Merge from 4.0.

This is the second of three changesets. It contains the pure bug fix.
It also contains the second after-review fixes.
The problem was that with gcc on x86, shifts are done modulo word size. 
'value' is 32 bits wide and shifting it by 32 bits is a no-op.
This was triggered by an evil distribution of character incidences. 
A distribution of 2917027827 characters made of 202 distinct values led to
34 occurrences of 32-bit Huffman codes.
This might have been the first time ever that write_bits() had to write
32-bit values. Since it can be expected that one day even 32 bits might
be insufficient, the third changeset suggests to enlarge some variables
to 64 bits.

BUILD/check-cpu:
  Remove the exit 0 from check-cpu

into neptunus.(none):/home/msvensson/mysql/mysql-4.1

 - Remove ha_archive::rename_table and move the fix to handler::rename_table


sql/examples/ha_archive.cc:
  Remove ha_archive::rename_table
sql/examples/ha_archive.h:
  Remove ha_archive::rename_table
sql/handler.cc:
  Fix handler::rename_table so it does not care about if the file to rename is missing

into neptunus.(none):/home/msvensson/mysql/mysql-4.1


configure.in:
  Auto merged

After merge fix.

sql/lock.cc:
  Auto merged
sql/mysql_priv.h:
  Auto merged

BUILD/check-cpu:
  Remove defaul flag i386 and add a check to see wheter to set check_cpu_cflags or not.

into mysql.com:/M41/mysql-4.1

into mysql.com:/usr/home/ram/work/4.1.b9803

 - Reverted removal of errorcheck mutex initialise, used in safe_mutex_init. 


include/my_pthread.h:
  Reverted the removal of errorcheck mutex initializer
mysys/my_thr_init.c:
  Reverted the removal of errorcheck mutex initializer
  Add destruction of mutex initializer

   - Made a script that selects the best compiler optimizations for the current cpu.
   - Use the script from BUILD/SETUP.sh 


BUILD/SETUP.sh:
  Run check-cpu and use the compiler optimization flags it outputs

Added protection against global read lock while creating and
initializing a delayed insert handler.
Allowed to ignore a global read lock when locking the table
inside the delayed insert handler.
Added some minor improvements.


sql/lock.cc:
  Bug#7823 - FLUSH TABLES WITH READ LOCK + INSERT DELAYED = deadlock
  Changed mysql_lock_tables() to allow for ignoring global read lock.
  Added functions to set/unset protection against global read lock.
sql/mysql_priv.h:
  Bug#7823 - FLUSH TABLES WITH READ LOCK + INSERT DELAYED = deadlock
  Changed existing and added new function declarations.
sql/sql_insert.cc:
  Bug#7823 - FLUSH TABLES WITH READ LOCK + INSERT DELAYED = deadlock
  Added and extended some comments.
  Added a protection against global read lock while a handler is
  created and initialized.
  Moved the unlock of the delayed insert object past its last usage
  in delayed_get_table().
  Changed the table locking in handle_delayed_insert() so that it
  does not wait for global read lock.

into neptunus.(none):/home/msvensson/mysql/mysql-4.1


configure.in:
  Auto merged

into neptunus.(none):/home/msvensson/mysql/mysql-4.1


configure.in:
  Auto merged

into neptunus.(none):/home/msvensson/mysql/mysql-4.1

into neptunus.(none):/home/msvensson/mysql/mysql-4.1

 - Display error if system readline or libedit can't be found


configure.in:
  Display an error if user has selected not to use the bundled libedit or readline, AND the system readline or libedit can't be found

  - Changed makelist.sh
  - Bump up required version of autoconf
  - Use new style to init mutex in my_thr_init


cmd-line-utils/libedit/makelist.sh:
  Changed file so it works also on windows cr/lf files.
configure.in:
  Bump up required AC version number so that correct version of aclocal and autoconf is selected.
include/my_pthread.h:
  Use PTHREAD_MUTEX_ADAPTIVE_NP to see if "fast" mutexes are available
  Remove "errorcheck" mutexes, since they are never used.
mysys/my_thr_init.c:
  Use new style functions to init mutex if PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP is defined
  Add comment what mutex "kind" means

into mysql.com:/home/jonas/src/mysql-4.1


ndb/src/ndbapi/ndberror.c:
  Auto merged

  Fix a problem in crash recovery of .ibd files on Windows if the user used lower_case_table_names=0 or 2; the directory scan in crash recovery forgot to put all paths to lower case, so that the tablespace name would be consistent with the internal data dictionary of InnoDB; remember that InnoDB puts internally all database names and table names to lower case on Windows, regardless of the value of lower_case_table_names


innobase/fil/fil0fil.c:
  Fix a problem in crash recovery of .ibd files on Windows if the user used lower_case_table_names=0 or 2; the directory scan in crash recovery forgot to put all paths to lower case, so that the tablespace name would be consistent with the internal data dictionary of InnoDB; remember that InnoDB puts internally all database names and table names to lower case on Windows, regardless of the value of lower_case_table_names
innobase/include/dict0dict.h:
  Fix a problem in crash recovery of .ibd files on Windows if the user used lower_case_table_names=0 or 2; the directory scan in crash recovery forgot to put all paths to lower case, so that the tablespace name would be consistent with the internal data dictionary of InnoDB; remember that InnoDB puts internally all database names and table names to lower case on Windows, regardless of the value of lower_case_table_names
innobase/dict/dict0dict.c:
  Fix a problem in crash recovery of .ibd files on Windows if the user used lower_case_table_names=0 or 2; the directory scan in crash recovery forgot to put all paths to lower case, so that the tablespace name would be consistent with the internal data dictionary of InnoDB; remember that InnoDB puts internally all database names and table names to lower case on Windows, regardless of the value of lower_case_table_names

