cherry-pick from 5.7:
  commit 6b24763
  Author: Manish Kumar <manish.4.kumar@oracle.com>
  Date:   Tue Mar 27 13:10:42 2012 +0530

  BUG#12977988 - ON STOP SLAVE: ERROR READING PACKET FROM SERVER: LOST CONNECTION
                 TO MYSQL SERVER
  BUG#11761457 - ERROR 2013 + "ERROR READING RELAY LOG EVENT" ON STOP SLAVEBUG#12977988 - ON STOP SLAVE: ERROR READING PACKET FROM SERVER: LOST CONNECTION
               TO MYSQL SERVER

Problem was that if old virtual column is computed and stored there
was no check if new column is really virtual column.

It is set in sql_yacc.yy.

Code flow hit incorrect branch while closing table instances before removal.
This branch expects thread to hold open table instance, whereas CREATE OR
REPLACE doesn't actually hold open table instance.

Before CREATE OR REPLACE TABLE it was impossible to hit this condition in
LTM_PRELOCKED mode, thus the problem didn't expose itself during DROP TABLE
or DROP DATABASE.

Fixed by adjusting condition to take into account LTM_PRELOCKED mode, which can
be set during CREATE OR REPLACE TABLE.

The following directives to ignore warnings where in the PerconaFT build in tokudb.
These generate errors when g++ ... -o xxx.so is used to compile are shared object.

As these don't actually hit any warnings they have been removed.

* -Wno-ignored-attributes
* -Wno-pointer-bool-conversion
Signed-off-by: Daniel Black <daniel.black@au.ibm.com>

coming from Percona's workaround for glibc bug
http://bugs.mysql.com/bug.php?id=82886

The crash is caused by macro uint3korr() accessing memory (1 byte) past
the end of allocated page. The macro is written such it reads 4 bytes
instead of 3 and discards the value of the last byte.

However, it is not always guaranteed that all uint3korr accesses will be
valid (i.e that the caller allocates an extra byte after the value).

In particular, the tree in Item_func_group_concat does not account for
any extra bytes that it would need for comparison of keys in some cases
(Field_newdate::cmp, Field_medium::cmp)

The fix change uint3korr so it does not access extra bytes.

commented out the "compressed columns" feature

This reverts commit 1daf746e.

Removed temporarily to make sure there are no legal problems.

The man page was already present in the debian release of MariaDB 10.0.

Make use of a different function to get the current tid.

Additionally, librt doesn't exist on OS X. Use System library instead.

- don't use stat() for file size, it doesn not handle large size
  use GetFileSizeEx() instead

- don't use lseek(), it can't handle large files, use _lseeki64() instead.

- Also, switch off OS file buffering for innochecksum on Windows,
to avoid thrashing file cache.

Clang/LLVM has more strict schemantics than gcc.

This patch quantifies the namesspace such that it will
compile using clang.

Prior to this patch name of the user was  read from environment variable
USER, with a fallback to 'ODBC', if the environment variable is not set.

The name of the env.variable is incorrect (USERNAME usually contains current
user's name,  but not USER), which made client to  always determine
current user as 'ODBC'.

The fix is to use GetUserName() instead.

From https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=838914

Fixes CMake so that when building a 32-bit mips binary on a 64-bit
mips machine, the target is not set as 32-bit, which apparently
confused some tests in mroonga.

The patch fixes 128-bit multiply on mips64.

This corrects a previous incorrect patch upstreamed from Debian.

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=838557

MIPS has a different errno for "directory not empty".

Fix PSI idle and socket instrumentation in threadpool

Patch provided by Honza Horak

OpenSSL problems, part II

  Now the null is tested using the result set getObject method.
  modified:   storage/connect/JdbcInterface.java
  modified:   storage/connect/jdbconn.cpp
  modified:   storage/connect/jdbconn.h

don't let identifiers with new lines to break a comment

pass them through as is

* use proper sql quoting rules for USE, while preserving
  as much of historical behavior as possible
* short commands (\u) behave as before

  Was because the quoting character was always '"' instead of being
  retrieve from the JDBC source.
  modified:   storage/connect/JdbcInterface.java
  modified:   storage/connect/jdbconn.cpp
  modified:   storage/connect/tabjdbc.cpp