Files copied from Percona-Server-5.5.20-rel24.1.tar.gz source tarball.

The replication slave sets first error 1913 and immediately after error
1595. Thus it is possible, but unlikely, to get 1913. The original test
seems to realise this, but uses an invalid error code - my guess is
that this was a temporary code used in a feature tree, which was then
forgotten to be fixed when merged to main. The removed "1923" is
something committed by mistake during tests.

The code was accessing a pointer in a mem_root that might be freed by
another concurrent thread. Fix by moving the access to be done while the
LOCK_thd_data is held, preventing the memory from being freed too early.

Fix of set_limit in case of an error (actually impossible case but better it will be right)

 - mysql-test-run.pl --valgrind complains when all tests succeed.

 - perfschema.all_instances fail on non-linux, where ENABLE_TEMP_POOL
   is not set and therefore BITMAP mutex is not used.

 - MDEV-132: main.mysqldump fails because it depends on exact size of stdio
   buffers.

 - MDEV-99: rpl.rpl_cant_read_event_incident fails due to a race where the
   slave manages to connect while the test case is in the middle of setting up
   the master, causing the slave to replicate extra/wrong events.

 - MDEV-133: rpl.rpl_rotate_purge_deadlock fails because it issues a
   DEBUG_SYNC SIGNAL immediately followed by RESET; this means that sometimes
   the intended receipient has no time to see the signal before it is cleared
   by the RESET, causing wait to timeout.

test passes.

Also, restrict symbol visibility in statically
built plugins, to minimize the chance for symbol 
name clashes with dynamic plugins.

fix pam.tets for 5.5

- Solaris readline always needs curses
- -rdynamic is not portable, replaced by SET_TARGET_PROPERTIES(...ENABLE_EXPORTS)

The reason for the failure is that the loaded library has the same exported symbols
as the builtin one. So the plugin uses innodb functions e.g srv_boot from mysqld
rather than plugin's own. This causes the crash.

On Unix systems with gcc4 later this error was so far worked around using GCC's
visibility attribute. However, in our case, we're using gcc3.

See related MySQL bug http://bugs.mysql.com/bug.php?id=48524 and
http://bugs.mysql.com/bug.php?id=52263

The fix is to restrict symbol visibility in the plugin using version script 
(called map file on Solaris).

Fix egrep syntax error in .deb preinst.

remove building pbxt by default in any configuration (need explicit ./configure --with-pbxt-storage-engine )

huge space usage), and there is no upstream support anymore.

file_contents.test. Also fix some old references to 5.3 in .deb packaging
found while debugging this.

The line in the file_contents.test removes all the '/lib' substrings from the
path, so file cannot be found if a path contains such a substring.
As i didn't find where it is needed, the line was just removed

per-file comments:
  mysql-test/t/file_contents.test
mdev57 5.5 main.file_contents fails on debian5-i386-fulltest.
        no '/lib' substring cutting.

In 5.5, ssl_do() no longer calls report_errors() in case of ssl error.
Since report_errors() iterated over the list of errors, this means that we
now report the first error in the list, rather than the last. Adjust the
--replace_regex line for OpenSSL build accordingly in the test case.

I want to avoid that upgrades silently change important config parameters
that users have come to rely on. This could happen if users changed their
my.cnf themselves, and then an upgrade introduces mariadb.cnf which silently
overrides the settings in my.cnf. Avoid this by having mariadb.cnf mostly
empty for now, and in the future we can add just new mariadb-specific
options there that do not break existing installations.

It needs recompilation for embedded server, as layout of THD is different in embedded.

mysql-common and places mariadb-specific stuff in /etc/mysql/conf.d/mariadb.cnf.
This should allow to co-exist with default Debian mysql-common package and
help resolve dependencies when installing mariadb among multiple available
versions of MySQL from different repositories.

This gives the wrong path when testing installed server, and we
set the correct path in my.cnf anyway.