Handlerton array is now created instead of using sys_table_types_st. All storage engines can now have inits and giant ifdef's are now gone for startup. No compeltely clean yet, handlertons will next be merged with sys_table_types. Federated and archive now have real cleanup if their inits fail.
"SELECT ... FOR UPDATE executed as consistent read inside LOCK TABLES"
Do not discard lock_type information as handler::start_stmt() may require knowledge.
(fixed by Antony)
Fix bug #12308 : do not roll back the whhole transaction in a lock wait timeout error, just roll back the latest SQL statement; note that the locks set in the latest SQL statements remain, as InnoDB does not know what locks were set in which SQL statement
Fix bug #12410 : InnoDB was too permissive with LOCK TABLE ... READ LOCAL, and alowed new inserts to the table; we now make READ LOCAL equivalent to READ for InnoDB; note that this will cause slightly more locking in mysqldump, but makes the InnoDB table dumps consistent with MyISAM table dumps; note that the real code change patch was accidentally pushed with my another patch 5 minutes ago
Fix bug #12852 : do not increment the open handle count to a table if the table does not have an .ibd file and InnoDB decides to return an error from the ::open() function; then the table can be dropped even if the user has tried to open it
Fix bug #12779 : never give a row count estimate of 0 to the MySQL query optimizer, as then left join optimizer may beleive it KNOWS that the table is empty; note that this fix may change query optimization of many other queries where one table is empty; note that the proper fix would be to make the query optimizer to know that the row count estimates it receives really are just estimates, it cannot assume they are certain
sql_print_warning() and sql_print_error() instead of fprintf to stderr.
Above functions are tuned for different platforms so that the behavior
is consistent around platforms. Using fprintf() different behavior can
be expected at least on Windows and Unix.
When PRIMARY KEY is present ha_innobase::cmp_ref() uses it to compare refs.
After comparing part of key it moves pointers to compare next part.
For varchar parts pointers were moved only by length of parts, not including
bytes containig part length itself. This results in wrong comparision and
wrong number of deleted records.
