diff options
Diffstat (limited to 'sql')
52 files changed, 4196 insertions, 904 deletions
diff --git a/sql/field.cc b/sql/field.cc index 16217e69de9..7c4f6c9ff5f 100644 --- a/sql/field.cc +++ b/sql/field.cc @@ -1308,7 +1308,8 @@ Field::Field(uchar *ptr_arg,uint32 length_arg,uchar *null_ptr_arg, field_name(field_name_arg), key_start(0), part_of_key(0), part_of_key_not_clustered(0), part_of_sortkey(0), unireg_check(unireg_check_arg), - field_length(length_arg), null_bit(null_bit_arg) + field_length(length_arg), null_bit(null_bit_arg), + is_created_from_null_item(FALSE) { flags=null_ptr ? 0: NOT_NULL_FLAG; comment.str= (char*) ""; @@ -5633,6 +5634,13 @@ String *Field_date::val_str(String *val_buffer, } +bool Field_date::get_time(MYSQL_TIME *ltime) +{ + bzero((char *)ltime, sizeof(MYSQL_TIME)); + return 0; +} + + int Field_date::cmp(const uchar *a_ptr, const uchar *b_ptr) { int32 a,b; diff --git a/sql/field.h b/sql/field.h index 53788560e68..e03131a10e3 100644 --- a/sql/field.h +++ b/sql/field.h @@ -90,6 +90,16 @@ public: uint32 flags; uint16 field_index; // field number in fields array uchar null_bit; // Bit used to test null bit + /** + If true, this field was created in create_tmp_field_from_item from a NULL + value. This means that the type of the field is just a guess, and the type + may be freely coerced to another type. + + @see create_tmp_field_from_item + @see Item_type_holder::get_real_type + + */ + bool is_created_from_null_item; Field(uchar *ptr_arg,uint32 length_arg,uchar *null_ptr_arg, uchar null_bit_arg, utype unireg_check_arg, @@ -1240,6 +1250,7 @@ public: double val_real(void); longlong val_int(void); String *val_str(String*,String *); + bool get_time(MYSQL_TIME *ltime); bool send_binary(Protocol *protocol); int cmp(const uchar *,const uchar *); void sort_string(uchar *buff,uint length); @@ -1258,6 +1269,10 @@ public: :Field_str(ptr_arg, 10, null_ptr_arg, null_bit_arg, unireg_check_arg, field_name_arg, cs) {} + Field_newdate(bool maybe_null_arg, const char *field_name_arg, + CHARSET_INFO *cs) + :Field_str((uchar*) 0,10, maybe_null_arg ? (uchar*) "": 0,0, + NONE, field_name_arg, cs) {} enum_field_types type() const { return MYSQL_TYPE_DATE;} enum_field_types real_type() const { return MYSQL_TYPE_NEWDATE; } enum ha_base_keytype key_type() const { return HA_KEYTYPE_UINT24; } diff --git a/sql/filesort.cc b/sql/filesort.cc index 1c1a3b8d9a3..12cc76b5852 100644 --- a/sql/filesort.cc +++ b/sql/filesort.cc @@ -41,7 +41,8 @@ if (my_b_write((file),(uchar*) (from),param->ref_length)) \ static char **make_char_array(char **old_pos, register uint fields, uint length, myf my_flag); -static BUFFPEK *read_buffpek_from_file(IO_CACHE *buffer_file, uint count); +static uchar *read_buffpek_from_file(IO_CACHE *buffer_file, uint count, + uchar *buf); static ha_rows find_all_keys(SORTPARAM *param,SQL_SELECT *select, uchar * *sort_keys, IO_CACHE *buffer_file, IO_CACHE *tempfile,IO_CACHE *indexfile); @@ -246,9 +247,14 @@ ha_rows filesort(THD *thd, TABLE *table, SORT_FIELD *sortorder, uint s_length, } else { - if (!table_sort.buffpek && table_sort.buffpek_len < maxbuffer && - !(table_sort.buffpek= - (uchar *) read_buffpek_from_file(&buffpek_pointers, maxbuffer))) + if (table_sort.buffpek && table_sort.buffpek_len < maxbuffer) + { + x_free(table_sort.buffpek); + table_sort.buffpek= 0; + } + if (!(table_sort.buffpek= + (uchar *) read_buffpek_from_file(&buffpek_pointers, maxbuffer, + table_sort.buffpek))) goto err; buffpek= (BUFFPEK *) table_sort.buffpek; table_sort.buffpek_len= maxbuffer; @@ -376,14 +382,16 @@ static char **make_char_array(char **old_pos, register uint fields, /** Read 'count' number of buffer pointers into memory. */ -static BUFFPEK *read_buffpek_from_file(IO_CACHE *buffpek_pointers, uint count) +static uchar *read_buffpek_from_file(IO_CACHE *buffpek_pointers, uint count, + uchar *buf) { - ulong length; - BUFFPEK *tmp; + ulong length= sizeof(BUFFPEK)*count; + uchar *tmp= buf; DBUG_ENTER("read_buffpek_from_file"); if (count > UINT_MAX/sizeof(BUFFPEK)) return 0; /* sizeof(BUFFPEK)*count will overflow */ - tmp=(BUFFPEK*) my_malloc(length=sizeof(BUFFPEK)*count, MYF(MY_WME)); + if (!tmp) + tmp= (uchar *)my_malloc(length, MYF(MY_WME)); if (tmp) { if (reinit_io_cache(buffpek_pointers,READ_CACHE,0L,0,0) || diff --git a/sql/ha_ndbcluster_binlog.cc b/sql/ha_ndbcluster_binlog.cc index 07b0d907229..841dce2d832 100644 --- a/sql/ha_ndbcluster_binlog.cc +++ b/sql/ha_ndbcluster_binlog.cc @@ -799,7 +799,7 @@ static int ndbcluster_create_ndb_apply_status_table(THD *thd) " log_name VARCHAR(255) BINARY NOT NULL, " " start_pos BIGINT UNSIGNED NOT NULL, " " end_pos BIGINT UNSIGNED NOT NULL, " - " PRIMARY KEY USING HASH (server_id) ) ENGINE=NDB"); + " PRIMARY KEY USING HASH (server_id) ) ENGINE=NDB CHARACTER SET latin1"); const int no_print_error[5]= {ER_TABLE_EXISTS_ERROR, 701, @@ -860,7 +860,7 @@ static int ndbcluster_create_schema_table(THD *thd) " id INT UNSIGNED NOT NULL," " version INT UNSIGNED NOT NULL," " type INT UNSIGNED NOT NULL," - " PRIMARY KEY USING HASH (db,name) ) ENGINE=NDB"); + " PRIMARY KEY USING HASH (db,name) ) ENGINE=NDB CHARACTER SET latin1"); const int no_print_error[5]= {ER_TABLE_EXISTS_ERROR, 701, @@ -4036,6 +4036,7 @@ restart: i_ndb->setReportThreshEventFreeMem(ndb_report_thresh_binlog_mem_usage); bzero((char*) &row, sizeof(row)); + thd->variables.character_set_client= &my_charset_latin1; injector::transaction trans; // pass table map before epoch { diff --git a/sql/ha_partition.cc b/sql/ha_partition.cc index bd0004f92cb..a74322acee1 100644 --- a/sql/ha_partition.cc +++ b/sql/ha_partition.cc @@ -367,7 +367,7 @@ bool ha_partition::initialise_partition(MEM_ROOT *mem_root) HA_CAN_GEOMETRY, HA_CAN_FULLTEXT, HA_CAN_SQL_HANDLER, HA_DUPLICATE_POS, HA_CAN_INSERT_DELAYED is disabled until further investigated. */ - m_table_flags= (ulong)m_file[0]->table_flags(); + m_table_flags= (ulong)m_file[0]->ha_table_flags(); m_low_byte_first= m_file[0]->low_byte_first(); m_pkey_is_clustered= TRUE; file_array= m_file; @@ -382,7 +382,7 @@ bool ha_partition::initialise_partition(MEM_ROOT *mem_root) } if (!file->primary_key_is_clustered()) m_pkey_is_clustered= FALSE; - m_table_flags&= file->table_flags(); + m_table_flags&= file->ha_table_flags(); } while (*(++file_array)); m_table_flags&= ~(HA_CAN_GEOMETRY | HA_CAN_FULLTEXT | HA_DUPLICATE_POS | HA_CAN_SQL_HANDLER | HA_CAN_INSERT_DELAYED | @@ -616,7 +616,7 @@ int ha_partition::drop_partitions(const char *path) sub_elem->partition_name, name_variant); file= m_file[part]; DBUG_PRINT("info", ("Drop subpartition %s", part_name_buff)); - if ((ret_error= file->delete_table((const char *) part_name_buff))) + if ((ret_error= file->ha_delete_table(part_name_buff))) error= ret_error; if (deactivate_ddl_log_entry(sub_elem->log_entry->entry_pos)) error= 1; @@ -629,7 +629,7 @@ int ha_partition::drop_partitions(const char *path) TRUE); file= m_file[i]; DBUG_PRINT("info", ("Drop partition %s", part_name_buff)); - if ((ret_error= file->delete_table((const char *) part_name_buff))) + if ((ret_error= file->ha_delete_table(part_name_buff))) error= ret_error; if (deactivate_ddl_log_entry(part_elem->log_entry->entry_pos)) error= 1; @@ -707,7 +707,7 @@ int ha_partition::rename_partitions(const char *path) sub_elem->partition_name, NORMAL_PART_NAME); DBUG_PRINT("info", ("Delete subpartition %s", norm_name_buff)); - if ((ret_error= file->delete_table((const char *) norm_name_buff))) + if ((ret_error= file->ha_delete_table(norm_name_buff))) error= ret_error; else if (deactivate_ddl_log_entry(sub_elem->log_entry->entry_pos)) error= 1; @@ -722,7 +722,7 @@ int ha_partition::rename_partitions(const char *path) part_elem->partition_name, NORMAL_PART_NAME, TRUE); DBUG_PRINT("info", ("Delete partition %s", norm_name_buff)); - if ((ret_error= file->delete_table((const char *) norm_name_buff))) + if ((ret_error= file->ha_delete_table(norm_name_buff))) error= ret_error; else if (deactivate_ddl_log_entry(part_elem->log_entry->entry_pos)) error= 1; @@ -778,7 +778,7 @@ int ha_partition::rename_partitions(const char *path) { file= m_reorged_file[part_count++]; DBUG_PRINT("info", ("Delete subpartition %s", norm_name_buff)); - if ((ret_error= file->delete_table((const char *) norm_name_buff))) + if ((ret_error= file->ha_delete_table(norm_name_buff))) error= ret_error; else if (deactivate_ddl_log_entry(sub_elem->log_entry->entry_pos)) error= 1; @@ -791,8 +791,8 @@ int ha_partition::rename_partitions(const char *path) TEMP_PART_NAME); DBUG_PRINT("info", ("Rename subpartition from %s to %s", part_name_buff, norm_name_buff)); - if ((ret_error= file->rename_table((const char *) part_name_buff, - (const char *) norm_name_buff))) + if ((ret_error= file->ha_rename_table(part_name_buff, + norm_name_buff))) error= ret_error; else if (deactivate_ddl_log_entry(sub_elem->log_entry->entry_pos)) error= 1; @@ -809,7 +809,7 @@ int ha_partition::rename_partitions(const char *path) { file= m_reorged_file[part_count++]; DBUG_PRINT("info", ("Delete partition %s", norm_name_buff)); - if ((ret_error= file->delete_table((const char *) norm_name_buff))) + if ((ret_error= file->ha_delete_table(norm_name_buff))) error= ret_error; else if (deactivate_ddl_log_entry(part_elem->log_entry->entry_pos)) error= 1; @@ -821,8 +821,8 @@ int ha_partition::rename_partitions(const char *path) TRUE); DBUG_PRINT("info", ("Rename partition from %s to %s", part_name_buff, norm_name_buff)); - if ((ret_error= file->rename_table((const char *) part_name_buff, - (const char *) norm_name_buff))) + if ((ret_error= file->ha_rename_table(part_name_buff, + norm_name_buff))) error= ret_error; else if (deactivate_ddl_log_entry(part_elem->log_entry->entry_pos)) error= 1; @@ -1036,9 +1036,9 @@ static int handle_opt_part(THD *thd, HA_CHECK_OPT *check_opt, DBUG_PRINT("enter", ("flag = %u", flag)); if (flag == OPTIMIZE_PARTS) - error= file->optimize(thd, check_opt); + error= file->ha_optimize(thd, check_opt); else if (flag == ANALYZE_PARTS) - error= file->analyze(thd, check_opt); + error= file->ha_analyze(thd, check_opt); else if (flag == CHECK_PARTS) error= file->ha_check(thd, check_opt); else if (flag == REPAIR_PARTS) @@ -1139,7 +1139,7 @@ int ha_partition::prepare_new_partition(TABLE *tbl, if ((error= set_up_table_before_create(tbl, part_name, create_info, 0, p_elem))) goto error; - if ((error= file->create(part_name, tbl, create_info))) + if ((error= file->ha_create(part_name, tbl, create_info))) goto error; create_flag= TRUE; if ((error= file->ha_open(tbl, part_name, m_mode, m_open_test_lock))) @@ -1150,13 +1150,13 @@ int ha_partition::prepare_new_partition(TABLE *tbl, assumes that external_lock() is last call that may fail here. Otherwise see description for cleanup_new_partition(). */ - if ((error= file->external_lock(current_thd, m_lock_type))) + if ((error= file->ha_external_lock(current_thd, m_lock_type))) goto error; DBUG_RETURN(0); error: if (create_flag) - VOID(file->delete_table(part_name)); + VOID(file->ha_delete_table(part_name)); DBUG_RETURN(error); } @@ -1574,14 +1574,18 @@ int ha_partition::copy_partitions(ulonglong *copied, ulonglong *deleted) } else { + THD *thd= ha_thd(); /* Copy record to new handler */ copied++; - if ((result= m_new_file[new_part]->write_row(m_rec0))) + tmp_disable_binlog(thd); /* Do not replicate the low-level changes. */ + result= m_new_file[new_part]->ha_write_row(m_rec0); + reenable_binlog(thd); + if (result) goto error; } } late_extra_no_cache(reorg_part); - file->rnd_end(); + file->ha_rnd_end(); reorg_part++; } DBUG_RETURN(FALSE); @@ -1698,16 +1702,15 @@ uint ha_partition::del_ren_cre_table(const char *from, { // Rename branch create_partition_name(to_buff, to, name_buffer_ptr, NORMAL_PART_NAME, FALSE); - error= (*file)->rename_table((const char*) from_buff, - (const char*) to_buff); + error= (*file)->ha_rename_table(from_buff, to_buff); } else if (table_arg == NULL) // delete branch - error= (*file)->delete_table((const char*) from_buff); + error= (*file)->ha_delete_table(from_buff); else { if ((error= set_up_table_before_create(table_arg, from_buff, create_info, i, NULL)) || - ((error= (*file)->create(from_buff, table_arg, create_info)))) + ((error= (*file)->ha_create(from_buff, table_arg, create_info)))) goto create_error; } name_buffer_ptr= strend(name_buffer_ptr) + 1; @@ -1722,7 +1725,7 @@ create_error: { create_partition_name(from_buff, from, name_buffer_ptr, NORMAL_PART_NAME, FALSE); - VOID((*file)->delete_table((const char*) from_buff)); + VOID((*file)->ha_delete_table((const char*) from_buff)); name_buffer_ptr= strend(name_buffer_ptr) + 1; } DBUG_RETURN(error); @@ -2314,7 +2317,7 @@ int ha_partition::open(const char *name, int mode, uint test_if_locked) } /* Recalculate table flags as they may change after open */ - m_table_flags= m_file[0]->table_flags(); + m_table_flags= m_file[0]->ha_table_flags(); file= m_file; do { @@ -2326,7 +2329,7 @@ int ha_partition::open(const char *name, int mode, uint test_if_locked) m_no_locks+= (*file)->lock_count(); name_buffer_ptr+= strlen(name_buffer_ptr) + 1; set_if_bigger(ref_length, ((*file)->ref_length)); - m_table_flags&= (*file)->table_flags(); + m_table_flags&= (*file)->ha_table_flags(); } while (*(++file)); m_table_flags&= ~(HA_CAN_GEOMETRY | HA_CAN_FULLTEXT | HA_DUPLICATE_POS | HA_CAN_SQL_HANDLER | HA_CAN_INSERT_DELAYED); @@ -2482,7 +2485,7 @@ repeat: { DBUG_PRINT("info", ("external_lock(thd, %d) iteration %d", lock_type, (int) (file - m_file))); - if ((error= (*file)->external_lock(thd, lock_type))) + if ((error= (*file)->ha_external_lock(thd, lock_type))) { if (F_UNLCK != lock_type) goto err_handler; @@ -2501,7 +2504,7 @@ repeat: err_handler: while (file-- != m_file) { - (*file)->external_lock(thd, F_UNLCK); + (*file)->ha_external_lock(thd, F_UNLCK); } DBUG_RETURN(error); } @@ -2694,6 +2697,7 @@ int ha_partition::write_row(uchar * buf) longlong func_value; bool autoincrement_lock= FALSE; my_bitmap_map *old_map; + THD *thd= ha_thd(); #ifdef NOT_NEEDED uchar *rec0= m_rec0; #endif @@ -2765,7 +2769,9 @@ int ha_partition::write_row(uchar * buf) } m_last_part= part_id; DBUG_PRINT("info", ("Insert in partition %d", part_id)); - error= m_file[part_id]->write_row(buf); + tmp_disable_binlog(thd); /* Do not replicate the low-level changes. */ + error= m_file[part_id]->ha_write_row(buf); + reenable_binlog(thd); exit: if (autoincrement_lock) pthread_mutex_unlock(&table_share->mutex); @@ -2806,6 +2812,7 @@ exit: int ha_partition::update_row(const uchar *old_data, uchar *new_data) { + THD *thd= ha_thd(); uint32 new_part_id, old_part_id; int error= 0; longlong func_value; @@ -2840,16 +2847,25 @@ int ha_partition::update_row(const uchar *old_data, uchar *new_data) if (new_part_id == old_part_id) { DBUG_PRINT("info", ("Update in partition %d", new_part_id)); - error= m_file[new_part_id]->update_row(old_data, new_data); + tmp_disable_binlog(thd); /* Do not replicate the low-level changes. */ + error= m_file[new_part_id]->ha_update_row(old_data, new_data); + reenable_binlog(thd); goto exit; } else { DBUG_PRINT("info", ("Update from partition %d to partition %d", old_part_id, new_part_id)); - if ((error= m_file[new_part_id]->write_row(new_data))) + tmp_disable_binlog(thd); /* Do not replicate the low-level changes. */ + error= m_file[new_part_id]->ha_write_row(new_data); + reenable_binlog(thd); + if (error) goto exit; - if ((error= m_file[old_part_id]->delete_row(old_data))) + + tmp_disable_binlog(thd); /* Do not replicate the low-level changes. */ + error= m_file[old_part_id]->ha_delete_row(old_data); + reenable_binlog(thd); + if (error) { #ifdef IN_THE_FUTURE (void) m_file[new_part_id]->delete_last_inserted_row(new_data); @@ -2896,6 +2912,7 @@ int ha_partition::delete_row(const uchar *buf) { uint32 part_id; int error; + THD *thd= ha_thd(); DBUG_ENTER("ha_partition::delete_row"); if ((error= get_part_for_delete(buf, m_rec0, m_part_info, &part_id))) @@ -2903,7 +2920,10 @@ int ha_partition::delete_row(const uchar *buf) DBUG_RETURN(error); } m_last_part= part_id; - DBUG_RETURN(m_file[part_id]->delete_row(buf)); + tmp_disable_binlog(thd); + error= m_file[part_id]->ha_delete_row(buf); + reenable_binlog(thd); + DBUG_RETURN(error); } @@ -2938,7 +2958,7 @@ int ha_partition::delete_all_rows() file= m_file; do { - if ((error= (*file)->delete_all_rows())) + if ((error= (*file)->ha_delete_all_rows())) DBUG_RETURN(error); } while (*(++file)); DBUG_RETURN(0); @@ -3980,7 +4000,7 @@ int ha_partition::partition_scan_set_up(uchar * buf, bool idx_read_flag) int ha_partition::handle_unordered_next(uchar *buf, bool is_next_same) { - handler *file= file= m_file[m_part_spec.start_part]; + handler *file= m_file[m_part_spec.start_part]; int error; DBUG_ENTER("ha_partition::handle_unordered_next"); @@ -3999,7 +4019,7 @@ int ha_partition::handle_unordered_next(uchar *buf, bool is_next_same) } else if (!(error= file->index_next(buf))) { - if (!(file->table_flags() & HA_READ_ORDER) || + if (!(file->ha_table_flags() & HA_READ_ORDER) || compare_key(end_range) <= 0) { m_last_part= m_part_spec.start_part; @@ -4077,7 +4097,7 @@ int ha_partition::handle_unordered_scan_next_partition(uchar * buf) } if (!error) { - if (!(file->table_flags() & HA_READ_ORDER) || + if (!