Fix for bug #10055 "Using stored function with information_schema causes empty
result set".
To enable full access to contents of I_S tables from stored functions
or statements that use them, we manipulate with thread's open tables
state and ensure that we won't cause deadlock when we open tables by
ignoring flushes and name-locks.
Building of contents of I_S.TABLES no longer requires locking of tables
since we use use handler::info() method with HA_STATUS_AUTO flag instead
of handler::update_auto_increment() for obtaining information about
auto-increment values. But this also means that handlers have to implement
support for HA_STATUS_AUTO flag (particularly InnoDB needs it).
mysql-test/r/alter_table.result:
Updated test results. This change was caused by the fact that now when
we build contents of I_S tables (and thus output of SHOW INDEX) we
don't use instances of tables which may be already opened and locked
by thread (we always use new instance).
mysql-test/r/information_schema.result:
Added test which checks how information about current auto-increment value for
table is reported in INFORMATION_SCHEMA.TABLES view.
mysql-test/r/sp.result:
Added test for bug #10055 "Using stored function with information_schema causes
empty result set".
mysql-test/t/information_schema.test:
Added test which checks how information about current auto-increment value for
table is reported in INFORMATION_SCHEMA.TABLES view.
mysql-test/t/sp.test:
Added test for bug #10055 "Using stored function with information_schema causes
empty result set".
sql/mysql_priv.h:
close_thread_tables():
Get rid of 'stopper' argument which is no longer used. Now when we need
to open and then close some table without touching tables which are already
opened we use THD::reset_n/restore_backup_open_tables_state() methods.
open_tables()/open_normal_and_derived_tables():
Added 'flags' argument to be able open tables even if some has done
a flush or hold namelock on them.
sql/sp.cc:
close_proc_table/open_proc_table_for_read/db_find_routine():
Replaced push_open_tables_state/pop_open_tables_state() methods which
were saving/restoring current open tables state in/from THD::open_state_list
with reset_n_backup_open_tables_state/restore_backup_open_tables_state()
methods which assume that backup storage for this state is allocated on
stack (or elsewhere) by their caller.
open_proc_table_for_read():
Since now we can have several open tables states stacked up we can't rely
rely on checking whether we have some tables open in previous state.
Instead we always assume that some tables are open and we need to ignore
flush while locking mysql.proc. We don't really need
MYSQL_LOCK_IGNORE_GLOBAL_READ_LOCK in this case since we open mysql.proc table
only for reading.
sql/sp.h:
Added declarations of open_proc_table_for_read()/close_proc_table() to be
able to use them in sql_show.cc.
sql/sql_base.cc:
close_thread_tables():
Get rid of 'stopper' argument which is no longer used. Now when we need
to open and then close some table without touching tables which are already
opened we use THD::reset_n/restore_backup_open_tables_state() methods.
open_tables()/open_normal_and_derived_tables():
Added 'flags' argument to be able open tables even if some has done
a flush or hold namelock on them.
sql/sql_class.cc:
Open_tables_state, THD:
Replaced push_open_tables_state/pop_open_tables_state() methods which
were saving/restoring current open tables state in/from THD::open_state_list
with reset_n_backup_open_tables_state/restore_backup_open_tables_state()
methods which assume that backup storage for this state is allocated on
stack (or elsewhere) by their caller.
sql/sql_class.h:
Open_tables_state, THD:
Replaced push_open_tables_state/pop_open_tables_state() methods which
were saving/restoring current open tables state in/from THD::open_state_list
with reset_n_backup_open_tables_state/restore_backup_open_tables_state()
methods which assume that backup storage for this state is allocated on
stack (or elsewhere) by their caller.
sql/sql_handler.cc:
open_tables()/open_normal_and_derived_tables():
Added 'flags' argument to be able open tables even if some has done
a flush or hold namelock on them.
sql/sql_prepare.cc:
open_tables()/open_normal_and_derived_tables():
Added 'flags' argument to be able open tables even if some has done
a flush or hold namelock on them.
sql/sql_show.cc:
get_all_tables():
Now we use THD::reset_n_/restore_backup_open_tables_state() for
saving/restoring open tables state instead of working with it directly
(This also allows us to have proper content of I_S system tables in
statements with stored functions and in stored functions). We also
ignore possible flushes when opening tables (we may create deadlock
otherwise). Also we do all needed manipulations with LEX in this function
and not in get_schema_tables_result() now.
get_schema_tables_record():
Let us use handler::info() method with HA_STATUS_AUTO flag for obtaining
information about table's auto-increment value. This allows to avoid locking
of tables which is needed when we use handler::update_auto_increment() method.
fill_schema_proc():
Now we use open_proc_table_for_read/close_proc_table() for access to
mysql.proc table (so we won't cause deadlock if we already have some
tables open and locked, this also allows us to have proper content in
ROUTINES system table in statements using stored functions/in stored
functions).
get_schema_tables_result():
Moved all manipulations with Open_tables_state and LEX needed for
safe opening of tables to ST_SCHEMA_TABLE::fill_table functions
(i.e. get_all_tables() and fill_schema_proc()).
sql/sql_update.cc:
open_tables()/open_normal_and_derived_tables():
Added 'flags' argument to be able open tables even if some has done
a flush or hold namelock on them.
Showing
Please register or sign in to comment