The function make_join_statistics checks whether eq_ref access uses only
constant expressions, and, if this is the case the function performs
constant row substitution. The code of this check must take into account
hidden components of extended secondary keys. 

Since event types can be >=128 and are read from a (possibly signed) char
pointer, we need to cast to unsigned char before extending to int, or we will
get an incorrect negative number. This was done in the main code path already,
but there is a rare case where we check for new events first without a lock
and then again with the lock. If the second check succeeds because a new event
turns up at just the right time, then we took a code path that was missing the
correct unsigned char cast, leading to incorrect handling of events for old
slave servers and possibly other grief.

(This was found from a sporadic failure in Buildbot of test case
rpl_mariadb_slave_capability).

The problem was that xtradb has innodb_purge_threads default 1 (plain
innodb defaults to 0).

The test sets a special debug variable and relies on it to force
purge to happen. But when using background purge threads, this does
not work, the debug code is not made to handle this, so occasionally
the test times out waiting for the purge to occur.

Fix by explicitly setting innodb_purgee_threads=0 for this test.

Change default for innodb_use_fallocate to FALSE, due to bugs in older Linux kernels (posix_fallocate() does not always guarantee that file size is like one specified)

with a bit of lame protection against abuse.

The official Debian Wheezy MySQL packages have versions like 5.5.30+dfsg-xxx.
Such version is larger than 5.5.30-yyy, so apt prefers it.
So use instead 5.5.30+maria-yyy, which is larger and can be pulled in
automatically by apt.

Also included are a couple of fixes for test failures in buildbot.

This fixes that by default LOAD DATA INFILE will not generate the error:
"Multi-statement transaction required more than 'max_binlog_cache_size' bytes of storage..."


mysql-test/suite/sys_vars/r/max_binlog_cache_size_basic.result:
  Updated test case
mysql-test/suite/sys_vars/r/max_binlog_stmt_cache_size_basic.result:
  Updated test case
sql/sys_vars.cc:
  Increase default value of max_binlog_cache_size and max_binlog_stmt_cache_size to ulonglong_max.

(because only there it  makes any sense)

  
MySQL bug http://bugs.mysql.com/bug.php?id=61713 was fixed in 5.5
  
Fix is to remove check for multiple entries returned by getaddrinfo(), and use the first entry that works  - i.e socket can be created.  

Unlike Oracle/MySQL's fix ,this one  is kept minimal : 
-  we do not prioritize IPv4 over IPv6,  orr other way around,  and just rely on operating system to sort getaddrinfo() entries in sensible order. There is RFC that defines  what is sensible order for getaddrinfo entries ( RFC 3484), and OS specific tweaks are also possible , like /etc/gai.conf o Linux.
-  also,  we do not force "0.0.0.0" address if bind-address is not given -  this would be a change in behavior of 5.5 at least on Windows, where passing NULL as  to getaddrinfo()  gives back IPv6-wildcard.

fixes for gcc 4.8 - compilation warnings and -fsanitize=address

fixes for gcc 4.8 -fsanitize=address

fulltext search was initialized for all MATCH ... AGAINST items
at the end of the JOIN::optimize(). But since 5.3 derived tables
are initialized lazily on first use, very late in the sub_select().

Skip Item_func_match::init_search initialization if the corresponding
table isn't open yet; repeat fulltext initialization for all
not-yet-initialized MATCH ... AGAINST items after creating derived tables.

currently get_mm_tree skipped the evaluation of this constant
and icorrectly proceeded. The correct behavior is to return a
NULL subtree, according to the IF branch being fixed - when it
evaluates the constant it returns a value, and doesn't continue
further.

(debug_sync value can have single quotes)

update 5.1 to replicate from 10.0 and
to show the server version (as of 10.0) correctly

sql-common/client.c:
  mdev:4088
sql/slave.cc:
  use the version number, not just the first character of the version string
  (we want 10 > 4 not "10" < "4").

init join->top_join_tab_count to be in sync with join->join_tab=stat,
otherwise a query can be killed in-between and join_tab's won't be deleted
(JOIN::cleanup won't call JOIN_TAB::cleanup)

- Let index_merge allocate table handlers on quick select's MEM_ROOT,
  not on statement's MEM_ROOT. 
  This is crucial for big "range checked for each record" queries, where 
  index_merge can be created and deleted many times during query exection. 
  We should not make O(#rows) allocations on statement's MEM_ROOT.

- When converting a subquery to a semi-join, propagate OPTION_SCHEMA_TABLE.

  
Analysis:
The reason for the inefficent plan was that Item_subselect::is_expensive()
didn't detect the special case when a subquery was optimized, but had no
join plan because it either has no table, or its tables have been optimized
away, or the optimizer detected that the result set is empty.
  
Solution:
Identify the special cases above in the Item_subselect::is_expensive(),
and consider such degenerate subqueries inexpensive.

This bug was introduced by the patch for WL#3220.
If the memory allocated for the tree to store unique elements
to be counted is not big enough to include all of them then
an external file is used to store the elements.
The unique elements are guaranteed not to be nulls. So, when 
reading them from the file we don't have to care about the null
flags of the read values. However, we should remove the flag 
at the very beginning of the process. If we don't do it and
if the last value written into the record buffer for the field
whose distinct values needs to be counted happens to be null,
then all values read from the file  are considered to be nulls
and are not counted in.
The fix does not remove a possible null flag for the read values.
Rather it just counts the values in the same way it was done
before WL #3220.