(file->ha_table_flags() & HA_READ_ORDER) || compare_key(end_range) <= 0) { m_last_part= i; @@ -5003,7 +5023,7 @@ int ha_partition::reset(void) file= m_file; do { - if ((tmp= (*file)->reset())) + if ((tmp= (*file)->ha_reset())) result= tmp; } while (*(++file)); DBUG_RETURN(result); @@ -5645,7 +5665,7 @@ void ha_partition::release_auto_increment() for (uint i= 0; i < m_tot_parts; i++) { - m_file[i]->release_auto_increment(); + m_file[i]->ha_release_auto_increment(); } DBUG_VOID_RETURN; } @@ -5681,7 +5701,7 @@ int ha_partition::disable_indexes(uint mode) for (file= m_file; *file; file++) { - if ((error= (*file)->disable_indexes(mode))) + if ((error= (*file)->ha_disable_indexes(mode))) break; } return error; @@ -5705,7 +5725,7 @@ int ha_partition::enable_indexes(uint mode) for (file= m_file; *file; file++) { - if ((error= (*file)->enable_indexes(mode))) + if ((error= (*file)->ha_enable_indexes(mode))) break; } return error; diff --git a/sql/handler.cc b/sql/handler.cc index 6489041eaf1..421708e5958 100644 --- a/sql/handler.cc +++ b/sql/handler.cc @@ -1477,7 +1477,7 @@ int ha_delete_table(THD *thd, handlerton *table_type, const char *path, DBUG_RETURN(ENOENT); path= check_lowercase_names(file, path, tmp_path); - if ((error= file->delete_table(path)) && generate_warning) + if ((error= file->ha_delete_table(path)) && generate_warning) { /* Because file->print_error() use my_error() to generate the error message @@ -1496,8 +1496,7 @@ int ha_delete_table(THD *thd, handlerton *table_type, const char *path, dummy_share.table_name.length= strlen(alias); dummy_table.alias= alias; - file->table_share= &dummy_share; - file->table= &dummy_table; + file->change_table_ptr(&dummy_table, &dummy_share); thd->push_internal_handler(&ha_delete_table_error_handler); file->print_error(error, 0); @@ -2506,6 +2505,12 @@ int handler::ha_check(THD *thd, HA_CHECK_OPT *check_opt) } +/** + Repair table: public interface. + + @sa handler::repair() +*/ + int handler::ha_repair(THD* thd, HA_CHECK_OPT* check_opt) { int result; @@ -2516,6 +2521,328 @@ int handler::ha_repair(THD* thd, HA_CHECK_OPT* check_opt) /** + Bulk update row: public interface. + + @sa handler::bulk_update_row() +*/ + +int +handler::ha_bulk_update_row(const uchar *old_data, uchar *new_data, + uint *dup_key_found) +{ + return bulk_update_row(old_data, new_data, dup_key_found); +} + + +/** + Delete all rows: public interface. + + @sa handler::delete_all_rows() +*/ + +int +handler::ha_delete_all_rows() +{ + return delete_all_rows(); +} + + +/** + Reset auto increment: public interface. + + @sa handler::reset_auto_increment() +*/ + +int +handler::ha_reset_auto_increment(ulonglong value) +{ + return reset_auto_increment(value); +} + + +/** + Backup table: public interface. + + @sa handler::backup() +*/ + +int +handler::ha_backup(THD* thd, HA_CHECK_OPT* check_opt) +{ + return backup(thd, check_opt); +} + + +/** + Restore table: public interface. + + @sa handler::restore() +*/ + +int +handler::ha_restore(THD* thd, HA_CHECK_OPT* check_opt) +{ + return restore(thd, check_opt); +} + + +/** + Optimize table: public interface. + + @sa handler::optimize() +*/ + +int +handler::ha_optimize(THD* thd, HA_CHECK_OPT* check_opt) +{ + return optimize(thd, check_opt); +} + + +/** + Analyze table: public interface. + + @sa handler::analyze() +*/ + +int +handler::ha_analyze(THD* thd, HA_CHECK_OPT* check_opt) +{ + return analyze(thd, check_opt); +} + + +/** + Check and repair table: public interface. + + @sa handler::check_and_repair() +*/ + +bool +handler::ha_check_and_repair(THD *thd) +{ + return check_and_repair(thd); +} + + +/** + Disable indexes: public interface. + + @sa handler::disable_indexes() +*/ + +int +handler::ha_disable_indexes(uint mode) +{ + return disable_indexes(mode); +} + + +/** + Enable indexes: public interface. + + @sa handler::enable_indexes() +*/ + +int +handler::ha_enable_indexes(uint mode) +{ + return enable_indexes(mode); +} + + +/** + Discard or import tablespace: public interface. + + @sa handler::discard_or_import_tablespace() +*/ + +int +handler::ha_discard_or_import_tablespace(my_bool discard) +{ + return discard_or_import_tablespace(discard); +} + + +/** + Prepare for alter: public interface. + + Called to prepare an *online* ALTER. + + @sa handler::prepare_for_alter() +*/ + +void +handler::ha_prepare_for_alter() +{ + prepare_for_alter(); +} + + +/** + Rename table: public interface. + + @sa handler::rename_table() +*/ + +int +handler::ha_rename_table(const char *from, const char *to) +{ + return rename_table(from, to); +} + + +/** + Delete table: public interface. + + @sa handler::delete_table() +*/ + +int +handler::ha_delete_table(const char *name) +{ + return delete_table(name); +} + + +/** + Drop table in the engine: public interface. + + @sa handler::drop_table() +*/ + +void +handler::ha_drop_table(const char *name) +{ + return drop_table(name); +} + + +/** + Create a table in the engine: public interface. + + @sa handler::create() +*/ + +int +handler::ha_create(const char *name, TABLE *form, HA_CREATE_INFO *info) +{ + return create(name, form, info); +} + + +/** + Create handler files for CREATE TABLE: public interface. + + @sa handler::create_handler_files() +*/ + +int +handler::ha_create_handler_files(const char *name, const char *old_name, + int action_flag, HA_CREATE_INFO *info) +{ + return create_handler_files(name, old_name, action_flag, info); +} + + +/** + Change partitions: public interface. + + @sa handler::change_partitions() +*/ + +int +handler::ha_change_partitions(HA_CREATE_INFO *create_info, + const char *path, + ulonglong *copied, + ulonglong *deleted, + const uchar *pack_frm_data, + size_t pack_frm_len) +{ + return change_partitions(create_info, path, copied, deleted, + pack_frm_data, pack_frm_len); +} + + +/** + Drop partitions: public interface. + + @sa handler::drop_partitions() +*/ + +int +handler::ha_drop_partitions(const char *path) +{ + return drop_partitions(path); +} + + +/** + Rename partitions: public interface. + + @sa handler::rename_partitions() +*/ + +int +handler::ha_rename_partitions(const char *path) +{ + return rename_partitions(path); +} + + +/** + Optimize partitions: public interface. + + @sa handler::optimize_partitions() +*/ + +int +handler::ha_optimize_partitions(THD *thd) +{ + return optimize_partitions(thd); +} + + +/** + Analyze partitions: public interface. + + @sa handler::analyze_partitions() +*/ + +int +handler::ha_analyze_partitions(THD *thd) +{ + return analyze_partitions(thd); +} + + +/** + Check partitions: public interface. + + @sa handler::check_partitions() +*/ + +int +handler::ha_check_partitions(THD *thd) +{ + return check_partitions(thd); +} + + +/** + Repair partitions: public interface. + + @sa handler::repair_partitions() +*/ + +int +handler::ha_repair_partitions(THD *thd) +{ + return repair_partitions(thd); +} + + +/** Tell the storage engine that it is allowed to "disable transaction" in the handler. It is a hint that ACID is not required - it is used in NDB for ALTER TABLE, for example, when data are copied to temporary table. @@ -2655,7 +2982,7 @@ int ha_create_table(THD *thd, const char *path, name= check_lowercase_names(table.file, share.path.str, name_buff); - error= table.file->create(name, &table, create_info); + error= table.file->ha_create(name, &table, create_info); VOID(closefrm(&table, 0)); if (error) { @@ -2726,7 +3053,7 @@ int ha_create_table_from_engine(THD* thd, const char *db, const char *name) create_info.table_options|= HA_OPTION_CREATE_FROM_ENGINE; check_lowercase_names(table.file, path, path); - error=table.file->create(path,&table,&create_info); + error=table.file->ha_create(path, &table, &create_info); VOID(closefrm(&table, 1)); DBUG_RETURN(error != 0); diff --git a/sql/handler.h b/sql/handler.h index 75c56e8339e..8e3073eb6ba 100644 --- a/sql/handler.h +++ b/sql/handler.h @@ -967,10 +967,6 @@ uint calculate_key_len(TABLE *, uint, const uchar *, key_part_map); class handler :public Sql_alloc { - friend class ha_partition; - friend int ha_delete_table(THD*,handlerton*,const char*,const char*, - const char*,bool); - public: typedef ulonglong Table_flags; protected: @@ -1118,6 +1114,40 @@ public: estimation_rows_to_insert= 0; return end_bulk_insert(); } + int ha_bulk_update_row(const uchar *old_data, uchar *new_data, + uint *dup_key_found); + int ha_delete_all_rows(); + int ha_reset_auto_increment(ulonglong value); + int ha_backup(THD* thd, HA_CHECK_OPT* check_opt); + int ha_restore(THD* thd, HA_CHECK_OPT* check_opt); + int ha_optimize(THD* thd, HA_CHECK_OPT* check_opt); + int ha_analyze(THD* thd, HA_CHECK_OPT* check_opt); + bool ha_check_and_repair(THD *thd); + int ha_disable_indexes(uint mode); + int ha_enable_indexes(uint mode); + int ha_discard_or_import_tablespace(my_bool discard); + void ha_prepare_for_alter(); + int ha_rename_table(const char *from, const char *to); + int ha_delete_table(const char *name); + void ha_drop_table(const char *name); + + int ha_create(const char *name, TABLE *form, HA_CREATE_INFO *info); + + int ha_create_handler_files(const char *name, const char *old_name, + int action_flag, HA_CREATE_INFO *info); + + int ha_change_partitions(HA_CREATE_INFO *create_info, + const char *path, + ulonglong *copied, + ulonglong *deleted, + const uchar *pack_frm_data, + size_t pack_frm_len); + int ha_drop_partitions(const char *path); + int ha_rename_partitions(const char *path); + int ha_optimize_partitions(THD *thd); + int ha_analyze_partitions(THD *thd); + int ha_check_partitions(THD *thd); + int ha_repair_partitions(THD *thd); void adjust_next_insert_id_after_explicit_value(ulonglong nr); int update_auto_increment(); @@ -1204,25 +1234,6 @@ public: */ virtual bool start_bulk_delete() { return 1; } /** - This method is similar to update_row, however the handler doesn't need - to execute the updates at this point in time. The handler can be certain - that another call to bulk_update_row will occur OR a call to - exec_bulk_update before the set of updates in this query is concluded. - - @param old_data Old record - @param new_data New record - @param dup_key_found Number of duplicate keys found - - @retval 0 Bulk delete used by handler - @retval 1 Bulk delete not used, normal operation used - */ - virtual int bulk_update_row(const uchar *old_data, uchar *new_data, - uint *dup_key_found) - { - DBUG_ASSERT(FALSE); - return HA_ERR_WRONG_COMMAND; - } - /** After this call all outstanding updates must be performed. The number of duplicate key errors are reported in the duplicate key parameter. It is allowed to continue to the batched update after this call, the @@ -1369,14 +1380,6 @@ public: virtual void try_semi_consistent_read(bool) {} virtual void unlock_row() {} virtual int start_stmt(THD *thd, thr_lock_type lock_type) {return 0;} - /** - This is called to delete all rows in a table - If the handler don't support this, then this function will - return HA_ERR_WRONG_COMMAND and MySQL will delete the rows one - by one. - */ - virtual int delete_all_rows() - { return (my_errno=HA_ERR_WRONG_COMMAND); } virtual void get_auto_increment(ulonglong offset, ulonglong increment, ulonglong nb_desired_values, ulonglong *first_value, @@ -1401,42 +1404,17 @@ public: next_insert_id= (prev_insert_id > 0) ? prev_insert_id : insert_id_for_cur_row; } - /** - Reset the auto-increment counter to the given value, i.e. the next row - inserted will get the given value. This is called e.g. after TRUNCATE - is emulated by doing a 'DELETE FROM t'. HA_ERR_WRONG_COMMAND is - returned by storage engines that don't support this operation. - */ - virtual int reset_auto_increment(ulonglong value) - { return HA_ERR_WRONG_COMMAND; } virtual void update_create_info(HA_CREATE_INFO *create_info) {} int check_old_types(); - virtual int backup(THD* thd, HA_CHECK_OPT* check_opt) - { return HA_ADMIN_NOT_IMPLEMENTED; } - /** - Restore assumes .frm file must exist, and that generate_table() has been - called; It will just copy the data file and run repair. - */ - virtual int restore(THD* thd, HA_CHECK_OPT* check_opt) - { return HA_ADMIN_NOT_IMPLEMENTED; } - virtual int optimize(THD* thd, HA_CHECK_OPT* check_opt) - { return HA_ADMIN_NOT_IMPLEMENTED; } - virtual int analyze(THD* thd, HA_CHECK_OPT* check_opt) - { return HA_ADMIN_NOT_IMPLEMENTED; } virtual int assign_to_keycache(THD* thd, HA_CHECK_OPT* check_opt) { return HA_ADMIN_NOT_IMPLEMENTED; } virtual int preload_keys(THD* thd, HA_CHECK_OPT* check_opt) { return HA_ADMIN_NOT_IMPLEMENTED; } /* end of the list of admin commands */ - virtual bool check_and_repair(THD *thd) { return HA_ERR_WRONG_COMMAND; } virtual int dump(THD* thd, int fd = -1) { return HA_ERR_WRONG_COMMAND; } - virtual int disable_indexes(uint mode) { return HA_ERR_WRONG_COMMAND; } - virtual int enable_indexes(uint mode) { return HA_ERR_WRONG_COMMAND; } virtual int indexes_are_disabled(void) {return 0;} - virtual int discard_or_import_tablespace(my_bool discard) - {return HA_ERR_WRONG_COMMAND;} virtual int net_read_dump(NET* net) { return HA_ERR_WRONG_COMMAND; } virtual char *update_table_comment(const char * comment) { return (char*) comment;} @@ -1494,7 +1472,6 @@ public: virtual ulong index_flags(uint idx, uint part, bool all_parts) const =0; - virtual void prepare_for_alter() { return; } virtual int add_index(TABLE *table_arg, KEY *key_info, uint num_of_keys) { return (HA_ERR_WRONG_COMMAND); } virtual int prepare_drop_index(TABLE *table_arg, uint *key_num, @@ -1526,46 +1503,11 @@ public: virtual bool is_crashed() const { return 0; } virtual bool auto_repair() const { return 0; } - /** - default rename_table() and delete_table() rename/delete files with a - given name and extensions from bas_ext() - */ - virtual int rename_table(const char *from, const char *to); - virtual int delete_table(const char *name); - virtual void drop_table(const char *name); - - virtual int create(const char *name, TABLE *form, HA_CREATE_INFO *info)=0; - #define CHF_CREATE_FLAG 0 #define CHF_DELETE_FLAG 1 #define CHF_RENAME_FLAG 2 #define CHF_INDEX_FLAG 3 - virtual int create_handler_files(const char *name, const char *old_name, - int action_flag, - HA_CREATE_INFO *create_info) - { return FALSE; } - - virtual int change_partitions(HA_CREATE_INFO *create_info, - const char *path, - ulonglong *copied, - ulonglong *deleted, - const uchar *pack_frm_data, - size_t pack_frm_len) - { return HA_ERR_WRONG_COMMAND; } - virtual int drop_partitions(const char *path) - { return HA_ERR_WRONG_COMMAND; } - virtual int rename_partitions(const char *path) - { return HA_ERR_WRONG_COMMAND; } - virtual int optimize_partitions(THD *thd) - { return HA_ERR_WRONG_COMMAND; } - virtual int analyze_partitions(THD *thd) - { return HA_ERR_WRONG_COMMAND; } - virtual int check_partitions(THD *thd) - { return HA_ERR_WRONG_COMMAND; } - virtual int repair_partitions(THD *thd) - { return HA_ERR_WRONG_COMMAND; } - /** @note lock_count() can return > 1 if the table is MERGE or partitioned. */ @@ -1680,22 +1622,6 @@ public: uint table_changes) { return COMPATIBLE_DATA_NO; } - /** These are only called from sql_select for internal temporary tables */ - virtual int write_row(uchar *buf __attribute__((unused))) - { - return HA_ERR_WRONG_COMMAND; - } - - virtual int update_row(const uchar *old_data __attribute__((unused)), - uchar *new_data __attribute__((unused))) - { - return HA_ERR_WRONG_COMMAND; - } - - virtual int delete_row(const uchar *buf __attribute__((unused))) - { - return HA_ERR_WRONG_COMMAND; - } /** use_hidden_primary_key() is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined @@ -1708,6 +1634,16 @@ protected: void ha_statistic_increment(ulong SSV::*offset) const; void **ha_data(THD *) const; THD *ha_thd(void) const; + + /** + Default rename_table() and delete_table() rename/delete files with a + given name and extensions from bas_ext(). + + These methods can be overridden, but their default implementation + provide useful functionality. + */ + virtual int rename_table(const char *from, const char *to); + virtual int delete_table(const char *name); private: /* Low-level primitives for storage engines. These should be @@ -1727,6 +1663,21 @@ private: */ virtual int rnd_init(bool scan)= 0; virtual int rnd_end() { return 0; } + virtual int write_row(uchar *buf __attribute__((unused))) + { + return HA_ERR_WRONG_COMMAND; + } + + virtual int update_row(const uchar *old_data __attribute__((unused)), + uchar *new_data __attribute__((unused))) + { + return HA_ERR_WRONG_COMMAND; + } + + virtual int delete_row(const uchar *buf __attribute__((unused))) + { + return HA_ERR_WRONG_COMMAND; + } /** Reset state of file to after 'open'. This function is called after every statement for all tables used @@ -1782,6 +1733,85 @@ private: { return HA_ERR_WRONG_COMMAND; } virtual int index_read_last(uchar * buf, const uchar * key, uint key_len) { return (my_errno= HA_ERR_WRONG_COMMAND); } + /** + This method is similar to update_row, however the handler doesn't need + to execute the updates at this point in time. The handler can be certain + that another call to bulk_update_row will occur OR a call to + exec_bulk_update before the set of updates in this query is concluded. + + @param old_data Old record + @param new_data New record + @param dup_key_found Number of duplicate keys found + + @retval 0 Bulk delete used by handler + @retval 1 Bulk delete not used, normal operation used + */ + virtual int bulk_update_row(const uchar *old_data, uchar *new_data, + uint *dup_key_found) + { + DBUG_ASSERT(FALSE); + return HA_ERR_WRONG_COMMAND; + } + /** + This is called to delete all rows in a table + If the handler don't support this, then this function will + return HA_ERR_WRONG_COMMAND and MySQL will delete the rows one + by one. + */ + virtual int delete_all_rows() + { return (my_errno=HA_ERR_WRONG_COMMAND); } + /** + Reset the auto-increment counter to the given value, i.e. the next row + inserted will get the given value. This is called e.g. after TRUNCATE + is emulated by doing a 'DELETE FROM t'. HA_ERR_WRONG_COMMAND is + returned by storage engines that don't support this operation. + */ + virtual int reset_auto_increment(ulonglong value) + { return HA_ERR_WRONG_COMMAND; } + virtual int backup(THD* thd, HA_CHECK_OPT* check_opt) + { return HA_ADMIN_NOT_IMPLEMENTED; } + /** + Restore assumes .frm file must exist, and that generate_table() has been + called; It will just copy the data file and run repair. + */ + virtual int restore(THD* thd, HA_CHECK_OPT* check_opt) + { return HA_ADMIN_NOT_IMPLEMENTED; } + virtual int