Bug fixes

Fix compiler issues

Final part of the error handling of ALTER TABLE for partitioning

Started writing detailed error handler of
ALTER TABLE for partitioning.
Philosophical issue, should one always attempt to make recovery
automatic or should one rely on manual means also.

Some more error handling

First step for handling errors in ALTER TABLE for partitioning

Added action code for recovery

Introduce log entry to handle replace of frm file

Make lots of fixes to handle the most complex case of reorganising
of partitions where two-phased processes are needed in some cases.

Free all memory with mutex and destroy mutex as final step

New function to set exec_log_entry
Moving code outside lock of table log
Fixing some initialisations of list reference to avoid
releasing wrong parts

into  c-9a08e253.1238-1-64736c10.cust.bredbandsbolaget.se:/home/pappa/wl2826

Bug fixes

Error handling
Crash if any error after reaching beyond certain point in
ALTER TABLE processing

More work on table logging of ALTER TABLE for partitioning

A little more complete handling

More fixes

A number of fixes

into  c-9a08e253.1238-1-64736c10.cust.bredbandsbolaget.se:/home/pappa/wl2826

Step 14: First version of table log for add/Drop partition

Close down table log also at exit of main thread

Lots of compilation fixes

Most of the code for handling the table log is in place now, except
the action part at recovery and proper error handling in some places.

More table log code

Lots of new code for table log

New methods to read/write/initialise table log

Updated error inject scripts

Inserted monty patch to allow for adding and dropping
DBUG keywords per thread

Removed session variables for error injects
started using DBUG macros for error injects

Removed partition state from parser

Remove partition state from frm file
Ensure we can still load tables created in 5.1.6 for the time
being

into  perch.ndb.mysql.com:/home/jonas/src/mysql-5.1-new

into  perch.ndb.mysql.com:/home/jonas/src/mysql-5.1-new

  Allow GET_TABINFO during BACKUP
  (otherwise LCP gets corrupt)