Do not let range estimator to return over 1 / 2 of total rows in table; use longlong in range estimation
btr0cur.h, ha_innobase.cc:
  Use longlong in range estimation, in case there are > 4 billion rows

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

  Fix bug: if SHOW INNODB STATUS or innodb_monitor printed several hundred transactions, the output was truncated and we forgot to release the kernel mutex

  Remove unintentionally pushed assert

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

into mysql.com:/space/my/mysql-3.23

   InnoDB, so test it after InnoDB)

  Backport from 4.0.4 the bug fix of the crash when a temporary table was created inside LOCK TABLES and used

  Fix the probable reason why InnoDB versions after July 2, 2002 did not start properly in some Win 95/98/ME computers

   with static glibc installed

into mysql.com:/space/my/mysql-3.23-spec

 - Updated scripts/make_binary_distribution.sh to include the correct
   ChangeLog and the LICENSE file for non-GPL builds

  Fix bug: the AUTO-INC lock was held to the end of trx if it was granted after a lock wait

 - some cosmetical fixups (typos, comments)

  static nss linkingg issue

Added missing mutex-lock around critical section in GRANT handling.

   of the static_nss stuff required for linking against the patched
   glibc)

  Fix typos in previous push

  Allow CREATE TABLE and DROP TABLE even if innodb_force_recovery is used, the user can drop a problematic table

  Change srv_flush_log_at_trx_commit to ulint, note that ibool is defined as ulint, so this is purely formal change
os0file.c:
  Start using unbuffered i/o again in Windows because sequential read using normal i/o was 4 times slower in XP

  Cleanup after last push

  update help innobase_flush_log_at_trx_commitcan have vals 0, 1, 2 
ha_innobase.h, ha_innobase.cc:
  make innobase_flush_log_at_trx_commit a long, like other params

  Fix a typo: should be buf_page_print

  innodb_flush_log_at_trx_commit can have values 0, 1, 2 starting from 3.23.52

  Remove an assertion in trx0undo.c which could erroneously fail when we ran out of tablespace

  Increase max latch wait time to 600 seconds before we assert