optimize(THD* thd, HA_CHECK_OPT* check_opt) + { return HA_ADMIN_NOT_IMPLEMENTED; } + virtual int analyze(THD* thd, HA_CHECK_OPT* check_opt) + { return HA_ADMIN_NOT_IMPLEMENTED; } + virtual bool check_and_repair(THD *thd) { return HA_ERR_WRONG_COMMAND; } + virtual int disable_indexes(uint mode) { return HA_ERR_WRONG_COMMAND; } + virtual int enable_indexes(uint mode) { return HA_ERR_WRONG_COMMAND; } + virtual int discard_or_import_tablespace(my_bool discard) + { return (my_errno=HA_ERR_WRONG_COMMAND); } + virtual void prepare_for_alter() { return; } + virtual void drop_table(const char *name); + virtual int create(const char *name, TABLE *form, HA_CREATE_INFO *info)=0; + + virtual int create_handler_files(const char *name, const char *old_name, + int action_flag, HA_CREATE_INFO *info) + { return FALSE; } + + virtual int change_partitions(HA_CREATE_INFO *create_info, + const char *path, + ulonglong *copied, + ulonglong *deleted, + const uchar *pack_frm_data, + size_t pack_frm_len) + { return HA_ERR_WRONG_COMMAND; } + virtual int drop_partitions(const char *path) + { return HA_ERR_WRONG_COMMAND; } + virtual int rename_partitions(const char *path) + { return HA_ERR_WRONG_COMMAND; } + virtual int optimize_partitions(THD *thd) + { return HA_ERR_WRONG_COMMAND; } + virtual int analyze_partitions(THD *thd) + { return HA_ERR_WRONG_COMMAND; } + virtual int check_partitions(THD *thd) + { return HA_ERR_WRONG_COMMAND; } + virtual int repair_partitions(THD *thd) + { return HA_ERR_WRONG_COMMAND; } }; diff --git a/sql/item.cc b/sql/item.cc index c33e5da0ca4..c546badd8e7 100644 --- a/sql/item.cc +++ b/sql/item.cc @@ -3418,7 +3418,7 @@ static Item** find_field_in_group_list(Item *find_item, ORDER *group_list) @param thd current thread @param ref column reference being resolved - @param select the sub-select that ref is resolved against + @param select the select that ref is resolved against @note The resolution procedure is: @@ -3478,6 +3478,7 @@ resolve_ref_in_select_and_group(THD *thd, Item_ident *ref, SELECT_LEX *select) } if (thd->variables.sql_mode & MODE_ONLY_FULL_GROUP_BY && + select->having_fix_field && select_ref != not_found_item && !group_by_ref) { /* @@ -4460,7 +4461,7 @@ Field *Item::tmp_table_field_from_field_type(TABLE *table, bool fixed_length) break; case MYSQL_TYPE_NEWDATE: case MYSQL_TYPE_DATE: - field= new Field_date(maybe_null, name, &my_charset_bin); + field= new Field_newdate(maybe_null, name, &my_charset_bin); break; case MYSQL_TYPE_TIME: field= new Field_time(maybe_null, name, &my_charset_bin); @@ -6695,6 +6696,8 @@ enum_field_types Item_type_holder::get_real_type(Item *item) */ Field *field= ((Item_field *) item)->field; enum_field_types type= field->real_type(); + if (field->is_created_from_null_item) + return MYSQL_TYPE_NULL; /* work around about varchar type field detection */ if (type == MYSQL_TYPE_STRING && field->type() == MYSQL_TYPE_VAR_STRING) return MYSQL_TYPE_VAR_STRING; @@ -6948,6 +6951,8 @@ Field *Item_type_holder::make_field_by_type(TABLE *table) if (field) field->init(table); return field; + case MYSQL_TYPE_NULL: + return make_string_field(table); default: break; } diff --git a/sql/item_cmpfunc.cc b/sql/item_cmpfunc.cc index 62ac7bc4751..dc868376796 100644 --- a/sql/item_cmpfunc.cc +++ b/sql/item_cmpfunc.cc @@ -975,7 +975,7 @@ get_datetime_value(THD *thd, Item ***item_arg, Item **cache_arg, *is_null= item->null_value; } if (*is_null) - return -1; + return ~(ulonglong) 0; /* Convert strings to the integer DATE/DATETIME representation. Even if both dates provided in strings we can't compare them directly as diff --git a/sql/item_func.cc b/sql/item_func.cc index aa7f5bbb0d5..748624c0357 100644 --- a/sql/item_func.cc +++ b/sql/item_func.cc @@ -1995,7 +1995,7 @@ void Item_func_round::fix_length_and_dec() int length_increase= ((decimals_delta <= 0) || truncate) ? 0:1; precision-= decimals_delta - length_increase; - decimals= decimals_to_set; + decimals= min(decimals_to_set, DECIMAL_MAX_SCALE); max_length= my_decimal_precision_to_length(precision, decimals, unsigned_flag); break; @@ -2094,18 +2094,18 @@ my_decimal *Item_func_round::decimal_op(my_decimal *decimal_value) { my_decimal val, *value= args[0]->val_decimal(&val); longlong dec= args[1]->val_int(); - if (dec > 0 || (dec < 0 && args[1]->unsigned_flag)) - { + if (dec >= 0 || args[1]->unsigned_flag) dec= min((ulonglong) dec, decimals); - decimals= (uint8) dec; // to get correct output - } else if (dec < INT_MIN) dec= INT_MIN; if (!(null_value= (args[0]->null_value || args[1]->null_value || my_decimal_round(E_DEC_FATAL_ERROR, value, (int) dec, - truncate, decimal_value) > 1))) + truncate, decimal_value) > 1))) + { + decimal_value->frac= decimals; return decimal_value; + } return 0; } diff --git a/sql/item_sum.cc b/sql/item_sum.cc index 19c71cbc48d..27b964a9e15 100644 --- a/sql/item_sum.cc +++ b/sql/item_sum.cc @@ -632,7 +632,7 @@ Field *Item_sum_hybrid::create_tmp_field(bool group, TABLE *table, */ switch (args[0]->field_type()) { case MYSQL_TYPE_DATE: - field= new Field_date(maybe_null, name, collation.collation); + field= new Field_newdate(maybe_null, name, collation.collation); break; case MYSQL_TYPE_TIME: field= new Field_time(maybe_null, name, collation.collation); @@ -2617,7 +2617,7 @@ void Item_sum_count_distinct::clear() else if (table) { table->file->extra(HA_EXTRA_NO_CACHE); - table->file->delete_all_rows(); + table->file->ha_delete_all_rows(); table->file->extra(HA_EXTRA_WRITE_CACHE); } } @@ -2863,44 +2863,51 @@ String *Item_sum_udf_str::val_str(String *str) concat of values from "group by" operation BUGS - DISTINCT and ORDER BY only works if ORDER BY uses all fields and only fields - in expression list Blobs doesn't work with DISTINCT or ORDER BY *****************************************************************************/ -/** - function of sort for syntax: GROUP_CONCAT(DISTINCT expr,...) + + +/** + Compares the values for fields in expr list of GROUP_CONCAT. + @note + + GROUP_CONCAT([DISTINCT] expr [,expr ...] + [ORDER BY {unsigned_integer | col_name | expr} + [ASC | DESC] [,col_name ...]] + [SEPARATOR str_val]) + + @return + @retval -1 : key1 < key2 + @retval 0 : key1 = key2 + @retval 1 : key1 > key2 */ -int group_concat_key_cmp_with_distinct(void* arg, uchar* key1, - uchar* key2) +int group_concat_key_cmp_with_distinct(void* arg, const void* key1, + const void* key2) { - Item_func_group_concat* grp_item= (Item_func_group_concat*)arg; - TABLE *table= grp_item->table; - Item **field_item, **end; + Item_func_group_concat *item_func= (Item_func_group_concat*)arg; + TABLE *table= item_func->table; - for (field_item= grp_item->args, end= field_item + grp_item->arg_count_field; - field_item < end; - field_item++) + for (uint i= 0; i < item_func->arg_count_field; i++) { + Item *item= item_func->args[i]; + /* + If field_item is a const item then either get_tp_table_field returns 0 + or it is an item over a const table. + */ + if (item->const_item()) + continue; /* We have to use get_tmp_table_field() instead of real_item()->get_tmp_table_field() because we want the field in the temporary table, not the original field */ - Field *field= (*field_item)->get_tmp_table_field(); - /* - If field_item is a const item then either get_tmp_table_field returns 0 - or it is an item over a const table. - */ - if (field && !(*field_item)->const_item()) - { - int res; - uint offset= (field->offset(field->table->record[0]) - - table->s->null_bytes); - if ((res= field->cmp(key1 + offset, key2 + offset))) - return res; - } + Field *field= item->get_tmp_table_field(); + int res; + uint offset= field->offset(field->table->record[0])-table->s->null_bytes; + if((res= field->cmp((uchar*)key1 + offset, (uchar*)key2 + offset))) + return res; } return 0; } @@ -2910,7 +2917,8 @@ int group_concat_key_cmp_with_distinct(void* arg, uchar* key1, function of sort for syntax: GROUP_CONCAT(expr,... ORDER BY col,... ) */ -int group_concat_key_cmp_with_order(void* arg, uchar* key1, uchar* key2) +int group_concat_key_cmp_with_order(void* arg, const void* key1, + const void* key2) { Item_func_group_concat* grp_item= (Item_func_group_concat*) arg; ORDER **order_item, **end; @@ -2936,7 +2944,7 @@ int group_concat_key_cmp_with_order(void* arg, uchar* key1, uchar* key2) int res; uint offset= (field->offset(field->table->record[0]) - table->s->null_bytes); - if ((res= field->cmp(key1 + offset, key2 + offset))) + if ((res= field->cmp((uchar*)key1 + offset, (uchar*)key2 + offset))) return (*order_item)->asc ? res : -res; } } @@ -2950,25 +2958,6 @@ int group_concat_key_cmp_with_order(void* arg, uchar* key1, uchar* key2) /** - function of sort for syntax: - GROUP_CONCAT(DISTINCT expr,... ORDER BY col,... ). - - @bug - This doesn't work in the case when the order by contains data that - is not part of the field list because tree-insert will not notice - the duplicated values when inserting things sorted by ORDER BY -*/ - -int group_concat_key_cmp_with_distinct_and_order(void* arg,uchar* key1, - uchar* key2) -{ - if (!group_concat_key_cmp_with_distinct(arg,key1,key2)) - return 0; - return(group_concat_key_cmp_with_order(arg,key1,key2)); -} - - -/** Append data from current leaf to item->result. */ @@ -3054,7 +3043,7 @@ Item_func_group_concat(Name_resolution_context *context_arg, bool distinct_arg, List<Item> *select_list, SQL_LIST *order_list, String *separator_arg) :tmp_table_param(0), warning(0), - separator(separator_arg), tree(0), table(0), + separator(separator_arg), tree(0), unique_filter(NULL), table(0), order(0), context(context_arg), arg_count_order(order_list ? order_list->elements : 0), arg_count_field(select_list->elements), @@ -3109,6 +3098,7 @@ Item_func_group_concat::Item_func_group_concat(THD *thd, warning(item->warning), separator(item->separator), tree(item->tree), + unique_filter(item->unique_filter), table(item->table), order(item->order), context(item->context), @@ -3159,6 +3149,11 @@ void Item_func_group_concat::cleanup() delete_tree(tree); tree= 0; } + if (unique_filter) + { + delete unique_filter; + unique_filter= NULL; + } if (warning) { char warn_buff[MYSQL_ERRMSG_SIZE]; @@ -3188,6 +3183,8 @@ void Item_func_group_concat::clear() no_appended= TRUE; if (tree) reset_tree(tree); + if (distinct) + unique_filter->reset(); /* No need to reset the table as we never call write_row */ } @@ -3211,9 +3208,19 @@ bool Item_func_group_concat::add() } null_value= FALSE; + bool row_eligible= TRUE; + + if (distinct) + { + /* Filter out duplicate rows. */ + uint count= unique_filter->elements_in_tree(); + unique_filter->unique_add(table->record[0] + table->s->null_bytes); + if (count == unique_filter->elements_in_tree()) + row_eligible= FALSE; + } TREE_ELEMENT *el= 0; // Only for safety - if (tree) + if (row_eligible && tree) el= tree_insert(tree, table->record[0] + table->s->null_bytes, 0, tree->custom_arg); /* @@ -3221,7 +3228,7 @@ bool Item_func_group_concat::add() we can dump the row here in case of GROUP_CONCAT(DISTINCT...) instead of doing tree traverse later. */ - if (!warning_for_row && + if (row_eligible && !warning_for_row && (!tree || (el->count == 1 && distinct && !arg_count_order))) dump_leaf_key(table->record[0] + table->s->null_bytes, 1, this); @@ -3297,7 +3304,6 @@ bool Item_func_group_concat::setup(THD *thd) { List<Item> list; SELECT_LEX *select_lex= thd->lex->current_select; - qsort_cmp2 compare_key; DBUG_ENTER("Item_func_group_concat::setup"); /* @@ -3387,38 +3393,33 @@ bool Item_func_group_concat::setup(THD *thd) table->file->extra(HA_EXTRA_NO_ROWS); table->no_rows= 1; + /* + Need sorting or uniqueness: init tree and choose a function to sort. + Don't reserve space for NULLs: if any of gconcat arguments is NULL, + the row is not added to the result. + */ + uint tree_key_length= table->s->reclength - table->s->null_bytes; - if (distinct || arg_count_order) + if (arg_count_order) { - /* - Need sorting: init tree and choose a function to sort. - Don't reserve space for NULLs: if any of gconcat arguments is NULL, - the row is not added to the result. - */ - uint tree_key_length= table->s->reclength - table->s->null_bytes; - tree= &tree_base; - if (arg_count_order) - { - if (distinct) - compare_key= (qsort_cmp2) group_concat_key_cmp_with_distinct_and_order; - else - compare_key= (qsort_cmp2) group_concat_key_cmp_with_order; - } - else - { - compare_key= (qsort_cmp2) group_concat_key_cmp_with_distinct; - } /* - Create a tree for sorting. The tree is used to sort and to remove - duplicate values (according to the syntax of this function). If there - is no DISTINCT or ORDER BY clauses, we don't create this tree. + Create a tree for sorting. The tree is used to sort (according to the + syntax of this function). If there is no ORDER BY clause, we don't + create this tree. */ init_tree(tree, (uint) min(thd->variables.max_heap_table_size, thd->variables.sortbuff_size/16), 0, - tree_key_length, compare_key, 0, NULL, (void*) this); + tree_key_length, + group_concat_key_cmp_with_order , 0, NULL, (void*) this); } + if (distinct) + unique_filter= new Unique(group_concat_key_cmp_with_distinct, + (void*)this, + tree_key_length, + thd->variables.max_heap_table_size); + DBUG_RETURN(FALSE); } @@ -3488,3 +3489,10 @@ void Item_func_group_concat::print(String *str) str->append(*separator); str->append(STRING_WITH_LEN("\')")); } + + +Item_func_group_concat::~Item_func_group_concat() +{ + if (unique_filter) + delete unique_filter; +} diff --git a/sql/item_sum.h b/sql/item_sum.h index b3a382012f1..a3582967736 100644 --- a/sql/item_sum.h +++ b/sql/item_sum.h @@ -1173,11 +1173,22 @@ class Item_func_group_concat : public Item_sum String *separator; TREE tree_base; TREE *tree; + + /** + If DISTINCT is used with this GROUP_CONCAT, this member is used to filter + out duplicates. + @see Item_func_group_concat::setup + @see Item_func_group_concat::add + @see Item_func_group_concat::clear + */ + Unique *unique_filter; TABLE *table; ORDER **order; Name_resolution_context *context; - uint arg_count_order; // total count of ORDER BY items - uint arg_count_field; // count of arguments + /** The number of ORDER BY items. */ + uint arg_count_order; + /** The number of selected items, aka the expr list. */ + uint arg_count_field; uint count_cut_values; bool distinct; bool warning_for_row; @@ -1190,13 +1201,10 @@ class Item_func_group_concat : public Item_sum */ Item_func_group_concat *original; - friend int group_concat_key_cmp_with_distinct(void* arg, uchar* key1, - uchar* key2); - friend int group_concat_key_cmp_with_order(void* arg, uchar* key1, - uchar* key2); - friend int group_concat_key_cmp_with_distinct_and_order(void* arg, - uchar* key1, - uchar* key2); + friend int group_concat_key_cmp_with_distinct(void* arg, const void* key1, + const void* key2); + friend int group_concat_key_cmp_with_order(void* arg, const void* key1, + const void* key2); friend int dump_leaf_key(uchar* key, element_count count __attribute__((unused)), Item_func_group_concat *group_concat_item); @@ -1207,7 +1215,7 @@ public: SQL_LIST *is_order, String *is_separator); Item_func_group_concat(THD *thd, Item_func_group_concat *item); - ~Item_func_group_concat() {} + ~Item_func_group_concat(); void cleanup(); enum Sumfunctype sum_func () const {return GROUP_CONCAT_FUNC;} diff --git a/sql/item_timefunc.cc b/sql/item_timefunc.cc index ce760b8fd2a..a0beadcd481 100644 --- a/sql/item_timefunc.cc +++ b/sql/item_timefunc.cc @@ -3360,6 +3360,8 @@ bool Item_func_last_day::get_date(MYSQL_TIME *ltime, uint fuzzy_date) ltime->day= days_in_month[month_idx]; if ( month_idx == 1 && calc_days_in_year(ltime->year) == 366) ltime->day= 29; + ltime->hour= ltime->minute= ltime->second= 0; + ltime->second_part= 0; ltime->time_type= MYSQL_TIMESTAMP_DATE; return 0; } diff --git a/sql/log.cc b/sql/log.cc index a0c35059902..ac8831495f5 100644 --- a/sql/log.cc +++ b/sql/log.cc @@ -1420,6 +1420,21 @@ static int binlog_prepare(handlerton *hton, THD *thd, bool all) return 0; } +/** + This function is called once after each statement. + + It has the responsibility to flush the transaction cache to the + binlog file on commits. + + @param hton The binlog handlerton. + @param thd The client thread that executes the transaction. + @param all true if this is the last statement before a COMMIT + statement; false if either this is a statement in a + transaction but not the last, or if this is a statement + not inside a BEGIN block and autocommit is on. + + @see handlerton::commit +*/ static int binlog_commit(handlerton *hton, THD *thd, bool all) { DBUG_ENTER("binlog_commit"); @@ -1432,7 +1447,15 @@ static int binlog_commit(handlerton *hton, THD *thd, bool all) trx_data->reset(); DBUG_RETURN(0); } - if (all) + /* + Write commit event if at least one of the following holds: + - the user sends an explicit COMMIT; or + - the autocommit flag is on, and we are not inside a BEGIN. + However, if the user has not sent an explicit COMMIT, and we are + either inside a BEGIN or run with autocommit off, then this is not + the end of a transaction and we should not write a commit event. + */ + if (all || !