of the API which may
impact interoperability. It also means that SRP verifier files created prior
to 1.1.1 would not be readable in 1.1.1 and vice versa.

Instead we expand our standard base64 processing with the capability to be
able to read and generate the SRP base64 variant.

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5925)

Change SRP functions to use EVP_EncodeUpdate/EVP_DecodeUpdate functions

Previously they were using EVP_EncodeBlock/EVP_DecodeBlock. These are low
level functions that do not handle padding characters. This was causing
the SRP code to fail. One side effect of using EVP_EncodeUpdate is that
it inserts newlines which is not what we need in SRP so we add a flag to
avoid that.

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5925)

Docs for OpenSSL_init_crypto: there is no way to specify another file

The documentation erroneously stated that one can change the default
configuration file name.

Fixes #5939

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5941)

TLSProxy/Proxy.pm: handle "impossible" failure to connect to s_server.

The failure is "impossible", because we have confirmation that s_server
listens, yet Mac OS X fails to connect. This avoids 10 minutes timeout
on Travis CI.

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5907)

TLSProxy/Proxy.pm: handle -1 as return value from waitpid.

On rare occasion 's_server | perl -ne print' can complete before
corresponding waitpid, which on Windows can results in -1 return
value. This is not an error, don't treat it like one. Collect
even return value from s_server.

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5907)

openssl rehash: use libcrypto variables for default dir

X509_get_default_cert_dir_env() returns the default environment
variable to check for certificate directories.
X509_get_default_cert_dir() returns the default configured certificate
directory.

Use these instead of hard coding our own values, and thereby be more
integrated with the rest of OpenSSL.

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5937)

openssl rehash: document -compat

Fixes #5902

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5937)

Configurations/*.tmpl: refine build_all_generated.

Purpose of build_all_generated is to execute all the rules that require
perl, so that one can copy the tree to system with compiler but without
perl. This commit removes last dependencies on perl.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5929)

Clarify default section in config.pod

This is a minor update which hopefully makes these particular lines
read a little easier.

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5938)

appveyor.yml: exercise build_all_generated.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5930)

.travis.yml: exercise build_all_generated

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5930)

Configuration: Simplify generating list of generated files in build file templates

Computing the value of the GENERATED variable in the build file
templates is somewhat overcomplicated, and because of possible
duplication errors, changes are potentially error prone.

Looking more closely at how this list is determined, it can be
observed that the exact list of files to check is consistently
available in all the values found in the %unified_info tables
'depends', 'sources' and 'shared_sources', and all that's needed is to
filter those values so only those present as keys in the 'generate'
table are left.

This computation is also common for all build files, so due to its
apparent complexity, we move it to common0.tmpl, with the result left
in a global variable (@generated), to be consumed by all build file
templates.

common0.tmpl is included among the files to process when creating
build files, but unlike common.tmpl, it comes first of all.

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5930)

EVP_MD_CTX_cleanup replaced with EVP_MD_CTX_reset

The EVP_MD_CTX_cleanup() function was merged into EVP_MD_CTX_reset()
which is called by EVP_MD_CTX_free().  Adjust the documentation to say
that the latter should be used to avoid leaking memory.

CLA: trivial

Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5921)

Fix minor typos in Configurations/README

Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5774)

Minor corrections for the RAND_DRBG API documentation

- added some explaining text to a sentence that lost its context.
- removed mention of per-ssl drbg
- fix whitespace errors

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5804)

OpenSSL::Test: add data_dir()

For test recipes that want to use the directory of the data directory
or a subdirectory thereof, rather than just individual files.

Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5928)

DRBG: fix memory leak on error in rand_drbg_get_entropy()

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5918)

bio/b_addr.c: resolve HP-UX compiler warnings.

The warning reads "[cast] may cause misaligned access". Even though
this can be application-supplied pointer, misaligned access shouldn't
happen, because structure type is "encoded" into data itself, and
application would customarily pass correctly aligned pointer. But
there is no harm in resolving the warning...

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5894)

Configurations/10-main.conf: further HP-UX cleanups/unifications.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Correct the equation for Y' in the comment of point_double function

CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5908)

Fix false positives of IS_*() macros for 8-bit ASCII characters

Fixes #5778, #5840

The various IS_*() macros did not work correctly for 8-bit ASCII
characters with the high bit set, because the CVT(a) preprocessor
macro and'ed the given ASCII value with 0x7F, effectively folding
the high value range 128-255 over the low value range 0-127.
As a consequence, some of the IS_*() erroneously returned TRUE.

This commit fixes the issue by adding range checks instead of
cutting off high order bits using a mask. In order avoid multiple
evaluation of macro arguments, most of the implementation was moved
from macros into a static function is_keytype().

Thanks to Румен Петров for reporting and analyzing the UTF-8 parsing
issue #5840.

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5903)

Fix the build_all_generated rule to include generated .map, .def and .opt files

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5900)

TLSProxy/Record.pm: remove dead condition and improve readability.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5887)

TLSProxy/Proxy.pm: refine partial packet handling.

Original logic was "if no records found *or* last one is truncated, then
leave complete records in queue." Trouble is that if we don't pass on
complete records and get complete packet in opposite direction, then
queued records will go back to sender. In other words complete records
should always be passed on. [Possible alternative would be to match
direction in reconstruct_record.]

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5887)

apps/{s_client.c|s_socket}.c: omit usleep calls.

Even though removed calls were oiriginally added on Windows, problem
they tried to mitigate is not Windows-specific.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5887)

apps/s_socket.c: disable the Nagle algorithm.

Without TCP_NODELAY alerts risk to be dropped between shutdown and close.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5887)

TLSProxy/Proxy.pm: harmonize inner loop with the way sockets are.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5887)

Configurations/10-main.conf: clean up HP-UX targets and add magic macros.

HP-UX provides sockets symbols with incompatible prototypes under same
name. This caused problems in 64-bit builds. Additional macros force
unambiguous symbols with unambiguous prototypes.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5742)

config: fix hpux64-parisc2-gcc detection.

hpux64-parisc2-gcc is chosen based on gcc's bitness, and it was overriden
unconditionally.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5742)

test/asn1_time_test.c: make it work on 64-bit HP-UX.

HP-UX gmtime fails with ERANGE past 19011213204552Z, so skip some tests.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5742)