into mysql.com:/M41/mysql-4.1


acinclude.m4:
  Auto merged
configure.in:
  Auto merged

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

acinclude.m4:
  Fix typo

no rows of the blackhold engine (for this engine to be a binlog propagator).
NOTE: blackhole.test currently hangs in 5.0 BUG#10175 so I'll merge without running this test; the
person who fixes the hang will correct the merged blackhole.result or ask me to).


mysql-test/r/blackhole.result:
  result update
mysql-test/t/blackhole.test:
  checking that statements get into binlog even when they were affecting
  no rows of the blackhold engine. So that this engine can serve as
  a binlog propagator (A->B->C replication where B has only blackhole tables,
  B receives binlog from A, applies it to blackhole tables, writes statements
  to its binlog which it sends to C; both A and C have non-blackhole tables). B is just a proxy.

include/my_global.h:
  As we have a new CPP symbol, this should be "undef"ed explicitly.

 - Implemented ha_archive::rename_table
 - Added testcases for rename


mysql-test/r/archive.result:
  Addd testcase for rename of archive table
mysql-test/t/archive.test:
  Addd testcase for rename of archive table
sql/examples/ha_archive.cc:
  Implement special version of rename table that does not care it the .arn file is missing
sql/examples/ha_archive.h:
  Implement special version of rename table that does not care it the .arn file is missing

acinclude.m4:
  Use AC_LANG_PUSH/POP instead of _SAVE/RESTORE
  Add test to get type of 'struct rlimit'
  Switch order of including stdlib.h and declaration being tested to
  match how it will be used in regular code.
configure.in:
  Call MYSQL_TYPE_STRUCT_RLIMIT macro
sql/mysqld.cc:
  Use STRUCT_RLIMIT for getting type of struct rlimit.

  - Print readline or libedit version


client/mysql.cc:
  Print the version of readline or libedit library used.

  post review fix.
  don't init hash value on restart operations
  make sure that lock mode is correct in entire que.


ndb/test/ndbapi/testOperations.cpp:
  Test case also for abort case.
    (abort if not lock owner still fails!)

into mysql.com:/home/dlenev/src/mysql-4.1-bg10040

"'func_sapdb' fails on 64bit build hosts")


sql/item_timefunc.cc:
  calc_time_diff():
    Fixed yet another signedness problem.

  add more tests to verify correctness of backups


ndb/test/run-test/daily-basic-tests.txt:
  ndb - test
    add more tests to verify correctness of backups

  remove incorrect assertion

  uninit variable

ndb/include/kernel/signaldata/TuxMaint.hpp:
  handle multipl index update failure atomically
ndb/src/kernel/blocks/dbtup/Dbtup.hpp:
  handle multipl index update failure atomically
ndb/src/kernel/blocks/dbtup/DbtupTrigger.cpp:
  handle multipl index update failure atomically
ndb/src/kernel/blocks/dbtup/Notes.txt:
  handle multipl index update failure atomically
ndb/src/kernel/blocks/dbtux/DbtuxNode.cpp:
  handle multipl index update failure atomically
ndb/src/ndbapi/ndberror.c:
  handle multipl index update failure atomically
ndb/test/ndbapi/testOIBasic.cpp:
  handle multipl index update failure atomically

Change "configure.in".

The branches for Tru64 (ex: OSF/1) and Linux are separated,
and for Linux 'getconf GNU_LIBPTHREAD_VERSION' is used to tell between Linuxthreads and NPTL.


configure.in:
  Step 3 of the switch to support configuration with NPTL:
  Change "configure.in".
  
  1) Separate the branches checking for thread libraries on OSF/1 (now: Tru64) and on Linux,
     as these are different things. In the Tru64 branch, remove the obsolete part for OSF/1 3.2,
     because that is gone since long (about a decade?); we are at 5 now. Adapt the comment.
  
  2) For Linux, the thread libraries are "Linuxthreads" (older) and "NPTL" (newer).
     Until now, we cared only about "Linuxthreads", and we did so by checking a comment in 
     the header file '/usr/include/pthread.h'.
     With NPTL as an alternative or only thread package, this is not secure any more,
     there are distributions where the header file still contains that comment but the lib is NPTL.
     For all current distributions, the way to check is the 'getconf' command.
  
     On old distributions, 'getconf' exists but does not understand the 'GNU_LIBPTHREAD_VERSION'
     parameter. It is unlikely that such an old system should have NPTL, so the header file check is
     sufficient if the getconf inquiry fails.
  
     For both Linuxthreads and NPTL, '-lpthread' is the linker option to use.
  
  3) To detect problems as early as possible, on Linux a link test using '-lpthread' is done.
     The code is copied from a later check where the thread functions are searched in '-lc',
     '-lpthread', and '-lpthreads', but this later search would not be executed after the lib
     was already determined. So if the test is not done here, it will be detected only during
     the real build. This is too late.

into mysql.com:/usr/local/home/marty/MySQL/mysql-4.1