(thd->options & (OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN))) { Query_log_event qev(thd, STRING_WITH_LEN("COMMIT"), TRUE, FALSE); qev.error_code= 0; // see comment in MYSQL_LOG::write(THD, IO_CACHE) @@ -1446,6 +1469,23 @@ static int binlog_commit(handlerton *hton, THD *thd, bool all) } } +/** + This function is called when a transaction involving a transactional + table is rolled back. + + It has the responsibility to flush the transaction cache to the + binlog file. However, if the transaction does not involve + non-transactional tables, nothing needs to be logged. + + @param hton The binlog handlerton. + @param thd The client thread that executes the transaction. + @param all true if this is the last statement before a COMMIT + statement; false if either this is a statement in a + transaction but not the last, or if this is a statement + not inside a BEGIN block and autocommit is on. + + @see handlerton::rollback +*/ static int binlog_rollback(handlerton *hton, THD *thd, bool all) { DBUG_ENTER("binlog_rollback"); @@ -3023,10 +3063,10 @@ err: void MYSQL_BIN_LOG::make_log_name(char* buf, const char* log_ident) { uint dir_len = dirname_length(log_file_name); - if (dir_len > FN_REFLEN) + if (dir_len >= FN_REFLEN) dir_len=FN_REFLEN-1; strnmov(buf, log_file_name, dir_len); - strmake(buf+dir_len, log_ident, FN_REFLEN - dir_len); + strmake(buf+dir_len, log_ident, FN_REFLEN - dir_len -1); } @@ -4015,32 +4055,42 @@ bool MYSQL_BIN_LOG::write(THD *thd, IO_CACHE *cache, Log_event *commit_event) if (my_b_tell(cache) > 0) { /* - Log "BEGIN" at the beginning of the transaction. - which may contain more than 1 SQL statement. + Log "BEGIN" at the beginning of every transaction. Here, a + transaction is either a BEGIN..COMMIT block or a single + statement in autocommit mode. */ - if (thd->options & (OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) - { - Query_log_event qinfo(thd, STRING_WITH_LEN("BEGIN"), TRUE, FALSE); - /* - Imagine this is rollback due to net timeout, after all statements of - the transaction succeeded. Then we want a zero-error code in BEGIN. - In other words, if there was a really serious error code it's already - in the statement's events, there is no need to put it also in this - internally generated event, and as this event is generated late it - would lead to false alarms. - This is safer than thd->clear_error() against kills at shutdown. - */ - qinfo.error_code= 0; - /* - Now this Query_log_event has artificial log_pos 0. It must be adjusted - to reflect the real position in the log. Not doing it would confuse the - slave: it would prevent this one from knowing where he is in the - master's binlog, which would result in wrong positions being shown to - the user, MASTER_POS_WAIT undue waiting etc. - */ - if (qinfo.write(&log_file)) - goto err; - } + Query_log_event qinfo(thd, STRING_WITH_LEN("BEGIN"), TRUE, FALSE); + /* + Imagine this is rollback due to net timeout, after all + statements of the transaction succeeded. Then we want a + zero-error code in BEGIN. In other words, if there was a + really serious error code it's already in the statement's + events, there is no need to put it also in this internally + generated event, and as this event is generated late it would + lead to false alarms. + + This is safer than thd->clear_error() against kills at shutdown. + */ + qinfo.error_code= 0; + /* + Now this Query_log_event has artificial log_pos 0. It must be + adjusted to reflect the real position in the log. Not doing it + would confuse the slave: it would prevent this one from + knowing where he is in the master's binlog, which would result + in wrong positions being shown to the user, MASTER_POS_WAIT + undue waiting etc. + */ + if (qinfo.write(&log_file)) + goto err; + + DBUG_EXECUTE_IF("crash_before_writing_xid", + { + if ((write_error= write_cache(cache, false, true))) + DBUG_PRINT("info", ("error writing binlog cache: %d", + write_error)); + DBUG_PRINT("info", ("crashing before writing xid")); + abort(); + }); if ((write_error= write_cache(cache, false, false))) goto err; diff --git a/sql/log_event.cc b/sql/log_event.cc index cf03dd5bf44..df0d1e8a020 100644 --- a/sql/log_event.cc +++ b/sql/log_event.cc @@ -36,7 +36,17 @@ #define FLAGSTR(V,F) ((V)&(F)?#F" ":"") -#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) && !defined(DBUG_OFF) && !defined(_lint) + +/* + Size of buffer for printing a double in format %.<PREC>g + + optional '-' + optional zero + '.' + PREC digits + 'e' + sign + + exponent digits + '\0' +*/ +#define FMT_G_BUFSIZE(PREC) (3 + (PREC) + 5 + 1) + + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) static const char *HA_ERR(int i) { switch (i) { @@ -90,7 +100,28 @@ static const char *HA_ERR(int i) case HA_ERR_LOGGING_IMPOSSIBLE: return "HA_ERR_LOGGING_IMPOSSIBLE"; case HA_ERR_CORRUPT_EVENT: return "HA_ERR_CORRUPT_EVENT"; } - return "<unknown error>"; + return 0; +} + +/** + macro to call from different branches of Rows_log_event::do_apply_event +*/ +static void inline slave_rows_error_report(enum loglevel level, int ha_error, + Relay_log_info const *rli, THD *thd, + TABLE *table, const char * type, + const char *log_name, ulong pos) +{ + const char *handler_error= HA_ERR(ha_error); + rli->report(level, thd->net.client_last_errno, + "Could not execute %s event on table %s.%s;" + "%s%s handler error %s; " + "the event's master log %s, end_log_pos %lu", + type, table->s->db.str, + table->s->table_name.str, + thd->net.client_last_error[0] != 0 ? thd->net.client_last_error : "", + thd->net.client_last_error[0] != 0 ? ";" : "", + handler_error == NULL? "<unknown>" : handler_error, + log_name, pos); } #endif @@ -460,9 +491,9 @@ static void print_set_option(IO_CACHE* file, uint32 bits_changed, returns the human readable name of the event's type */ -const char* Log_event::get_type_str() +const char* Log_event::get_type_str(Log_event_type type) { - switch(get_type_code()) { + switch(type) { case START_EVENT_V3: return "Start_v3"; case STOP_EVENT: return "Stop"; case QUERY_EVENT: return "Query"; @@ -480,6 +511,9 @@ const char* Log_event::get_type_str() case USER_VAR_EVENT: return "User var"; case FORMAT_DESCRIPTION_EVENT: return "Format_desc"; case TABLE_MAP_EVENT: return "Table_map"; + case PRE_GA_WRITE_ROWS_EVENT: return "Write_rows_event_old"; + case PRE_GA_UPDATE_ROWS_EVENT: return "Update_rows_event_old"; + case PRE_GA_DELETE_ROWS_EVENT: return "Delete_rows_event_old"; case WRITE_ROWS_EVENT: return "Write_rows"; case UPDATE_ROWS_EVENT: return "Update_rows"; case DELETE_ROWS_EVENT: return "Delete_rows"; @@ -490,6 +524,11 @@ const char* Log_event::get_type_str() } } +const char* Log_event::get_type_str() +{ + return get_type_str(get_type_code()); +} + /* Log_event::Log_event() @@ -997,6 +1036,8 @@ Log_event* Log_event::read_log_event(const char* buf, uint event_len, DBUG_ENTER("Log_event::read_log_event(char*,...)"); DBUG_ASSERT(description_event != 0); DBUG_PRINT("info", ("binlog_version: %d", description_event->binlog_version)); + DBUG_DUMP("data", (unsigned char*) buf, event_len); + /* Check the integrity */ if (event_len < EVENT_LEN_OFFSET || buf[EVENT_TYPE_OFFSET] >= ENUM_END_EVENT || @@ -1006,94 +1047,134 @@ Log_event* Log_event::read_log_event(const char* buf, uint event_len, DBUG_RETURN(NULL); // general sanity check - will fail on a partial read } - switch(buf[EVENT_TYPE_OFFSET]) { - case QUERY_EVENT: - ev = new Query_log_event(buf, event_len, description_event, QUERY_EVENT); - break; - case LOAD_EVENT: - ev = new Load_log_event(buf, event_len, description_event); - break; - case NEW_LOAD_EVENT: - ev = new Load_log_event(buf, event_len, description_event); - break; - case ROTATE_EVENT: - ev = new Rotate_log_event(buf, event_len, description_event); - break; + uint event_type= buf[EVENT_TYPE_OFFSET]; + if (event_type > description_event->number_of_event_types && + event_type != FORMAT_DESCRIPTION_EVENT) + { + /* + It is unsafe to use the description_event if its post_header_len + array does not include the event type. + */ + DBUG_PRINT("error", ("event type %d found, but the current " + "Format_description_log_event supports only %d event " + "types", event_type, + description_event->number_of_event_types)); + ev= NULL; + } + else + { + /* + In some previuos versions (see comment in + Format_description_log_event::Format_description_log_event(char*,...)), + event types were assigned different id numbers than in the + present version. In order to replicate from such versions to the + present version, we must map those event type id's to our event + type id's. The mapping is done with the event_type_permutation + array, which was set up when the Format_description_log_event + was read. + */ + if (description_event->event_type_permutation) + { + IF_DBUG({ + int new_event_type= + description_event->event_type_permutation[event_type]; + DBUG_PRINT("info", + ("converting event type %d to %d (%s)", + event_type, new_event_type, + get_type_str((Log_event_type)new_event_type))); + }); + event_type= description_event->event_type_permutation[event_type]; + } + + switch(event_type) { + case QUERY_EVENT: + ev = new Query_log_event(buf, event_len, description_event, QUERY_EVENT); + break; + case LOAD_EVENT: + ev = new Load_log_event(buf, event_len, description_event); + break; + case NEW_LOAD_EVENT: + ev = new Load_log_event(buf, event_len, description_event); + break; + case ROTATE_EVENT: + ev = new Rotate_log_event(buf, event_len, description_event); + break; #ifdef HAVE_REPLICATION - case SLAVE_EVENT: /* can never happen (unused event) */ - ev = new Slave_log_event(buf, event_len); - break; + case SLAVE_EVENT: /* can never happen (unused event) */ + ev = new Slave_log_event(buf, event_len); + break; #endif /* HAVE_REPLICATION */ - case CREATE_FILE_EVENT: - ev = new Create_file_log_event(buf, event_len, description_event); - break; - case APPEND_BLOCK_EVENT: - ev = new Append_block_log_event(buf, event_len, description_event); - break; - case DELETE_FILE_EVENT: - ev = new Delete_file_log_event(buf, event_len, description_event); - break; - case EXEC_LOAD_EVENT: - ev = new Execute_load_log_event(buf, event_len, description_event); - break; - case START_EVENT_V3: /* this is sent only by MySQL <=4.x */ - ev = new Start_log_event_v3(buf, description_event); - break; - case STOP_EVENT: - ev = new Stop_log_event(buf, description_event); - break; - case INTVAR_EVENT: - ev = new Intvar_log_event(buf, description_event); - break; - case XID_EVENT: - ev = new Xid_log_event(buf, description_event); - break; - case RAND_EVENT: - ev = new Rand_log_event(buf, description_event); - break; - case USER_VAR_EVENT: - ev = new User_var_log_event(buf, description_event); - break; - case FORMAT_DESCRIPTION_EVENT: - ev = new Format_description_log_event(buf, event_len, description_event); - break; + case CREATE_FILE_EVENT: + ev = new Create_file_log_event(buf, event_len, description_event); + break; + case APPEND_BLOCK_EVENT: + ev = new Append_block_log_event(buf, event_len, description_event); + break; + case DELETE_FILE_EVENT: + ev = new Delete_file_log_event(buf, event_len, description_event); + break; + case EXEC_LOAD_EVENT: + ev = new Execute_load_log_event(buf, event_len, description_event); + break; + case START_EVENT_V3: /* this is sent only by MySQL <=4.x */ + ev = new Start_log_event_v3(buf, description_event); + break; + case STOP_EVENT: + ev = new Stop_log_event(buf, description_event); + break; + case INTVAR_EVENT: + ev = new Intvar_log_event(buf, description_event); + break; + case XID_EVENT: + ev = new Xid_log_event(buf, description_event); + break; + case RAND_EVENT: + ev = new Rand_log_event(buf, description_event); + break; + case USER_VAR_EVENT: + ev = new User_var_log_event(buf, description_event); + break; + case FORMAT_DESCRIPTION_EVENT: + ev = new Format_description_log_event(buf, event_len, description_event); + break; #if defined(HAVE_REPLICATION) - case PRE_GA_WRITE_ROWS_EVENT: - ev = new Write_rows_log_event_old(buf, event_len, description_event); - break; - case PRE_GA_UPDATE_ROWS_EVENT: - ev = new Update_rows_log_event_old(buf, event_len, description_event); - break; - case PRE_GA_DELETE_ROWS_EVENT: - ev = new Delete_rows_log_event_old(buf, event_len, description_event); - break; - case WRITE_ROWS_EVENT: - ev = new Write_rows_log_event(buf, event_len, description_event); - break; - case UPDATE_ROWS_EVENT: - ev = new Update_rows_log_event(buf, event_len, description_event); - break; - case DELETE_ROWS_EVENT: - ev = new Delete_rows_log_event(buf, event_len, description_event); - break; - case TABLE_MAP_EVENT: - ev = new Table_map_log_event(buf, event_len, description_event); - break; + case PRE_GA_WRITE_ROWS_EVENT: + ev = new Write_rows_log_event_old(buf, event_len, description_event); + break; + case PRE_GA_UPDATE_ROWS_EVENT: + ev = new Update_rows_log_event_old(buf, event_len, description_event); + break; + case PRE_GA_DELETE_ROWS_EVENT: + ev = new Delete_rows_log_event_old(buf, event_len, description_event); + break; + case WRITE_ROWS_EVENT: + ev = new Write_rows_log_event(buf, event_len, description_event); + break; + case UPDATE_ROWS_EVENT: + ev = new Update_rows_log_event(buf, event_len, description_event); + break; + case DELETE_ROWS_EVENT: + ev = new Delete_rows_log_event(buf, event_len, description_event); + break; + case TABLE_MAP_EVENT: + ev = new Table_map_log_event(buf, event_len, description_event); + break; #endif - case BEGIN_LOAD_QUERY_EVENT: - ev = new Begin_load_query_log_event(buf, event_len, description_event); - break; - case EXECUTE_LOAD_QUERY_EVENT: - ev= new Execute_load_query_log_event(buf, event_len, description_event); - break; - case INCIDENT_EVENT: - ev = new Incident_log_event(buf, event_len, description_event); - break; - default: - DBUG_PRINT("error",("Unknown event code: %d", - (int) buf[EVENT_TYPE_OFFSET])); - ev= NULL; - break; + case BEGIN_LOAD_QUERY_EVENT: + ev = new Begin_load_query_log_event(buf, event_len, description_event); + break; + case EXECUTE_LOAD_QUERY_EVENT: + ev= new Execute_load_query_log_event(buf, event_len, description_event); + break; + case INCIDENT_EVENT: + ev = new Incident_log_event(buf, event_len, description_event); + break; + default: + DBUG_PRINT("error",("Unknown event code: %d", + (int) buf[EVENT_TYPE_OFFSET])); + ev= NULL; + break; + } } DBUG_PRINT("read_event", ("%s(type_code: %d; event_len: %d)", @@ -1632,6 +1713,7 @@ Query_log_event::Query_log_event(THD* thd_arg, const char* query_arg, DBUG_ASSERT(thd_arg->variables.character_set_client->number < 256*256); DBUG_ASSERT(thd_arg->variables.collation_connection->number < 256*256); DBUG_ASSERT(thd_arg->variables.collation_server->number < 256*256); + DBUG_ASSERT(thd_arg->variables.character_set_client->mbminlen == 1); int2store(charset, thd_arg->variables.character_set_client->number); int2store(charset+2, thd_arg->variables.collation_connection->number); int2store(charset+4, thd_arg->variables.collation_server->number); @@ -2135,7 +2217,7 @@ void Query_log_event::print(FILE* file, PRINT_EVENT_INFO* print_event_info) print_query_header(&cache, print_event_info); my_b_write(&cache, (uchar*) query, q_len); - my_b_printf(&cache, "%s\n", print_event_info->delimiter); + my_b_printf(&cache, "\n%s\n", print_event_info->delimiter); } #endif /* MYSQL_CLIENT */ @@ -2580,6 +2662,14 @@ void Start_log_event_v3::print(FILE* file, PRINT_EVENT_INFO* print_event_info) my_b_printf(&cache,"ROLLBACK%s\n", print_event_info->delimiter); #endif } + if (temp_buf && + print_event_info->base64_output_mode != BASE64_OUTPUT_NEVER && + !print_event_info->short_form) + { + my_b_printf(&cache, "BINLOG '\n"); + print_base64(&cache, print_event_info, FALSE); + print_event_info->printed_fd_event= TRUE; + } DBUG_VOID_RETURN; } #endif /* MYSQL_CLIENT */ @@ -2715,7 +2805,7 @@ int Start_log_event_v3::do_apply_event(Relay_log_info const *rli) Format_description_log_event:: Format_description_log_event(uint8 binlog_ver, const char* server_ver) - :Start_log_event_v3() + :Start_log_event_v3(), event_type_permutation(0) { binlog_version= binlog_ver; switch (binlog_ver) { @@ -2842,7 +2932,7 @@ Format_description_log_event(const char* buf, const Format_description_log_event* description_event) - :Start_log_event_v3(buf, description_event) + :Start_log_event_v3(buf, description_event), event_type_permutation(0) { DBUG_ENTER("Format_description_log_event::Format_description_log_event(char*,...)"); buf+= LOG_EVENT_MINIMAL_HEADER_LEN; @@ -2857,6 +2947,65 @@ Format_description_log_event(const char* buf, number_of_event_types* sizeof(*post_header_len), MYF(0)); calc_server_version_split(); + + /* + In some previous versions, the events were given other event type + id numbers than in the present version. When replicating from such + a version, we therefore set up an array that maps those id numbers + to the id numbers of the present server. + + If post_header_len is null, it means malloc failed, and is_valid + will fail, so there is no need to do anything. + + The trees which have wrong event id's are: + mysql-5.1-wl2325-5.0-drop6p13-alpha, mysql-5.1-wl2325-5.0-drop6, + mysql-5.1-wl2325-5.0, mysql-5.1-wl2325-no-dd (`grep -C2 + BEGIN_LOAD_QUERY_EVENT /home/bk/ * /sql/log_event.h`). The + corresponding version (`grep mysql, configure.in` in those trees) + strings are 5.2.2-a_drop6p13-alpha, 5.2.2-a_drop6p13c, + 5.1.5-a_drop5p20, 5.1.2-a_drop5p5. + */ + if (post_header_len && + (strncmp(server_version, "5.1.2-a_drop5", 13) == 0 || + strncmp(server_version, "5.1.5-a_drop5", 13) == 0 || + strncmp(server_version, "5.2.2-a_drop6", 13) == 0)) + { + if (number_of_event_types != 22) + { + DBUG_PRINT("info", (" number_of_event_types=%d", + number_of_event_types)); + /* this makes is_valid() return false. */ + my_free(post_header_len, MYF(MY_ALLOW_ZERO_PTR)); + post_header_len= NULL; + DBUG_VOID_RETURN; + } + static const uint8 perm[23]= + { + UNKNOWN_EVENT, START_EVENT_V3, QUERY_EVENT, STOP_EVENT, ROTATE_EVENT, + INTVAR_EVENT, LOAD_EVENT, SLAVE_EVENT, CREATE_FILE_EVENT, + APPEND_BLOCK_EVENT, EXEC_LOAD_EVENT, DELETE_FILE_EVENT, + NEW_LOAD_EVENT, + RAND_EVENT, USER_VAR_EVENT, + FORMAT_DESCRIPTION_EVENT, + TABLE_MAP_EVENT, + PRE_GA_WRITE_ROWS_EVENT, + PRE_GA_UPDATE_ROWS_EVENT, + PRE_GA_DELETE_ROWS_EVENT, + XID_EVENT, + BEGIN_LOAD_QUERY_EVENT, + EXECUTE_LOAD_QUERY_EVENT, + }; + event_type_permutation= perm; + /* + Since we use (permuted) event id's to index the post_header_len + array, we need to permute the post_header_len array too. + */ + uint8 post_header_len_temp[23]; + for (int i= 1; i < 23; i++) + post_header_len_temp[perm[i] - 1]= post_header_len[i - 1]; + for (int i= 0; i < 22; i++) + post_header_len[i] = post_header_len_temp[i]; + } DBUG_VOID_RETURN; } @@ -4267,6 +4416,7 @@ Xid_log_event(const char* buf, #ifndef MYSQL_CLIENT bool Xid_log_event::write(IO_CACHE* file) { + DBUG_EXECUTE_IF("do_not_write_xid", return 0;); return write_header(file, sizeof(xid)) || my_b_safe_write(file, (uchar*) &xid, sizeof(xid)); } @@ -4520,8 +4670,10 @@ void User_var_log_event::print(FILE* file, PRINT_EVENT_INFO* print_event_info) switch (type) { case REAL_RESULT: double real_val; + char real_buf[FMT_G_BUFSIZE(14)]; float8get(real_val, val); - my_b_printf(&cache, ":=%.14g%s\n", real_val, print_event_info->delimiter); + my_sprintf(real_buf, (real_buf, "%.14g", real_val)); + my_b_printf(&cache, ":=%s%s\n", real_buf, print_event_info->delimiter); break; case INT_RESULT: char int_buf[22]; @@ -4925,7 +5077,7 @@ Create_file_log_event(THD* thd_arg, sql_exchange* ex, const char* db_arg, const char* table_name_arg, List<Item>& fields_arg, enum enum_duplicates handle_dup, bool ignore, - char* block_arg, uint block_len_arg, bool using_trans) + uchar* block_arg, uint block_len_arg, bool using_trans) :Load_log_event(thd_arg,ex,db_arg,table_name_arg,fields_arg,handle_dup, ignore, using_trans), fake_base(0), block(block_arg), event_buf(0), block_len(block_len_arg), @@ -5023,8 +5175,8 @@ Create_file_log_event::Create_file_log_event(const char* buf, uint len, Load_log_event::get_data_size() + create_file_header_len + 1); if (len < block_offset) - return; - block = (char*)buf + block_offset; + DBUG_VOID_RETURN; + block = (uchar*)buf + block_offset; block_len = len - block_offset; } else @@ -5182,7 +5334,7 @@ err: #ifndef MYSQL_CLIENT Append_block_log_event::Append_block_log_event(THD *thd_arg, const char *db_arg, - char *block_arg, + uchar *block_arg, uint block_len_arg, bool using_trans) :Log_event(thd_arg,0, using_trans), block(block_arg), @@ -5208,7 +5360,7 @@ Append_block_log_event::Append_block_log_event(const char* buf, uint len, if (len < total_header_len) DBUG_VOID_RETURN; file_id= uint4korr(buf + common_header_len + AB_FILE_ID_OFFSET); - block= (char*)buf + total_header_len; + block= (uchar*)buf + total_header_len; block_len= len - total_header_len; DBUG_VOID_RETURN; } @@ -5600,7 +5752,7 @@ err: #ifndef MYSQL_CLIENT Begin_load_query_log_event:: -Begin_load_query_log_event(THD* thd_arg, const char* db_arg, char* block_arg, +Begin_load_query_log_event(THD* thd_arg, const char* db_arg, uchar* block_arg, uint block_len_arg, bool using_trans) :Append_block_log_event(thd_arg, db_arg, block_arg, block_len_arg, using_trans) @@ -5732,12 +5884,12 @@ void Execute_load_query_log_event::print(FILE* file, my_b_printf(&cache, " REPLACE"); my_b_printf(&cache, " INTO"); my_b_write(&cache, (uchar*) query + fn_pos_end, q_len-fn_pos_end); - my_b_printf(&cache, "%s\n", print_event_info->delimiter); + my_b_printf(&cache, "\n%s\n", print_event_info->delimiter); } else { my_b_write(&cache, (uchar*) query, q_len); - my_b_printf(&cache, "%s\n", print_event_info->delimiter); + my_b_printf(&cache, "\n%s\n", print_event_info->delimiter); } if (!print_event_info->short_form) @@ -6180,7 +6332,6 @@ int Rows_log_event::do_apply_event(Relay_log_info const *rli) { DBUG_ENTER("Rows_log_event::do_apply_event(Relay_log_info*)"); int error= 0; - /* If m_table_id == ~0UL, then we have a dummy event that does not contain any data. In that case, we just remove all tables in the @@ -6226,6 +6377,24 @@ int Rows_log_event::do_apply_event(Relay_log_info const *rli) */ lex_start(thd); + /* + There are a few flags that are replicated with each row event. + Make sure to set/clear them before executing the main body of + the event. + */ + if (get_flags(NO_FOREIGN_KEY_CHECKS_F)) + thd->options|= OPTION_NO_FOREIGN_KEY_CHECKS; + else + thd->options&= ~OPTION_NO_FOREIGN_KEY_CHECKS; + + if (get_flags(RELAXED_UNIQUE_CHECKS_F)) + thd->options|= OPTION_RELAXED_UNIQUE_CHECKS; + else + thd->options&= ~OPTION_RELAXED_UNIQUE_CHECKS; + /* A small test to verify that objects have consistent types */ + DBUG_ASSERT(sizeof(thd->options) == sizeof(OPTION_RELAXED_UNIQUE_CHECKS)); + + while ((error= lock_tables(thd, rli->tables_to_lock, rli->tables_to_lock_count, &need_reopen))) { @@ -6360,22 +6529,6 @@ int Rows_log_event::do_apply_event(Relay_log_info const *rli) So we call set_time(), like in SBR. Presently it changes nothing. */ thd->set_time((time_t)when); - /* - There are a few flags that are replicated with each row event. - Make sure to set/clear them before executing the main body of - the event. - */ - if (get_flags(NO_FOREIGN_KEY_CHECKS_F)) - thd->options|= OPTION_NO_FOREIGN_KEY_CHECKS; - else - thd->options&= ~OPTION_NO_FOREIGN_KEY_CHECKS; - - if (get_flags(RELAXED_UNIQUE_CHECKS_F)) - thd->options|= OPTION_RELAXED_UNIQUE_CHECKS; - else - thd->options&= ~OPTION_RELAXED_UNIQUE_CHECKS; - /* A small test to verify that objects have consistent types */ - DBUG_ASSERT(sizeof(thd->options) == sizeof(OPTION_RELAXED_UNIQUE_CHECKS)); /* Now we are in a statement and will stay in a statement until we @@ -6408,8 +6561,9 @@ int Rows_log_event::do_apply_event(Relay_log_info const *rli) if (!get_flags(COMPLETE_ROWS_F)) bitmap_intersect(table->write_set,&m_cols); + this->slave_exec_mode= slave_exec_mode_options; // fix the mode + // Do event specific preparations - error= do_before_row_operations(rli); // row processing loop @@ -6428,23 +6582,41 @@ int Rows_log_event::do_apply_event(Relay_log_info const *rli) { case 0: break; - - /* Some recoverable errors */ + /* + The following list of "idempotent" errors + means that an error from the list might happen + because of idempotent (more than once) + applying of a binlog file. + Notice, that binlog has a ddl operation its + second applying may cause + + case HA_ERR_TABLE_DEF_CHANGED: + case HA_ERR_CANNOT_ADD_FOREIGN: + + which are not included into to the list. + */ case HA_ERR_RECORD_CHANGED: case HA_ERR_RECORD_DELETED: case HA_ERR_KEY_NOT_FOUND: case HA_ERR_END_OF_FILE: - /* Idempotency support: OK if tuple does not exist */ + case HA_ERR_FOUND_DUPP_KEY: + case HA_ERR_FOUND_DUPP_UNIQUE: + case HA_ERR_FOREIGN_DUPLICATE_KEY: + case HA_ERR_NO_REFERENCED_ROW: + case HA_ERR_ROW_IS_REFERENCED: + DBUG_PRINT("info", ("error: %s", HA_ERR(error))); - error= 0; + if (bit_is_set(slave_exec_mode, SLAVE_EXEC_MODE_IDEMPOTENT) == 1) + { + if (global_system_variables.log_warnings) + slave_rows_error_report(WARNING_LEVEL, error, rli, thd, table, + get_type_str(), + RPL_LOG_NAME, (ulong) log_pos); + error= 0; + } break; - + default: - rli->report(ERROR_LEVEL, - thd->is_error() ? thd->main_da.sql_errno() : 0, - "Error in %s event: row application failed. %s", - get_type_str(), - thd->is_error() ? thd->main_da.message() : ""); thd->is_slave_error= 1; break; } @@ -6488,17 +6660,14 @@ int Rows_log_event::do_apply_event(Relay_log_info const *rli) */ if (rli->tables_to_lock && get_flags(STMT_END_F)) const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); - + if (error) { /* error has occured during the transaction */ - rli->report(ERROR_LEVEL, - thd->is_error() ? thd->main_da.sql_errno() : 0, - "Error in %s event: error during transaction execution " - "on table %s.%s. %s", - get_type_str(), table->s->db.str, - table->s->table_name.str, - thd->is_error() ? thd->main_da.message() : ""); - + slave_rows_error_report(ERROR_LEVEL, error, rli, thd, table, + get_type_str(), RPL_LOG_NAME, (ulong) log_pos); + } + if (error) + { /* If one day we honour --skip-slave-errors in row-based replication, and the error should be skipped, then we would clear mappings, rollback, @@ -6610,6 +6779,7 @@ Rows_log_event::do_update_pos(Relay_log_info *rli) */ thd->reset_current_stmt_binlog_row_based(); + rli->cleanup_context(thd, 0); if (error == 0) { @@ -7299,43 +7469,50 @@ Write_rows_log_event::do_before_row_operations(const Slave_reporting_capability { int error= 0; - /* - We are using REPLACE semantics and not INSERT IGNORE semantics - when writing rows, that is: new rows replace old rows. We need to - inform the storage engine that it should use this behaviour. + /** + todo: to introduce a property for the event (handler?) which forces + applying the event in the replace (idempotent) fashion. */ + if (bit_is_set(slave_exec_mode, SLAVE_EXEC_MODE_IDEMPOTENT) == 1 || + m_table->s->db_type()->db_type == DB_TYPE_NDBCLUSTER) + { + /* + We are using REPLACE semantics and not INSERT IGNORE semantics + when writing rows, that is: new rows replace old rows. We need to + inform the storage engine that it should use this behaviour. + */ + + /* Tell the storage engine that we are using REPLACE semantics. */ + thd->lex->duplicates= DUP_REPLACE; + + /* + Pretend we're executing a REPLACE command: this is needed for + InnoDB and NDB Cluster since they are not (properly) checking the + lex->duplicates flag. + */ + thd->lex->sql_command= SQLCOM_REPLACE; + /* + Do not raise the error flag in case of hitting to an unique attribute + */ + m_table->file->extra(HA_EXTRA_IGNORE_DUP_KEY); + /* + NDB specific: update from ndb master wrapped as Write_rows + so that the event should be applied to replace slave's row + */ + m_table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE); + /* + NDB specific: if update from ndb master wrapped as Write_rows + does not find the row it's assumed idempotent binlog applying + is taking place; don't raise the error. + */ + m_table->file->extra(HA_EXTRA_IGNORE_NO_KEY); + /* + TODO: the cluster team (Tomas?) says that it's better if the engine knows + how many rows are going to be inserted, then it can allocate needed memory + from the start. + */ + } - /* Tell the storage engine that we are using REPLACE semantics. */ - thd->lex->duplicates= DUP_REPLACE; - - /* - Pretend we're executing a REPLACE command: this is needed for - InnoDB and NDB Cluster since they are not (properly) checking the - lex->duplicates flag. - */ - thd->lex->sql_command= SQLCOM_REPLACE; - /* - Do not raise the error flag in case of hitting to an unique attribute - */ - m_table->file->extra(HA_EXTRA_IGNORE_DUP_KEY); - /* - NDB specific: update from ndb master wrapped as Write_rows - */ - /* - so that the event should be applied to replace slave's row - */ - m_table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE); - /* - NDB specific: if update from ndb master wrapped as Write_rows - does not find the row it's assumed idempotent binlog applying - is taking place; don't raise the error. - */ - m_table->file->extra(HA_EXTRA_IGNORE_NO_KEY); - /* - TODO: the cluster team (Tomas?) says that it's better if the engine knows - how many rows are going to be inserted, then it can allocate needed memory - from the start. - */ m_table->file->ha_start_bulk_insert(0); /* We need TIMESTAMP_NO_AUTO_SET otherwise ha_write_row() will not use fill @@ -7357,18 +7534,23 @@ Write_rows_log_event::do_before_row_operations(const Slave_reporting_capability } int -Write_rows_log_event::do_after_row_operations(const Slave_reporting_capability *const, +Write_rows_log_event::do_after_row_operations(const Slave_reporting_capability *const, int error) { int local_error= 0; - m_table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY); - m_table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE); - /* - reseting the extra with - table->file->extra(HA_EXTRA_NO_IGNORE_NO_KEY); - fires bug#27077 - todo: explain or fix - */ + if (bit_is_set(slave_exec_mode, SLAVE_EXEC_MODE_IDEMPOTENT) == 1 || + m_table->s->db_type()->db_type == DB_TYPE_NDBCLUSTER) + { + m_table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY); + m_table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE); + /* + resetting the extra with + table->file->extra(HA_EXTRA_NO_IGNORE_NO_KEY); + fires bug#27077 + explanation: file->reset() performs this duty + ultimately. Still todo: fix + */ + } if ((local_error= m_table->file->ha_end_bulk_insert())) { m_table->file->print_error(local_error, MYF(0)); @@ -7487,23 +7669,22 @@ Rows_log_event::write_row(const Relay_log_info *const rli, while ((error= table->file->ha_write_row(table->record[0]))) { - if (error == HA_ERR_LOCK_DEADLOCK || error == HA_ERR_LOCK_WAIT_TIMEOUT) - { - table->file->print_error(error, MYF(0)); /* to check at exec_relay_log_event */ - DBUG_RETURN(error); - } - if ((keynum= table->file->get_dup_key(error)) < 0) + if (error == HA_ERR_LOCK_DEADLOCK || + error == HA_ERR_LOCK_WAIT_TIMEOUT || + (keynum= table->file->get_dup_key(error)) < 0 || + !overwrite) { - DBUG_PRINT("info",("Can't locate duplicate key (get_dup_key returns %d)",keynum)); - table->file->print_error(error, MYF(0)); + DBUG_PRINT("info",("get_dup_key returns %d)", keynum)); /* - We failed to retrieve the duplicate key + Deadlock, waiting for lock or just an error from the handler + such as HA_ERR_FOUND_DUPP_KEY when overwrite is false. + Retrieval of the duplicate key number may fail - either because the error was not "duplicate key" error - or because the information which key is not available */ + table->file->print_error(error, MYF(0)); DBUG_RETURN(error); } - /* We need to retrieve the old row into record[1] to be able to either update or delete the offending record. We either: @@ -7641,14 +7822,16 @@ int Write_rows_log_event::do_exec_row(const Relay_log_info *const rli) { DBUG_ASSERT(m_table != NULL); - int error= write_row(rli, TRUE /* overwrite */); - + int error= + write_row(rli, /* if 1 then overwrite */ + bit_is_set(slave_exec_mode, SLAVE_EXEC_MODE_IDEMPOTENT) == 1); + if (error && !thd->is_error()) { DBUG_ASSERT(0); my_error(ER_UNKNOWN_ERROR, MYF(0)); } - + return error; } diff --git a/sql/log_event.h b/sql/log_event.h index 4bd496af2a4..efb8675780e 100644 --- a/sql/log_event.h +++ b/sql/log_event.h @@ -567,6 +567,15 @@ class Format_description_log_event; class Relay_log_info; #ifdef MYSQL_CLIENT +enum enum_base64_output_mode { + BASE64_OUTPUT_NEVER= 0, + BASE64_OUTPUT_AUTO= 1, + BASE64_OUTPUT_ALWAYS= 2, + BASE64_OUTPUT_UNSPEC= 3, + /* insert new output modes here */ + BASE64_OUTPUT_MODE_COUNT +}; + /* A structure for mysqlbinlog to know how to print events @@ -600,7 +609,8 @@ typedef struct st_print_event_info st_print_event_info() :flags2_inited(0), sql_mode_inited(0), auto_increment_increment(1),auto_increment_offset(1), charset_inited(0), - lc_time_names_number(0), charset_database_number(0) + lc_time_names_number(0), charset_database_number(0), + base64_output_mode(BASE64_OUTPUT_UNSPEC), printed_fd_event(FALSE) { /* Currently we only use static PRINT_EVENT_INFO objects, so zeroed at @@ -627,7 +637,14 @@ typedef struct st_print_event_info /* Settings on how to print the events */ bool short_form; - bool base64_output; + enum_base64_output_mode base64_output_mode; + /* + This is set whenever a Format_description_event is printed. + Later, when an event is printed in base64, this flag is tested: if + no Format_description_event has been seen, it is unsafe to print + the base64 event, so an error message is generated. + */ + bool printed_fd_event; my_off_t hexdump_from; uint8 common_header_len; char delimiter[16]; @@ -809,6 +826,12 @@ public: bool cache_stmt; + /** + A storage to cache the global system variable's value. + Handling of a separate event will be governed its member. + */ + ulong slave_exec_mode; + #ifndef MYSQL_CLIENT THD* thd; @@ -930,7 +953,13 @@ public: const char **error, const Format_description_log_event *description_event); - /* returns the human readable name of the event's type */ + /** + Returns the human readable name of the given event type. + */ + static const char* get_type_str(Log_event_type type); + /** + Returns the human readable name of this event's type. + */ const char* get_type_str(); /* Return start of query time or current time */ @@ -2077,12 +2106,16 @@ public: /* The list of post-headers' lengthes */ uint8 *post_header_len; uchar server_version_split[3]; + const uint8 *event_type_permutation; Format_description_log_event(uint8 binlog_ver, const char* server_ver=0); Format_description_log_event(const char* buf, uint event_len, const Format_description_log_event *description_event); - ~Format_description_log_event() { my_free((uchar*)post_header_len, MYF(0)); } + ~Format_description_log_event() + { + my_free((uchar*)post_header_len, MYF(MY_ALLOW_ZERO_PTR)); + } Log_event_type get_type_code() { return FORMAT_DESCRIPTION_EVENT;} #ifndef MYSQL_CLIENT bool write(IO_CACHE* file); @@ -2486,7 +2519,7 @@ protected: */ bool fake_base; public: - char* block; + uchar* block; const char *event_buf; uint block_len; uint file_id; @@ -2497,7 +2530,7 @@ public: const char* table_name_arg, List<Item>& fields_arg, enum enum_duplicates handle_dup, bool ignore, - char* block_arg, uint block_len_arg, + uchar* block_arg, uint block_len_arg, bool using_trans); #ifdef HAVE_REPLICATION void pack_info(Protocol* protocol); @@ -2552,7 +2585,7 @@ private: class Append_block_log_event: public Log_event { public: - char* block; + uchar* block; uint block_len; uint file_id; /* @@ -2569,7 +2602,7 @@ public: const char* db; #ifndef MYSQL_CLIENT - Append_block_log_event(THD* thd, const char* db_arg, char* block_arg, + Append_block_log_event(THD* thd, const char* db_arg, uchar* block_arg, uint block_len_arg, bool using_trans); #ifdef HAVE_REPLICATION void pack_info(Protocol* protocol); @@ -2693,7 +2726,7 @@ class Begin_load_query_log_event: public Append_block_log_event public: #ifndef MYSQL_CLIENT Begin_load_query_log_event(THD* thd_arg, const char *db_arg, - char* block_arg, uint block_len_arg, + uchar* block_arg, uint block_len_arg, bool using_trans); #ifdef HAVE_REPLICATION Begin_load_query_log_event(THD* thd); diff --git a/sql/log_event_old.cc b/sql/log_event_old.cc index 6d5d86e42fe..7621bdc6291 100644 --- a/sql/log_event_old.cc +++ b/sql/log_event_old.cc @@ -11,9 +11,9 @@ // Old implementation of do_apply_event() int -Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli) +Old_rows_log_event::do_apply_event(Old_rows_log_event *ev, const Relay_log_info *rli) { - DBUG_ENTER("Rows_log_event::do_apply_event(st_relay_log_info*)"); + DBUG_ENTER("Old_rows_log_event::do_apply_event(st_relay_log_info*)"); int error= 0; THD *thd= ev->thd; uchar const *row_start= ev->m_rows_buf; @@ -30,7 +30,7 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli This one is supposed to be set: just an extra check so that nothing strange has happened. */ - DBUG_ASSERT(ev->get_flags(Rows_log_event::STMT_END_F)); + DBUG_ASSERT(ev->get_flags(Old_rows_log_event::STMT_END_F)); const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); close_thread_tables(thd); @@ -148,7 +148,7 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli thd->lock= 0; thd->is_slave_error= 1; const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); - DBUG_RETURN(Rows_log_event::ERR_BAD_TABLE_DEF); + DBUG_RETURN(Old_rows_log_event::ERR_BAD_TABLE_DEF); } } } @@ -163,8 +163,8 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli TODO [/Matz]: Maybe the query cache should not be invalidated here? It might be that a table is not changed, even though it was locked for the statement. We do know that each - Rows_log_event contain at least one row, so after processing one - Rows_log_event, we can invalidate the query cache for the + Old_rows_log_event contain at least one row, so after processing one + Old_rows_log_event, we can invalidate the query cache for the associated table. */ for (TABLE_LIST *ptr= rli->tables_to_lock ; ptr ; ptr= ptr->next_global) @@ -200,12 +200,12 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli Make sure to set/clear them before executing the main body of the event. */ - if (ev->get_flags(Rows_log_event::NO_FOREIGN_KEY_CHECKS_F)) + if (ev->get_flags(Old_rows_log_event::NO_FOREIGN_KEY_CHECKS_F)) thd->options|= OPTION_NO_FOREIGN_KEY_CHECKS; else thd->options&= ~OPTION_NO_FOREIGN_KEY_CHECKS; - if (ev->get_flags(Rows_log_event::RELAXED_UNIQUE_CHECKS_F)) + if (ev->get_flags(Old_rows_log_event::RELAXED_UNIQUE_CHECKS_F)) thd->options|= OPTION_RELAXED_UNIQUE_CHECKS; else thd->options&= ~OPTION_RELAXED_UNIQUE_CHECKS; @@ -275,7 +275,7 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli We need to delay this clear until the table def is no longer needed. The table def is needed in unpack_row(). */ - if (rli->tables_to_lock && ev->get_flags(Rows_log_event::STMT_END_F)) + if (rli->tables_to_lock && ev->get_flags(Old_rows_log_event::STMT_END_F)) const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); if (error) @@ -311,7 +311,7 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli */ if (table && (table->s->primary_key == MAX_KEY) && !ev->cache_stmt && - ev->get_flags(Rows_log_event::STMT_END_F) == Rows_log_event::RLE_NO_FLAGS) + ev->get_flags(Old_rows_log_event::STMT_END_F) == Old_rows_log_event::RLE_NO_FLAGS) { /* ------------ Temporary fix until WL#2975 is implemented --------- @@ -323,7 +323,7 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli present, and idempotency is not guaranteed (no PK) so we risk that repeating leads to double insert. So we desperately try to continue, hope we'll eventually leave this buggy situation (by - executing the final Rows_log_event). If we are in a hopeless + executing the final Old_rows_log_event). If we are in a hopeless wait (reached end of last relay log and nothing gets appended there), we timeout after one minute, and notify DBA about the problem. When WL#2975 is implemented, just remove the member @@ -336,6 +336,7 @@ Old_rows_log_event::do_apply_event(Rows_log_event *ev, const Relay_log_info *rli } #endif + #if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) /* @@ -350,6 +351,7 @@ last_uniq_key(TABLE *table, uint keyno) return 1; } + /* Compares table->record[0] and table->record[1] @@ -428,6 +430,7 @@ record_compare_exit: return result; } + /* Copy "extra" columns from record[1] to record[0]. @@ -516,6 +519,7 @@ copy_extra_record_fields(TABLE *table, DBUG_RETURN(0); // All OK } + /* Replace the provided record in the database. @@ -668,6 +672,7 @@ replace_record(THD *thd, TABLE *table, DBUG_RETURN(error); } + /** Find the row given by 'key', if the table has keys, or else use a table scan to find (and fetch) the row. @@ -879,6 +884,7 @@ static int find_and_fetch_row(TABLE *table, uchar *key) DBUG_RETURN(0); } + /********************************************************** Row handling primitives for Write_rows_log_event_old **********************************************************/ @@ -944,6 +950,7 @@ int Write_rows_log_event_old::do_before_row_operations(TABLE *table) return error; } + int Write_rows_log_event_old::do_after_row_operations(TABLE *table, int error) { int local_error= 0; @@ -962,6 +969,7 @@ int Write_rows_log_event_old::do_after_row_operations(TABLE *table, int error) return error? error : local_error; } + int Write_rows_log_event_old::do_prepare_row(THD *thd_arg, Relay_log_info const *rli, @@ -981,6 +989,7 @@ Write_rows_log_event_old::do_prepare_row(THD *thd_arg, return error; } + int Write_rows_log_event_old::do_exec_row(TABLE *table) { DBUG_ASSERT(table != NULL); @@ -988,6 +997,7 @@ int Write_rows_log_event_old::do_exec_row(TABLE *table) return error; } + /********************************************************** Row handling primitives for Delete_rows_log_event_old **********************************************************/ @@ -1029,6 +1039,7 @@ int Delete_rows_log_event_old::do_before_row_operations(TABLE *table) return error; } + int Delete_rows_log_event_old::do_after_row_operations(TABLE *table, int error) { /*error= ToDo:find out what this should really be, this triggers close_scan in nbd, returning error?*/ @@ -1041,6 +1052,7 @@ int Delete_rows_log_event_old::do_after_row_operations(TABLE *table, int error) return error; } + int Delete_rows_log_event_old::do_prepare_row(THD *thd_arg, Relay_log_info const *rli, @@ -1074,6 +1086,7 @@ Delete_rows_log_event_old::do_prepare_row(THD *thd_arg, return error; } + int Delete_rows_log_event_old::do_exec_row(TABLE *table) { int error; @@ -1091,6 +1104,7 @@ int Delete_rows_log_event_old::do_exec_row(TABLE *table) return error; } + /********************************************************** Row handling primitives for Update_rows_log_event_old **********************************************************/ @@ -1124,6 +1138,7 @@ int Update_rows_log_event_old::do_before_row_operations(TABLE *table) return error; } + int Update_rows_log_event_old::do_after_row_operations(TABLE *table, int error) { /*error= ToDo:find out what this should really be, this triggers close_scan in nbd, returning error?*/ @@ -1136,6 +1151,7 @@ int Update_rows_log_event_old::do_after_row_operations(TABLE *table, int error) return error; } + int Update_rows_log_event_old::do_prepare_row(THD *thd_arg, Relay_log_info const *rli, TABLE *table, @@ -1179,6 +1195,7 @@ int Update_rows_log_event_old::do_prepare_row(THD *thd_arg, return error; } + int Update_rows_log_event_old::do_exec_row(TABLE *table) { DBUG_ASSERT(table != NULL); @@ -1217,3 +1234,1676 @@ int Update_rows_log_event_old::do_exec_row(TABLE *table) } #endif + + +/************************************************************************** + Rows_log_event member functions +**************************************************************************/ + +#ifndef MYSQL_CLIENT +Old_rows_log_event::Old_rows_log_event(THD *thd_arg, TABLE *tbl_arg, ulong tid, + MY_BITMAP const *cols, + bool is_transactional) + : Log_event(thd_arg, 0, is_transactional), + m_row_count(0), + m_table(tbl_arg), + m_table_id(tid), + m_width(tbl_arg ? tbl_arg->s->fields : 1), + m_rows_buf(0), m_rows_cur(0), m_rows_end(0), m_flags(0) +#ifdef HAVE_REPLICATION + , m_curr_row(NULL), m_curr_row_end(NULL), m_key(NULL) +#endif +{ + + // This constructor should not be reached. + assert(0); + + /* + We allow a special form of dummy event when the table, and cols + are null and the table id is ~0UL. This is a temporary + solution, to be able to terminate a started statement in the + binary log: the extraneous events will be removed in the future. + */ + DBUG_ASSERT(tbl_arg && tbl_arg->s && tid != ~0UL || + !tbl_arg && !cols && tid == ~0UL); + + if (thd_arg->options & OPTION_NO_FOREIGN_KEY_CHECKS) + set_flags(NO_FOREIGN_KEY_CHECKS_F); + if (thd_arg->options & OPTION_RELAXED_UNIQUE_CHECKS) + set_flags(RELAXED_UNIQUE_CHECKS_F); + /* if bitmap_init fails, caught in is_valid() */ + if (likely(!bitmap_init(&m_cols, + m_width <= sizeof(m_bitbuf)*8 ? m_bitbuf : NULL, + m_width, + false))) + { + /* Cols can be zero if this is a dummy binrows event */ + if (likely(cols != NULL)) + { + memcpy(m_cols.bitmap, cols->bitmap, no_bytes_in_map(cols)); + create_last_word_mask(&m_cols); + } + } + else + { + // Needed because bitmap_init() does not set it to null on failure + m_cols.bitmap= 0; + } +} +#endif + + +Old_rows_log_event::Old_rows_log_event(const char *buf, uint event_len, + Log_event_type event_type, + const Format_description_log_event + *description_event) + : Log_event(buf, description_event), + m_row_count(0), +#ifndef MYSQL_CLIENT + m_table(NULL), +#endif + m_table_id(0), m_rows_buf(0), m_rows_cur(0), m_rows_end(0) +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + , m_curr_row(NULL), m_curr_row_end(NULL), m_key(NULL) +#endif +{ + DBUG_ENTER("Old_rows_log_event::Old_Rows_log_event(const char*,...)"); + uint8 const common_header_len= description_event->common_header_len; + uint8 const post_header_len= description_event->post_header_len[event_type-1]; + + DBUG_PRINT("enter",("event_len: %u common_header_len: %d " + "post_header_len: %d", + event_len, common_header_len, + post_header_len)); + + const char *post_start= buf + common_header_len; + DBUG_DUMP("post_header", (uchar*) post_start, post_header_len); + post_start+= RW_MAPID_OFFSET; + if (post_header_len == 6) + { + /* Master is of an intermediate source tree before 5.1.4. Id is 4 bytes */ + m_table_id= uint4korr(post_start); + post_start+= 4; + } + else + { + m_table_id= (ulong) uint6korr(post_start); + post_start+= RW_FLAGS_OFFSET; + } + + m_flags= uint2korr(post_start); + + uchar const *const var_start= + (const uchar *)buf + common_header_len + post_header_len; + uchar const *const ptr_width= var_start; + uchar *ptr_after_width= (uchar*) ptr_width; + DBUG_PRINT("debug", ("Reading from %p", ptr_after_width)); + m_width = net_field_length(&ptr_after_width); + DBUG_PRINT("debug", ("m_width=%lu", m_width)); + /* if bitmap_init fails, catched in is_valid() */ + if (likely(!bitmap_init(&m_cols, + m_width <= sizeof(m_bitbuf)*8 ? m_bitbuf : NULL, + m_width, + false))) + { + DBUG_PRINT("debug", ("Reading from %p", ptr_after_width)); + memcpy(m_cols.bitmap, ptr_after_width, (m_width + 7) / 8); + create_last_word_mask(&m_cols); + ptr_after_width+= (m_width + 7) / 8; + DBUG_DUMP("m_cols", (uchar*) m_cols.bitmap, no_bytes_in_map(&m_cols)); + } + else + { + // Needed because bitmap_init() does not set it to null on failure + m_cols.bitmap= NULL; + DBUG_VOID_RETURN; + } + + const uchar* const ptr_rows_data= (const uchar*) ptr_after_width; + size_t const data_size= event_len - (ptr_rows_data - (const uchar *) buf); + DBUG_PRINT("info",("m_table_id: %lu m_flags: %d m_width: %lu data_size: %lu", + m_table_id, m_flags, m_width, (ulong) data_size)); + DBUG_DUMP("rows_data", (uchar*) ptr_rows_data, data_size); + + m_rows_buf= (uchar*) my_malloc(data_size, MYF(MY_WME)); + if (likely((bool)m_rows_buf)) + { +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + m_curr_row= m_rows_buf; +#endif + m_rows_end= m_rows_buf + data_size; + m_rows_cur= m_rows_end; + memcpy(m_rows_buf, ptr_rows_data, data_size); + } + else + m_cols.bitmap= 0; // to not free it + + DBUG_VOID_RETURN; +} + + +Old_rows_log_event::~Old_rows_log_event() +{ + if (m_cols.bitmap == m_bitbuf) // no my_malloc happened + m_cols.bitmap= 0; // so no my_free in bitmap_free + bitmap_free(&m_cols); // To pair with bitmap_init(). + my_free((uchar*)m_rows_buf, MYF(MY_ALLOW_ZERO_PTR)); +} + + +int Old_rows_log_event::get_data_size() +{ + uchar buf[sizeof(m_width)+1]; + uchar *end= net_store_length(buf, (m_width + 7) / 8); + + DBUG_EXECUTE_IF("old_row_based_repl_4_byte_map_id_master", + return 6 + no_bytes_in_map(&m_cols) + (end - buf) + + (m_rows_cur - m_rows_buf);); + int data_size= ROWS_HEADER_LEN; + data_size+= no_bytes_in_map(&m_cols); + data_size+= end - buf; + + data_size+= (m_rows_cur - m_rows_buf); + return data_size; +} + + +#ifndef MYSQL_CLIENT +int Old_rows_log_event::do_add_row_data(uchar *row_data, size_t length) +{ + /* + When the table has a primary key, we would probably want, by default, to + log only the primary key value instead of the entire "before image". This + would save binlog space. TODO + */ + DBUG_ENTER("Old_rows_log_event::do_add_row_data"); + DBUG_PRINT("enter", ("row_data: 0x%lx length: %lu", (ulong) row_data, + (ulong) length)); + /* + Don't print debug messages when running valgrind since they can + trigger false warnings. + */ +#ifndef HAVE_purify + DBUG_DUMP("row_data", row_data, min(length, 32)); +#endif + + DBUG_ASSERT(m_rows_buf <= m_rows_cur); + DBUG_ASSERT(!m_rows_buf || m_rows_end && m_rows_buf < m_rows_end); + DBUG_ASSERT(m_rows_cur <= m_rows_end); + + /* The cast will always work since m_rows_cur <= m_rows_end */ + if (static_cast<size_t>(m_rows_end - m_rows_cur) <= length) + { + size_t const block_size= 1024; + my_ptrdiff_t const cur_size= m_rows_cur - m_rows_buf; + my_ptrdiff_t const new_alloc= + block_size * ((cur_size + length + block_size - 1) / block_size); + + uchar* const new_buf= (uchar*)my_realloc((uchar*)m_rows_buf, (uint) new_alloc, + MYF(MY_ALLOW_ZERO_PTR|MY_WME)); + if (unlikely(!new_buf)) + DBUG_RETURN(HA_ERR_OUT_OF_MEM); + + /* If the memory moved, we need to move the pointers */ + if (new_buf != m_rows_buf) + { + m_rows_buf= new_buf; + m_rows_cur= m_rows_buf + cur_size; + } + + /* + The end pointer should always be changed to point to the end of + the allocated memory. + */ + m_rows_end= m_rows_buf + new_alloc; + } + + DBUG_ASSERT(m_rows_cur + length <= m_rows_end); + memcpy(m_rows_cur, row_data, length); + m_rows_cur+= length; + m_row_count++; + DBUG_RETURN(0); +} +#endif + + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) +int Old_rows_log_event::do_apply_event(Relay_log_info const *rli) +{ + DBUG_ENTER("Old_rows_log_event::do_apply_event(Relay_log_info*)"); + int error= 0; + + /* + If m_table_id == ~0UL, then we have a dummy event that does not + contain any data. In that case, we just remove all tables in the + tables_to_lock list, close the thread tables, and return with + success. + */ + if (m_table_id == ~0UL) + { + /* + This one is supposed to be set: just an extra check so that + nothing strange has happened. + */ + DBUG_ASSERT(get_flags(STMT_END_F)); + + const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); + close_thread_tables(thd); + thd->clear_error(); + DBUG_RETURN(0); + } + + /* + 'thd' has been set by exec_relay_log_event(), just before calling + do_apply_event(). We still check here to prevent future coding + errors. + */ + DBUG_ASSERT(rli->sql_thd == thd); + + /* + If there is no locks taken, this is the first binrow event seen + after the table map events. We should then lock all the tables + used in the transaction and proceed with execution of the actual + event. + */ + if (!thd->lock) + { + bool need_reopen= 1; /* To execute the first lap of the loop below */ + + /* + lock_tables() reads the contents of thd->lex, so they must be + initialized. Contrary to in + Table_map_log_event::do_apply_event() we don't call + mysql_init_query() as that may reset the binlog format. + */ + lex_start(thd); + + while ((error= lock_tables(thd, rli->tables_to_lock, + rli->tables_to_lock_count, &need_reopen))) + { + if (!need_reopen) + { + if (thd->is_slave_error || thd->is_fatal_error) + { + /* + Error reporting borrowed from Query_log_event with many excessive + simplifications (we don't honour --slave-skip-errors) + */ + uint actual_error= thd->net.client_last_errno; + rli->report(ERROR_LEVEL, actual_error, + "Error '%s' in %s event: when locking tables", + (actual_error ? thd->net.client_last_error : + "unexpected success or fatal error"), + get_type_str()); + thd->is_fatal_error= 1; + } + else + { + rli->report(ERROR_LEVEL, error, + "Error in %s event: when locking tables", + get_type_str()); + } + const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); + DBUG_RETURN(error); + } + + /* + So we need to reopen the tables. + + We need to flush the pending RBR event, since it keeps a + pointer to an open table. + + ALTERNATIVE SOLUTION (not implemented): Extract a pointer to + the pending RBR event and reset the table pointer after the + tables has been reopened. + + NOTE: For this new scheme there should be no pending event: + need to add code to assert that is the case. + */ + thd->binlog_flush_pending_rows_event(false); + TABLE_LIST *tables= rli->tables_to_lock; + close_tables_for_reopen(thd, &tables); + + uint tables_count= rli->tables_to_lock_count; + if ((error= open_tables(thd, &tables, &tables_count, 0))) + { + if (thd->is_slave_error || thd->is_fatal_error) + { + /* + Error reporting borrowed from Query_log_event with many excessive + simplifications (we don't honour --slave-skip-errors) + */ + uint actual_error= thd->net.client_last_errno; + rli->report(ERROR_LEVEL, actual_error, + "Error '%s' on reopening tables", + (actual_error ? thd->net.client_last_error : + "unexpected success or fatal error")); + thd->is_slave_error= 1; + } + const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); + DBUG_RETURN(error); + } + } + + /* + When the open and locking succeeded, we check all tables to + ensure that they still have the correct type. + + We can use a down cast here since we know that every table added + to the