- rename upgrade_wizard.exe to mysql_upgrade_wizard.exe
- have shortcut to upgrade wizard in the menu folder

If upgradable instances are found, bring a new dialog  to inform user about it. This gives user a
chance to deselect "database instance" feature early,  because experience shows nobody really looks at features and their in their description  in "customize setup" dialog. This also tells  user 
that existing instances can  be upgraded.

and mysql_upgrade_service, after LPBUG#688404 was 
fixed.

The reason for the crash is misalignment on SSE instruciton
in setjmp(). The root cause is PBXT debug malloc(),  which
unlike  OS malloc  does not guarantee 16 bytes alignment.

So the fix for now is disable PBXT debug malloc on Windows.
It was obsolete anyway, as  it does not provide additional
benefits to C runtime debug routines (always used in debug 
compilation) or to pageheap, available at runtime.

mysqld --install without any parameters.

In such case, service name is always MYSQL, as service
binary path is "path\to\mysqld.exe" "MySQL". Guess data
directory it is either from my.ini (which  is assumed to
be in the installation root), or just data directory 
under install root.
 

Allow MSI and  NSIS side-by-side installation if  installed NSIS package
differs in  "major.minor" version numbers. Still disallow MSI and NSIS 
if major.minor versions  of both packages match.

BUILD_RELEASE to disable graceful fallbacks
if WiX or MFC is not available.

(Philip's review)

Take into account that services registered by  MySQL do not have
 .exe extension  in service binary path.

Problem: user without privileges can have an half-complete
installation, if he manages to click on "Ignore" for all  errors
in the installer. 

As a result, he will miss ARP registry keys, and uninstall 
will not be possible using  "Add/Remove Programs" applet.

on Win7 with the most strict user account control setting 
(secure desktop)

Fix: use job object for child process only, not for current
process itself.

* Spell username correctly as "NT AUTHORITY\NetworkService"
* Also, use well-known SIDs for predefined user when assigning directory
ACLs (the names differ in localized Windows)

The  general technique to generate MSI using CMake is taken from MySQL 5.5

Additional features not present in 5.5 installer : 
-optionally creating a new database
(as Windows service), using new mysql_install_db.exe to do the job
 
- optional upgrade of existing services from old MySQL or Maria installation.
This work is actually done by the upgrade wizard that is launched at the 
end of installation.

to uzpgrade existing MySQL/Maria services to higher version.

To be used in installer (but also can be used outside of installer too)

New utility to upgrade Windows service to higher MariaDB version.
Its functionality includes changing service definition as well as
running mysql_upgrade.

on Windows.

Some parameters not present in traditional mysql_install_db are present
e.g --port, --default-user (whether to create a new users) or 
--service (windows service name)

to be able to build MSI based installer

Fixes part of LP:705213 (Other part is to be pushed into 5.1)

storage/ndb/include/kernel/signaldata/FsOpenReq.hpp:
  Fixes compiler warnings
storage/ndb/include/ndbapi/NdbDictionary.hpp:
  Update to right typedef
storage/ndb/include/util/NdbSqlUtil.hpp:
  Remove not needed, conflicting code.
storage/ndb/src/kernel/blocks/dbtup/DbtupExecQuery.cpp:
  Added cast to get rid of not critical const.

sql/sql_base.cc:
  fill_record should be used for only one table (no need list).

support building with -all-static (no dlopen and dlclose make few
related declarations unused or "statement have no effect") and -Werror

don't pass NULL as an argument where va_list is expected.

Fixes failures in merge.test

Fix for LP#697610 ha_index_prev(uchar*): Assertion `inited==INDEX' failed with HANDLER + InnoDB in maria-5.3


mysql-test/suite/handler/innodb.result:
  Added test case
mysql-test/suite/handler/innodb.test:
  Added test case
sql/handler.h:
  Move setting/resetting of active_index to ha_index_init()/ha_index_end() to simplify handler functions index_init()/index_end()
  Fixed that get_index() returns MAX_KEY if index is not inited (this fixed LP#697610)
storage/federated/ha_federated.cc:
  Settting of active_index is not needed anymore
storage/maria/ma_pagecache.c:
  Added error message if we have too little memory for Maria page cache

storage/maria/ma_pagecache.c:
  - Ensure that we write also the last buffer in flush if buffer overflows.
  - Don't set PCBLOCK_IN_FLUSH if we don't flush the page
  - Added some new safety asserts.