Cause of the bug is uninitialized items before evaluation HAVING clasue in case of empty result.

Problem was a crash in internal temporary (Maria) files when row length exceeded 65535

mysql-test/suite/maria/r/maria3.result:
  Added test case
mysql-test/suite/maria/t/maria3.test:
  Added test case
storage/maria/ma_open.c:
  Added support for row length > 65535.
  This fixes crash when using tables with longer row lengths.

the mysql-5.6 code line.

mysql-test/suite/innodb_plugin/t/innodb_gis.test:
  Don't run test if innodb_plugin isn't used.

sql/signal_handler.cc:
  Added missing signal values.

Fixed supression in mysql-test-run so it also works on windows.

mysql-test/mysql-test-run.pl:
  Fixed supression so it also works on windows.
mysql-test/valgrind.supp:
  More general handling of memory loss in dlclose (backported from 5.2)
sql/signal_handler.cc:
  Added newlines around link to how to do bug reports

"safe exception patch".
Remove misleading comments suggesting about signal() Windows, the routine here
is part of a exception handler, and sig parameter is an exception code.

Fixed README with link to source
Merged InnoDB change to XtraDB

README:
  Added information of where to find MariaDB code
storage/archive/ha_archive.cc:
  Removed memset() of rows, a MariaDB checksum's doesn't touch not used data.

This happend when you have more than 1024 open Aria tables during checkpoint.


mysql-test/mysql-test-run.pl:
  Fixed that variable names are consistent between external and internal server.
mysql-test/suite/maria/suite.pm:
  Test for aria-block-size instead of 'aria' as 'aria' is not set for embedded server.
  This should be ok for aria tests, as aria is never disabled for these.
storage/maria/ma_checkpoint.c:
  Fixed bug when there are more than 1024 open Aria tables during checkpoint.

The bug was fixed with the code back-ported from the patch for LP bug 800184
pushed into mariadb-5.3.

Protocol documentation (http://forge.mysql.com/wiki/MySQL_Internals_ClientServer_Protocol)
says that initial packet sent by client if client wants SSL, consists of client capability flags only
(4 bytes or 2 bytes edependent on protocol versionl). 
Some clients happen to send more in the initial SSL packet (C client, Python connector), while others (Java, .NET) follow the docs and send only client capability flags.

A change that broke Java client was a newly introduced check that frst client packet
has 32 or more bytes. This is generally wrong, if client capability flags contains CLIENT_SSL.

Also, fixed the code such that read max client packet size and charset in the first packet prior to SSL handshake.  With SSL, clients do not have to send this info, they can only send client flags.

This is now fixed such that max packet size and charset are not read prior to SSL handshake, in case of SSL they are read from the "complete" client authentication packet after SSL initialization.

For single table update/insert added deep check of single tables (single_table_updatable()).
For multi-table view insert added additional check of target table (check_view_single_update).
Multi-update was correct.

Test suite for all cases added.

The function subselect_uniquesubquery_engine::copy_ref_key has to take into
account that when EXPLAIN is processed the array of store_key object created
for any TABLE_REF may contain elements for constant items. These items should
be ignored by thefunction.

1. on windows: don't hang when there's no console, that is, _getch() returns -1. 
2. on windows: _getch() returns an int, not char.
   to distinguish between (char)255 and (int)-1
3. everywhere. isspace(pos[-1]) == ' ' never worked,
   isspace() returns a boolean, not a char. the never-worked loop was
   removed to preserve the existing behavior.

test added.

The cause of this bug was the same as for bug 902356 fixed for 5.3.