tables_to_lock is a RPL_TABLE_LIST. + */ + + { + RPL_TABLE_LIST *ptr= rli->tables_to_lock; + for ( ; ptr ; ptr= static_cast<RPL_TABLE_LIST*>(ptr->next_global)) + { + if (ptr->m_tabledef.compatible_with(rli, ptr->table)) + { + mysql_unlock_tables(thd, thd->lock); + thd->lock= 0; + thd->is_slave_error= 1; + const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); + DBUG_RETURN(ERR_BAD_TABLE_DEF); + } + } + } + + /* + ... and then we add all the tables to the table map and remove + them from tables to lock. + + We also invalidate the query cache for all the tables, since + they will now be changed. + + TODO [/Matz]: Maybe the query cache should not be invalidated + here? It might be that a table is not changed, even though it + was locked for the statement. We do know that each + Old_rows_log_event contain at least one row, so after processing one + Old_rows_log_event, we can invalidate the query cache for the + associated table. + */ + for (TABLE_LIST *ptr= rli->tables_to_lock ; ptr ; ptr= ptr->next_global) + { + const_cast<Relay_log_info*>(rli)->m_table_map.set_table(ptr->table_id, ptr->table); + } +#ifdef HAVE_QUERY_CACHE + query_cache.invalidate_locked_for_write(rli->tables_to_lock); +#endif + } + + TABLE* + table= + m_table= const_cast<Relay_log_info*>(rli)->m_table_map.get_table(m_table_id); + + if (table) + { + /* + table == NULL means that this table should not be replicated + (this was set up by Table_map_log_event::do_apply_event() + which tested replicate-* rules). + */ + + /* + It's not needed to set_time() but + 1) it continues the property that "Time" in SHOW PROCESSLIST shows how + much slave is behind + 2) it will be needed when we allow replication from a table with no + TIMESTAMP column to a table with one. + So we call set_time(), like in SBR. Presently it changes nothing. + */ + thd->set_time((time_t)when); + /* + There are a few flags that are replicated with each row event. + Make sure to set/clear them before executing the main body of + the event. + */ + if (get_flags(NO_FOREIGN_KEY_CHECKS_F)) + thd->options|= OPTION_NO_FOREIGN_KEY_CHECKS; + else + thd->options&= ~OPTION_NO_FOREIGN_KEY_CHECKS; + + if (get_flags(RELAXED_UNIQUE_CHECKS_F)) + thd->options|= OPTION_RELAXED_UNIQUE_CHECKS; + else + thd->options&= ~OPTION_RELAXED_UNIQUE_CHECKS; + /* A small test to verify that objects have consistent types */ + DBUG_ASSERT(sizeof(thd->options) == sizeof(OPTION_RELAXED_UNIQUE_CHECKS)); + + /* + Now we are in a statement and will stay in a statement until we + see a STMT_END_F. + + We set this flag here, before actually applying any rows, in + case the SQL thread is stopped and we need to detect that we're + inside a statement and halting abruptly might cause problems + when restarting. + */ + const_cast<Relay_log_info*>(rli)->set_flag(Relay_log_info::IN_STMT); + + if ( m_width == table->s->fields && bitmap_is_set_all(&m_cols)) + set_flags(COMPLETE_ROWS_F); + + /* + Set tables write and read sets. + + Read_set contains all slave columns (in case we are going to fetch + a complete record from slave) + + Write_set equals the m_cols bitmap sent from master but it can be + longer if slave has extra columns. + */ + + DBUG_PRINT_BITSET("debug", "Setting table's write_set from: %s", &m_cols); + + bitmap_set_all(table->read_set); + bitmap_set_all(table->write_set); + if (!get_flags(COMPLETE_ROWS_F)) + bitmap_intersect(table->write_set,&m_cols); + + // Do event specific preparations + + error= do_before_row_operations(rli); + + // row processing loop + + while (error == 0 && m_curr_row < m_rows_end) + { + /* in_use can have been set to NULL in close_tables_for_reopen */ + THD* old_thd= table->in_use; + if (!table->in_use) + table->in_use= thd; + + error= do_exec_row(rli); + + table->in_use = old_thd; + switch (error) + { + case 0: + break; + + /* Some recoverable errors */ + case HA_ERR_RECORD_CHANGED: + case HA_ERR_KEY_NOT_FOUND: /* Idempotency support: OK if + tuple does not exist */ + error= 0; + break; + + default: + rli->report(ERROR_LEVEL, thd->net.client_last_errno, + "Error in %s event: row application failed. %s", + get_type_str(), + thd->net.client_last_error ? thd->net.client_last_error : ""); + thd->is_slave_error= 1; + break; + } + + /* + If m_curr_row_end was not set during event execution (e.g., because + of errors) we can't proceed to the next row. If the error is transient + (i.e., error==0 at this point) we must call unpack_current_row() to set + m_curr_row_end. + */ + + DBUG_PRINT("info", ("error: %d", error)); + DBUG_PRINT("info", ("curr_row: 0x%lu; curr_row_end: 0x%lu; rows_end: 0x%lu", + (ulong) m_curr_row, (ulong) m_curr_row_end, (ulong) m_rows_end)); + + if (!m_curr_row_end && !error) + unpack_current_row(rli); + + // at this moment m_curr_row_end should be set + DBUG_ASSERT(error || m_curr_row_end != NULL); + DBUG_ASSERT(error || m_curr_row < m_curr_row_end); + DBUG_ASSERT(error || m_curr_row_end <= m_rows_end); + + m_curr_row= m_curr_row_end; + + } // row processing loop + + DBUG_EXECUTE_IF("STOP_SLAVE_after_first_Rows_event", + const_cast<Relay_log_info*>(rli)->abort_slave= 1;); + error= do_after_row_operations(rli, error); + if (!cache_stmt) + { + DBUG_PRINT("info", ("Marked that we need to keep log")); + thd->options|= OPTION_KEEP_LOG; + } + } // if (table) + + /* + We need to delay this clear until here bacause unpack_current_row() uses + master-side table definitions stored in rli. + */ + if (rli->tables_to_lock && get_flags(STMT_END_F)) + const_cast<Relay_log_info*>(rli)->clear_tables_to_lock(); + + if (error) + { /* error has occured during the transaction */ + rli->report(ERROR_LEVEL, thd->net.client_last_errno, + "Error in %s event: error during transaction execution " + "on table %s.%s. %s", + get_type_str(), table->s->db.str, + table->s->table_name.str, + thd->net.client_last_error ? thd->net.client_last_error : ""); + + /* + If one day we honour --skip-slave-errors in row-based replication, and + the error should be skipped, then we would clear mappings, rollback, + close tables, but the slave SQL thread would not stop and then may + assume the mapping is still available, the tables are still open... + So then we should clear mappings/rollback/close here only if this is a + STMT_END_F. + For now we code, knowing that error is not skippable and so slave SQL + thread is certainly going to stop. + rollback at the caller along with sbr. + */ + thd->reset_current_stmt_binlog_row_based(); + const_cast<Relay_log_info*>(rli)->cleanup_context(thd, error); + thd->is_slave_error= 1; + DBUG_RETURN(error); + } + + /* + This code would ideally be placed in do_update_pos() instead, but + since we have no access to table there, we do the setting of + last_event_start_time here instead. + */ + if (table && (table->s->primary_key == MAX_KEY) && + !cache_stmt && get_flags(STMT_END_F) == RLE_NO_FLAGS) + { + /* + ------------ Temporary fix until WL#2975 is implemented --------- + + This event is not the last one (no STMT_END_F). If we stop now + (in case of terminate_slave_thread()), how will we restart? We + have to restart from Table_map_log_event, but as this table is + not transactional, the rows already inserted will still be + present, and idempotency is not guaranteed (no PK) so we risk + that repeating leads to double insert. So we desperately try to + continue, hope we'll eventually leave this buggy situation (by + executing the final Old_rows_log_event). If we are in a hopeless + wait (reached end of last relay log and nothing gets appended + there), we timeout after one minute, and notify DBA about the + problem. When WL#2975 is implemented, just remove the member + Relay_log_info::last_event_start_time and all its occurrences. + */ + const_cast<Relay_log_info*>(rli)->last_event_start_time= my_time(0); + } + + DBUG_RETURN(0); +} + + +Log_event::enum_skip_reason +Old_rows_log_event::do_shall_skip(Relay_log_info *rli) +{ + /* + If the slave skip counter is 1 and this event does not end a + statement, then we should not start executing on the next event. + Otherwise, we defer the decision to the normal skipping logic. + */ + if (rli->slave_skip_counter == 1 && !get_flags(STMT_END_F)) + return Log_event::EVENT_SKIP_IGNORE; + else + return Log_event::do_shall_skip(rli); +} + +int +Old_rows_log_event::do_update_pos(Relay_log_info *rli) +{ + DBUG_ENTER("Old_rows_log_event::do_update_pos"); + int error= 0; + + DBUG_PRINT("info", ("flags: %s", + get_flags(STMT_END_F) ? "STMT_END_F " : "")); + + if (get_flags(STMT_END_F)) + { + /* + This is the end of a statement or transaction, so close (and + unlock) the tables we opened when processing the + Table_map_log_event starting the statement. + + OBSERVER. This will clear *all* mappings, not only those that + are open for the table. There is not good handle for on-close + actions for tables. + + NOTE. Even if we have no table ('table' == 0) we still need to be + here, so that we increase the group relay log position. If we didn't, we + could have a group relay log position which lags behind "forever" + (assume the last master's transaction is ignored by the slave because of + replicate-ignore rules). + */ + thd->binlog_flush_pending_rows_event(true); + + /* + If this event is not in a transaction, the call below will, if some + transactional storage engines are involved, commit the statement into + them and flush the pending event to binlog. + If this event is in a transaction, the call will do nothing, but a + Xid_log_event will come next which will, if some transactional engines + are involved, commit the transaction and flush the pending event to the + binlog. + */ + error= ha_autocommit_or_rollback(thd, 0); + + /* + Now what if this is not a transactional engine? we still need to + flush the pending event to the binlog; we did it with + thd->binlog_flush_pending_rows_event(). Note that we imitate + what is done for real queries: a call to + ha_autocommit_or_rollback() (sometimes only if involves a + transactional engine), and a call to be sure to have the pending + event flushed. + */ + + thd->reset_current_stmt_binlog_row_based(); + rli->cleanup_context(thd, 0); + if (error == 0) + { + /* + Indicate that a statement is finished. + Step the group log position if we are not in a transaction, + otherwise increase the event log position. + */ + rli->stmt_done(log_pos, when); + + /* + Clear any errors pushed in thd->net.client_last_err* if for + example "no key found" (as this is allowed). This is a safety + measure; apparently those errors (e.g. when executing a + Delete_rows_log_event_old of a non-existing row, like in + rpl_row_mystery22.test, thd->net.client_last_error = "Can't + find record in 't1'" and last_errno=1032) do not become + visible. We still prefer to wipe them out. + */ + thd->clear_error(); + } + else + rli->report(ERROR_LEVEL, error, + "Error in %s event: commit of row events failed, " + "table `%s`.`%s`", + get_type_str(), m_table->s->db.str, + m_table->s->table_name.str); + } + else + { + rli->inc_event_relay_log_pos(); + } + + DBUG_RETURN(error); +} + +#endif /* !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) */ + + +#ifndef MYSQL_CLIENT +bool Old_rows_log_event::write_data_header(IO_CACHE *file) +{ + uchar buf[ROWS_HEADER_LEN]; // No need to init the buffer + + // This method should not be reached. + assert(0); + + DBUG_ASSERT(m_table_id != ~0UL); + DBUG_EXECUTE_IF("old_row_based_repl_4_byte_map_id_master", + { + int4store(buf + 0, m_table_id); + int2store(buf + 4, m_flags); + return (my_b_safe_write(file, buf, 6)); + }); + int6store(buf + RW_MAPID_OFFSET, (ulonglong)m_table_id); + int2store(buf + RW_FLAGS_OFFSET, m_flags); + return (my_b_safe_write(file, buf, ROWS_HEADER_LEN)); +} + + +bool Old_rows_log_event::write_data_body(IO_CACHE*file) +{ + /* + Note that this should be the number of *bits*, not the number of + bytes. + */ + uchar sbuf[sizeof(m_width)]; + my_ptrdiff_t const data_size= m_rows_cur - m_rows_buf; + + // This method should not be reached. + assert(0); + + bool res= false; + uchar *const sbuf_end= net_store_length(sbuf, (size_t) m_width); + DBUG_ASSERT(static_cast<size_t>(sbuf_end - sbuf) <= sizeof(sbuf)); + + DBUG_DUMP("m_width", sbuf, (size_t) (sbuf_end - sbuf)); + res= res || my_b_safe_write(file, sbuf, (size_t) (sbuf_end - sbuf)); + + DBUG_DUMP("m_cols", (uchar*) m_cols.bitmap, no_bytes_in_map(&m_cols)); + res= res || my_b_safe_write(file, (uchar*) m_cols.bitmap, + no_bytes_in_map(&m_cols)); + DBUG_DUMP("rows", m_rows_buf, data_size); + res= res || my_b_safe_write(file, m_rows_buf, (size_t) data_size); + + return res; + +} +#endif + + +#if defined(HAVE_REPLICATION) && !defined(MYSQL_CLIENT) +void Old_rows_log_event::pack_info(Protocol *protocol) +{ + char buf[256]; + char const *const flagstr= + get_flags(STMT_END_F) ? " flags: STMT_END_F" : ""; + size_t bytes= my_snprintf(buf, sizeof(buf), + "table_id: %lu%s", m_table_id, flagstr); + protocol->store(buf, bytes, &my_charset_bin); +} +#endif + + +#ifdef MYSQL_CLIENT +void Old_rows_log_event::print_helper(FILE *file, + PRINT_EVENT_INFO *print_event_info, + char const *const name) +{ + IO_CACHE *const head= &print_event_info->head_cache; + IO_CACHE *const body= &print_event_info->body_cache; + if (!print_event_info->short_form) + { + bool const last_stmt_event= get_flags(STMT_END_F); + print_header(head, print_event_info, !last_stmt_event); + my_b_printf(head, "\t%s: table id %lu%s\n", + name, m_table_id, + last_stmt_event ? " flags: STMT_END_F" : ""); + print_base64(body, print_event_info, !last_stmt_event); + } + + if (get_flags(STMT_END_F)) + { + copy_event_cache_to_file_and_reinit(head, file); + copy_event_cache_to_file_and_reinit(body, file); + } +} +#endif + + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) +/** + Write the current row into event's table. + + The row is located in the row buffer, pointed by @c m_curr_row member. + Number of columns of the row is stored in @c m_width member (it can be + different from the number of columns in the table to which we insert). + Bitmap @c m_cols indicates which columns are present in the row. It is assumed + that event's table is already open and pointed by @c m_table. + + If the same record already exists in the table it can be either overwritten + or an error is reported depending on the value of @c overwrite flag + (error reporting not yet implemented). Note that the matching record can be + different from the row we insert if we use primary keys to identify records in + the table. + + The row to be inserted can contain values only for selected columns. The + missing columns are filled with default values using @c prepare_record() + function. If a matching record is found in the table and @c overwritte is + true, the missing columns are taken from it. + + @param rli Relay log info (needed for row unpacking). + @param overwrite + Shall we overwrite if the row already exists or signal + error (currently ignored). + + @returns Error code on failure, 0 on success. + + This method, if successful, sets @c m_curr_row_end pointer to point at the + next row in the rows buffer. This is done when unpacking the row to be + inserted. + + @note If a matching record is found, it is either updated using + @c ha_update_row() or first deleted and then new record written. +*/ + +int +Old_rows_log_event::write_row(const Relay_log_info *const rli, + const bool overwrite) +{ + DBUG_ENTER("write_row"); + DBUG_ASSERT(m_table != NULL && thd != NULL); + + TABLE *table= m_table; // pointer to event's table + int error; + int keynum; + auto_afree_ptr<char> key(NULL); + + /* fill table->record[0] with default values */ + + if ((error= prepare_record(rli, table, m_width, + TRUE /* check if columns have def. values */))) + DBUG_RETURN(error); + + /* unpack row into table->record[0] */ + error= unpack_current_row(rli); // TODO: how to handle errors? + +#ifndef DBUG_OFF + DBUG_DUMP("record[0]", table->record[0], table->s->reclength); + DBUG_PRINT_BITSET("debug", "write_set = %s", table->write_set); + DBUG_PRINT_BITSET("debug", "read_set = %s", table->read_set); +#endif + + /* + Try to write record. If a corresponding record already exists in the table, + we try to change it using ha_update_row() if possible. Otherwise we delete + it and repeat the whole process again. + + TODO: Add safety measures against infinite looping. + */ + + while ((error= table->file->ha_write_row(table->record[0]))) + { + if (error == HA_ERR_LOCK_DEADLOCK || error == HA_ERR_LOCK_WAIT_TIMEOUT) + { + table->file->print_error(error, MYF(0)); /* to check at exec_relay_log_event */ + DBUG_RETURN(error); + } + if ((keynum= table->file->get_dup_key(error)) < 0) + { + DBUG_PRINT("info",("Can't locate duplicate key (get_dup_key returns %d)",keynum)); + table->file->print_error(error, MYF(0)); + /* + We failed to retrieve the duplicate key + - either because the error was not "duplicate key" error + - or because the information which key is not available + */ + DBUG_RETURN(error); + } + + /* + We need to retrieve the old row into record[1] to be able to + either update or delete the offending record. We either: + + - use rnd_pos() with a row-id (available as dupp_row) to the + offending row, if that is possible (MyISAM and Blackhole), or else + + - use index_read_idx() with the key that is duplicated, to + retrieve the offending row. + */ + if (table->file->ha_table_flags() & HA_DUPLICATE_POS) + { + DBUG_PRINT("info",("Locating offending record using rnd_pos()")); + error= table->file->rnd_pos(table->record[1], table->file->dup_ref); + if (error) + { + DBUG_PRINT("info",("rnd_pos() returns error %d",error)); + table->file->print_error(error, MYF(0)); + DBUG_RETURN(error); + } + } + else + { + DBUG_PRINT("info",("Locating offending record using index_read_idx()")); + + if (table->file->extra(HA_EXTRA_FLUSH_CACHE)) + { + DBUG_PRINT("info",("Error