implemented in the client library and server.
Warning: this makes the new client library and server incompatible with all 
previous versions and 5.0.

Collation gets lost in CREATE TABLE (...) DEFAULT CHARACTER SET xxx COLLATE yyy

This is fixed now.
Test is added into the test suite.

Forgot to do it in the previous commit.

and for (2) converting the columns:

1.  ALTER TABLE t1 [DEFAULT] CHARACTER SET xxx [COLLATE yyy]
2.  ALTER TABLE t1 CONVERT TO CHARACTER SET xxx [COLLATE yyy]

ChangeSet
  1.1707 04/03/19 12:36:55 ram@gw.mysql.r18.ru +4 -0
  Fix for the bug #2976: NATURAL JOIN produces duplicate columns.

Fixes after last merge

into mysql.com:/home/my/mysql-4.1

into mysql.com:/home/my/mysql-4.1

into mysql.com:/home/my/mysql-4.0

SET character_set_xxx=NULL
and
SET collation_xxx=NULL

  More clear error message.

No need to check for result existence any more, store_result functions
now are shorter.
cli_read_binary_rows rewritten to handle MYSQL_DATA directly.

into deer.(none):/home/hf/work/mysql-4.1.clean

we should call check_user in NO_EMBEDDED_ACCESS_CHECKS case
to do necessary initializations

into mysql.com:/home/mysql_src/mysql-4.1

that it tested the privilege in ::update() whereas it should be
in ::check() (see email from Serg, subject
"Re: bk commit - 4.1 tree (guilhem:1.1706)").
So I add instead a check_func function to sys_var_thd_bit.
I do the same addition to sys_var_thd_ulong, to unify handling
of PSEUDO_THREAD_ID with the one of SQL_LOG_BIN. So class
sys_var_pseudo_thread_id is not needed anymore, removing it.

"Running ANALYZE TABLE on bdb table inside a transaction hangs server thread"

into eagle.mysql.r18.ru:/home/vva/work/BUG_2342/mysql-4.0

"Running ANALYZE TABLE on bdb table inside a transaction hangs server thread"

1. added new status HA_ADMIN_REJECT and processing of it in mysql_admin_table
2. got ha_berkley::analyze to return HA_ADMIN_REJECT if there are any 
 transactions with the table.. 

into mysql.com:/home/mysql_src/mysql-4.1

user is not SUPER, instead of setting the value but ignoring it
silently internally.

init the binlog_cache (THD::transaction.trans_log).
I have checked all places where trans_log is used, because as now
it may not be inited in some cases, we have to be cautious
(will forward this commit mail to Heikki).

into sinisa.nasamreza.org:/mnt/work/mysql-4.0

into hundin.mysql.fi:/home/heikki/mysql-4.0

  Backport from 4.1: make mem_current_allocated_memory global variable

  Let mem_current_allocated_memory be a global variable, because it is used in an .ic file if UNIV_MEM_DEBUG is defined

into sinisa.nasamreza.org:/mnt/work/mysql-4.0

  result of the test case for FORCE INDEX on ORDER BY
order_by.test:
  test case for FORCE INDEX on ORDER BY
sql_select.cc:
  Changing behaviour that MySQL server takes FORCE INDEX clause into account when optimising ORDER BY clause

into sanja.is.com.ua:/home/bell/mysql/bk/work-4.1