-LONGLONG_MIN is the undefined behavior in C.
longlong2decimal() used to do this:

  int longlong2decimal(longlong from, decimal_t *to) {
    if ((to->sign= from < 0))
      return ull2dec(-from, to);
    return ull2dec(from, to);

and later in ull2dec() (DIG_BASE is 1000000000):

  static int ull2dec(ulonglong from, decimal_t *to) {
    for (intg1=1; from >= DIG_BASE; intg1++, from/=DIG_BASE) {}

this breaks in gcc-5 at -O3. Here ull2dec is inlined into
longlong2decimal. And gcc-5 believes that 'from' in the
inlined ull2dec is always a positive integer (indeed, if it was
negative, then -from was used instead). So gcc-5 uses
*signed* comparison with DIG_BASE.

Fix: make a special case for LONGLONG_MIN, don't negate it

MDEV-8045 Assertion `0' fails in Protocol::end_statement on CREATE VIEW after another connection aborted

0 is a valid value for a file descriptor

On EOF vio_read returns 0, it's not an error so the errno
is not reset. If the previous error was EINTR the client
will loop forever. See also man recv.

Merge branch 'openquery:mdev-7781-allow-install-uninstall-plugins-during-bootstrap' into 10.0
Undo MySQL fix for bug#46261

Merge branch 'openquery:mdev-7390-alter-online-table-xx-possible-10.0' into 10.0

Merge branch 'pastcomputer:10.0-oqgraph-6282-6345-6784-test' into 10.0

And don't ignore SSL_CTX_set_tmp_dh() failures

change openssl_1 test not to use non-FIPS ciphers

MDEV-7695 MariaDB - ssl - fips: can not connect with --ssl-cipher=DHE-RSA-AES256-SHA - handshake failure

Change 512bit DH key to 1024bit to meet FIPS requirements

Tell OpenSSL to use MD5 even if FIPS prohibits it.
This is fine as long as we do not use MD5 for cryptographical
purposes (md5 is used internally for P_S message digests and for view
checksums)

MDEV-7697 Client reports ERROR 2006 (MySQL server has gone away) or ERROR 2013 (Lost connection to MySQL server during query) while executing AES* functions under SSL

Clear OpenSSL error queue after an error in AES_ENCRYPT/AES_DECRYPT.
Otherwise it might affect current ssl-encrypted connection.

1. After a period of wait (where last_master_timestamp=0)
   do NOT restore the last_master_timestamp to the timestamp
   of the last executed event (which would mean we've just
   executed it, and we're that much behind the master).

2. Update last_master_timestamp before executing the event,
   not after.

Take the approach from the this commit (but with a different test
case that actually makes sense):

commit 0c75ab453fb8c5439576af8fe5add7a1b89f1569
Author: Luis Soares <luis.soares@sun.com>
Date:   Thu Apr 15 17:39:31 2010 +0100

    BUG#52166: Seconds_Behind_Master spikes after long idle period

which failed main.features test in embedded, because
status variables were not sorted

(to help 'git diff' show the correct function for hunks)

The issue comes from not taking all possibilities to match an entry
within the roles_mapping HASH, when updating the data structure.

MDEV-7038 Assertion `status_var.memory_used == 0' failed in THD::~THD() on disconnect after executing EXPLAIN for multi-table UPDATE

Added test case that caught this bug. It is no longer reproducible in
the current tree.

Problem was that std::vector was allocated using calloc instead of
new, this caused vector constructor not being called and vector
metadata not initialized.