Merge of ha_innobase.cc from 4.0; This fixes that MySQL doesn't allocate too big buffers when sorting innobase rows
Changed name of variables
NO_FOREIGN_KEY_CHECKS -> FOREIGN_KEY_CHECKS
RELAXED_UNIQUE_CHECKS -> UNIQUE_CHECKS
to avoid confusion with 'not...' variables.
Write name of adminstrator command to slow log.
Added mysqldumpslow to RPM

  Backport the AIX crash fix changes from 4.0

  Add diagnostic code to track a probable table corruption reported by a user

into sergbook.mysql.com:/usr/home/serg/Abk/mysql

bugfix: uniques did not contributed to MYI tot_length estimation, which resulted in "Index file full" error on huge SELECT DISTINCT's

  Make the typecast trick pthread_t -> os_thread_id_t safer, but a full fix needs usage of appropriate Posix functions

into hundin.mysql.fi:/home/heikki/mysql3

  Fix a bug in the storage of the SQL NULL value in 64-bit computers: the value code must fit in 32 bits

  Move the assertion in foreign keys to the right place

  Revert the test changes to wait for a full push of new auto-inc algorithm

  Do not perform ON DELETE action for a FOREIGN KEY constraint if we are doing an UPDATE, not a DELETE
dict0dict.c:
  test

Fix wrong usage of constant which could cause mysqld to use index when doing an update/delete on small tables.
code cleanup

  Align a buffer for i/o from a raw device

  Fix a bug in the previous commit change and align also the buffer in simulataled aio for possible use in i/o to raw devices

  Align the buffer used in initing a data file to zero; this may be needed if the data file is actually a raw device

into sinisa.nasamreza.org:/mnt/work/mysql

  Suppress prints by slave.cc when there is a normal reconnect after the master has been quiet for the time of the net read timeout; this removes the Error 1159 messages many users have complained about in the replication of 3.23.51

  Increase max number of waiting threads to 10000 and put diagnostic prints if this value is exceeded

  On Monty's suggestion move ha_drop_database back inside LOCK_mysql_create_db, but we release LOCK_open during it to let possible running queries complete. I tested this and it seemed to work also in complex concurrent DELETE, SELECT, DROP DATABASE, and CREATE DATABASE situations

  from TeamOS/2 Italy
- removed some obsolete files on Yuri's request
- Added missing os2 related files outside the os2 subdirectory
  to the distribution

  Revert Monty's change to the positioning of ha_drop_database(): otherwise a running query on a table to be dropped in DROP DATABASE always causes the server to hang

ppc64 and x86-64, fixed a compile bug when building with TCP wrapper
support on Linux

Fix for access problem (merge from 4.0)

  Update my.cnf example files to reflect InnoDB autoextend and also simplify them for InnoDB

  Forgot to release an x-latch in DROP DATABASE

  Protect InnoDB monitor with a mutex

I wish I had connection stop  / start so that I could have made a 
test case.

  Remove debug code