when setting HA_EXTRA_FLUSH_CACHE")); + DBUG_RETURN(my_errno); + } + + if (key.get() == NULL) + { + key.assign(static_cast<char*>(my_alloca(table->s->max_unique_length))); + if (key.get() == NULL) + { + DBUG_PRINT("info",("Can't allocate key buffer")); + DBUG_RETURN(ENOMEM); + } + } + + key_copy((uchar*)key.get(), table->record[0], table->key_info + keynum, + 0); + error= table->file->index_read_idx_map(table->record[1], keynum, + (const uchar*)key.get(), + HA_WHOLE_KEY, + HA_READ_KEY_EXACT); + if (error) + { + DBUG_PRINT("info",("index_read_idx() returns error %d",error)); + table->file->print_error(error, MYF(0)); + DBUG_RETURN(error); + } + } + + /* + Now, record[1] should contain the offending row. That + will enable us to update it or, alternatively, delete it (so + that we can insert the new row afterwards). + */ + + /* + If row is incomplete we will use the record found to fill + missing columns. + */ + if (!get_flags(COMPLETE_ROWS_F)) + { + restore_record(table,record[1]); + error= unpack_current_row(rli); + } + +#ifndef DBUG_OFF + DBUG_PRINT("debug",("preparing for update: before and after image")); + DBUG_DUMP("record[1] (before)", table->record[1], table->s->reclength); + DBUG_DUMP("record[0] (after)", table->record[0], table->s->reclength); +#endif + + /* + REPLACE is defined as either INSERT or DELETE + INSERT. If + possible, we can replace it with an UPDATE, but that will not + work on InnoDB if FOREIGN KEY checks are necessary. + + I (Matz) am not sure of the reason for the last_uniq_key() + check as, but I'm guessing that it's something along the + following lines. + + Suppose that we got the duplicate key to be a key that is not + the last unique key for the table and we perform an update: + then there might be another key for which the unique check will + fail, so we're better off just deleting the row and inserting + the correct row. + */ + if (last_uniq_key(table, keynum) && + !table->file->referenced_by_foreign_key()) + { + DBUG_PRINT("info",("Updating row using ha_update_row()")); + error=table->file->ha_update_row(table->record[1], + table->record[0]); + switch (error) { + + case HA_ERR_RECORD_IS_THE_SAME: + DBUG_PRINT("info",("ignoring HA_ERR_RECORD_IS_THE_SAME error from" + " ha_update_row()")); + error= 0; + + case 0: + break; + + default: + DBUG_PRINT("info",("ha_update_row() returns error %d",error)); + table->file->print_error(error, MYF(0)); + } + + DBUG_RETURN(error); + } + else + { + DBUG_PRINT("info",("Deleting offending row and trying to write new one again")); + if ((error= table->file->ha_delete_row(table->record[1]))) + { + DBUG_PRINT("info",("ha_delete_row() returns error %d",error)); + table->file->print_error(error, MYF(0)); + DBUG_RETURN(error); + } + /* Will retry ha_write_row() with the offending row removed. */ + } + } + + DBUG_RETURN(error); +} + + +/** + Locate the current row in event's table. + + The current row is pointed by @c m_curr_row. Member @c m_width tells how many + columns are there in the row (this can be differnet from the number of columns + in the table). It is assumed that event's table is already open and pointed + by @c m_table. + + If a corresponding record is found in the table it is stored in + @c m_table->record[0]. Note that when record is located based on a primary + key, it is possible that the record found differs from the row being located. + + If no key is specified or table does not have keys, a table scan is used to + find the row. In that case the row should be complete and contain values for + all columns. However, it can still be shorter than the table, i.e. the table + can contain extra columns not present in the row. It is also possible that + the table has fewer columns than the row being located. + + @returns Error code on failure, 0 on success. + + @post In case of success @c m_table->record[0] contains the record found. + Also, the internal "cursor" of the table is positioned at the record found. + + @note If the engine allows random access of the records, a combination of + @c position() and @c rnd_pos() will be used. + */ + +int Old_rows_log_event::find_row(const Relay_log_info *rli) +{ + DBUG_ENTER("find_row"); + + DBUG_ASSERT(m_table && m_table->in_use != NULL); + + TABLE *table= m_table; + int error; + + /* unpack row - missing fields get default values */ + + // TODO: shall we check and report errors here? + prepare_record(NULL,table,m_width,FALSE /* don't check errors */); + error= unpack_current_row(rli); + +#ifndef DBUG_OFF + DBUG_PRINT("info",("looking for the following record")); + DBUG_DUMP("record[0]", table->record[0], table->s->reclength); +#endif + + if ((table->file->ha_table_flags() & HA_PRIMARY_KEY_REQUIRED_FOR_POSITION) && + table->s->primary_key < MAX_KEY) + { + /* + Use a more efficient method to fetch the record given by + table->record[0] if the engine allows it. We first compute a + row reference using the position() member function (it will be + stored in table->file->ref) and the use rnd_pos() to position + the "cursor" (i.e., record[0] in this case) at the correct row. + + TODO: Add a check that the correct record has been fetched by + comparing with the original record. Take into account that the + record on the master and slave can be of different + length. Something along these lines should work: + + ADD>>> store_record(table,record[1]); + int error= table->file->rnd_pos(table->record[0], table->file->ref); + ADD>>> DBUG_ASSERT(memcmp(table->record[1], table->record[0], + table->s->reclength) == 0); + + */ + DBUG_PRINT("info",("locating record using primary key (position)")); + int error= table->file->rnd_pos_by_record(table->record[0]); + if (error) + { + DBUG_PRINT("info",("rnd_pos returns error %d",error)); + table->file->print_error(error, MYF(0)); + } + DBUG_RETURN(error); + } + + // We can't use position() - try other methods. + + /* + We need to retrieve all fields + TODO: Move this out from this function to main loop + */ + table->use_all_columns(); + + /* + Save copy of the record in table->record[1]. It might be needed + later if linear search is used to find exact match. + */ + store_record(table,record[1]); + + if (table->s->keys > 0) + { + DBUG_PRINT("info",("locating record using primary key (index_read)")); + + /* We have a key: search the table using the index */ + if (!table->file->inited && (error= table->file->ha_index_init(0, FALSE))) + { + DBUG_PRINT("info",("ha_index_init returns error %d",error)); + table->file->print_error(error, MYF(0)); + DBUG_RETURN(error); + } + + /* Fill key data for the row */ + + DBUG_ASSERT(m_key); + key_copy(m_key, table->record[0], table->key_info, 0); + + /* + Don't print debug messages when running valgrind since they can + trigger false warnings. + */ +#ifndef HAVE_purify + DBUG_DUMP("key data", m_key, table->key_info->key_length); +#endif + + /* + We need to set the null bytes to ensure that the filler bit are + all set when returning. There are storage engines that just set + the necessary bits on the bytes and don't set the filler bits + correctly. + */ + my_ptrdiff_t const pos= + table->s->null_bytes > 0 ? table->s->null_bytes - 1 : 0; + table->record[0][pos]= 0xFF; + + if ((error= table->file->index_read_map(table->record[0], m_key, + HA_WHOLE_KEY, + HA_READ_KEY_EXACT))) + { + DBUG_PRINT("info",("no record matching the key found in the table")); + table->file->print_error(error, MYF(0)); + table->file->ha_index_end(); + DBUG_RETURN(error); + } + + /* + Don't print debug messages when running valgrind since they can + trigger false warnings. + */ +#ifndef HAVE_purify + DBUG_PRINT("info",("found first matching record")); + DBUG_DUMP("record[0]", table->record[0], table->s->reclength); +#endif + /* + Below is a minor "optimization". If the key (i.e., key number + 0) has the HA_NOSAME flag set, we know that we have found the + correct record (since there can be no duplicates); otherwise, we + have to compare the record with the one found to see if it is + the correct one. + + CAVEAT! This behaviour is essential for the replication of, + e.g., the mysql.proc table since the correct record *shall* be + found using the primary key *only*. There shall be no + comparison of non-PK columns to decide if the correct record is + found. I can see no scenario where it would be incorrect to + chose the row to change only using a PK or an UNNI. + */ + if (table->key_info->flags & HA_NOSAME) + { + table->file->ha_index_end(); + DBUG_RETURN(0); + } + + /* + In case key is not unique, we still have to iterate over records found + and find the one which is identical to the row given. A copy of the + record we are looking for is stored in record[1]. + */ + DBUG_PRINT("info",("non-unique index, scanning it to find matching record")); + + while (record_compare(table)) + { + /* + We need to set the null bytes to ensure that the filler bit + are all set when returning. There are storage engines that + just set the necessary bits on the bytes and don't set the + filler bits correctly. + + TODO[record format ndb]: Remove this code once NDB returns the + correct record format. + */ + if (table->s->null_bytes > 0) + { + table->record[0][table->s->null_bytes - 1]|= + 256U - (1U << table->s->last_null_bit_pos); + } + + if ((error= table->file->index_next(table->record[0]))) + { + DBUG_PRINT("info",("no record matching the given row found")); + table->file->print_error(error, MYF(0)); + table->file->ha_index_end(); + DBUG_RETURN(error); + } + } + + /* + Have to restart the scan to be able to fetch the next row. + */ + table->file->ha_index_end(); + } + else + { + DBUG_PRINT("info",("locating record using table scan (rnd_next)")); + + int restart_count= 0; // Number of times scanning has restarted from top + + /* We don't have a key: search the table using rnd_next() */ + if ((error= table->file->ha_rnd_init(1))) + { + DBUG_PRINT("info",("error initializing table scan" + " (ha_rnd_init returns %d)",error)); + table->file->print_error(error, MYF(0)); + DBUG_RETURN(error); + } + + /* Continue until we find the right record or have made a full loop */ + do + { + error= table->file->rnd_next(table->record[0]); + + switch (error) { + + case 0: + case HA_ERR_RECORD_DELETED: + break; + + case HA_ERR_END_OF_FILE: + if (++restart_count < 2) + table->file->ha_rnd_init(1); + break; + + default: + DBUG_PRINT("info", ("Failed to get next record" + " (rnd_next returns %d)",error)); + table->file->print_error(error, MYF(0)); + table->file->ha_rnd_end(); + DBUG_RETURN(error); + } + } + while (restart_count < 2 && record_compare(table)); + + /* + Note: above record_compare will take into accout all record fields + which might be incorrect in case a partial row was given in the event + */ + + /* + Have to restart the scan to be able to fetch the next row. + */ + if (restart_count == 2) + DBUG_PRINT("info", ("Record not found")); + else + DBUG_DUMP("record found", table->record[0], table->s->reclength); + table->file->ha_rnd_end(); + + DBUG_ASSERT(error == HA_ERR_END_OF_FILE || error == 0); + DBUG_RETURN(error); + } + + DBUG_RETURN(0); +} + +#endif + + +/************************************************************************** + Write_rows_log_event member functions +**************************************************************************/ + +/* + Constructor used to build an event for writing to the binary log. + */ +#if !defined(MYSQL_CLIENT) +Write_rows_log_event_old::Write_rows_log_event_old(THD *thd_arg, + TABLE *tbl_arg, + ulong tid_arg, + MY_BITMAP const *cols, + bool is_transactional) + : Old_rows_log_event(thd_arg, tbl_arg, tid_arg, cols, is_transactional) +{ + + // This constructor should not be reached. + assert(0); + +} +#endif + + +/* + Constructor used by slave to read the event from the binary log. + */ +#ifdef HAVE_REPLICATION +Write_rows_log_event_old::Write_rows_log_event_old(const char *buf, + uint event_len, + const Format_description_log_event + *description_event) +: Old_rows_log_event(buf, event_len, PRE_GA_WRITE_ROWS_EVENT, + description_event) +{ +} +#endif + + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) +int +Write_rows_log_event_old::do_before_row_operations(const Slave_reporting_capability *const) +{ + int error= 0; + + /* + We are using REPLACE semantics and not INSERT IGNORE semantics + when writing rows, that is: new rows replace old rows. We need to + inform the storage engine that it should use this behaviour. + */ + + /* Tell the storage engine that we are using REPLACE semantics. */ + thd->lex->duplicates= DUP_REPLACE; + + /* + Pretend we're executing a REPLACE command: this is needed for + InnoDB and NDB Cluster since they are not (properly) checking the + lex->duplicates flag. + */ + thd->lex->sql_command= SQLCOM_REPLACE; + /* + Do not raise the error flag in case of hitting to an unique attribute + */ + m_table->file->extra(HA_EXTRA_IGNORE_DUP_KEY); + /* + NDB specific: update from ndb master wrapped as Write_rows + */ + /* + so that the event should be applied to replace slave's row + */ + m_table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE); + /* + NDB specific: if update from ndb master wrapped as Write_rows + does not find the row it's assumed idempotent binlog applying + is taking place; don't raise the error. + */ + m_table->file->extra(HA_EXTRA_IGNORE_NO_KEY); + /* + TODO: the cluster team (Tomas?) says that it's better if the engine knows + how many rows are going to be inserted, then it can allocate needed memory + from the start. + */ + m_table->file->ha_start_bulk_insert(0); + /* + We need TIMESTAMP_NO_AUTO_SET otherwise ha_write_row() will not use fill + any TIMESTAMP column with data from the row but instead will use + the event's current time. + As we replicate from TIMESTAMP to TIMESTAMP and slave has no extra + columns, we know that all TIMESTAMP columns on slave will receive explicit + data from the row, so TIMESTAMP_NO_AUTO_SET is ok. + When we allow a table without TIMESTAMP to be replicated to a table having + more columns including a TIMESTAMP column, or when we allow a TIMESTAMP + column to be replicated into a BIGINT column and the slave's table has a + TIMESTAMP column, then the slave's TIMESTAMP column will take its value + from set_time() which we called earlier (consistent with SBR). And then in + some cases we won't want TIMESTAMP_NO_AUTO_SET (will require some code to + analyze if explicit data is provided for slave's TIMESTAMP columns). + */ + m_table->timestamp_field_type= TIMESTAMP_NO_AUTO_SET; + return error; +} + + +int +Write_rows_log_event_old::do_after_row_operations(const Slave_reporting_capability *const, + int error) +{ + int local_error= 0; + m_table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY); + m_table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE); + /* + reseting the extra with + table->file->extra(HA_EXTRA_NO_IGNORE_NO_KEY); + fires bug#27077 + todo: explain or fix + */ + if ((local_error= m_table->file->ha_end_bulk_insert())) + { + m_table->file->print_error(local_error, MYF(0)); + } + return error? error : local_error; +} + + +int +Write_rows_log_event_old::do_exec_row(const Relay_log_info *const rli) +{ + DBUG_ASSERT(m_table != NULL); + int error= write_row(rli, TRUE /* overwrite */); + + if (error && !thd->net.client_last_errno) + thd->net.client_last_errno= error; + + return error; +} + +#endif /* !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) */ + + +#ifdef MYSQL_CLIENT +void Write_rows_log_event_old::print(FILE *file, + PRINT_EVENT_INFO* print_event_info) +{ + Old_rows_log_event::print_helper(file, print_event_info, "Write_rows_old"); +} +#endif + + +/************************************************************************** + Delete_rows_log_event member functions +**************************************************************************/ + +/* + Constructor used to build an event for writing to the binary log. + */ + +#ifndef MYSQL_CLIENT +Delete_rows_log_event_old::Delete_rows_log_event_old(THD *thd_arg, + TABLE *tbl_arg, + ulong tid, + MY_BITMAP const *cols, + bool is_transactional) + : Old_rows_log_event(thd_arg, tbl_arg, tid, cols, is_transactional), + m_after_image(NULL), m_memory(NULL) +{ + + // This constructor should not be reached. + assert(0); + +} +#endif /* #if !defined(MYSQL_CLIENT) */ + + +/* + Constructor used by slave to read the event from the binary log. + */ +#ifdef HAVE_REPLICATION +Delete_rows_log_event_old::Delete_rows_log_event_old(const char *buf, + uint event_len, + const Format_description_log_event + *description_event) + : Old_rows_log_event(buf, event_len, PRE_GA_DELETE_ROWS_EVENT, + description_event), + m_after_image(NULL), m_memory(NULL) +{ +} +#endif + + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + +int +Delete_rows_log_event_old::do_before_row_operations(const Slave_reporting_capability *const) +{ + if ((m_table->file->ha_table_flags() & HA_PRIMARY_KEY_REQUIRED_FOR_POSITION) && + m_table->s->primary_key < MAX_KEY) + { + /* + We don't need to allocate any memory for m_key since it is not used. + */ + return 0; + } + + if (m_table->s->keys > 0) + { + // Allocate buffer for key searches + m_key= (uchar*)my_malloc(m_table->key_info->key_length, MYF(MY_WME)); + if (!m_key) + return HA_ERR_OUT_OF_MEM; + } + return 0; +} + + +int +Delete_rows_log_event_old::do_after_row_operations(const Slave_reporting_capability *const, + int error) +{ + /*error= ToDo:find out what this should really be, this triggers close_scan in nbd, returning error?*/ + m_table->file->ha_index_or_rnd_end(); + my_free(m_key, MYF(MY_ALLOW_ZERO_PTR)); + m_key= NULL; + + return error; +} + + +int Delete_rows_log_event_old::do_exec_row(const Relay_log_info *const rli) +{ + int error; + DBUG_ASSERT(m_table != NULL); + + if (!(error= find_row(rli))) + { + /* + Delete the record found, located in record[0] + */ + error= m_table->file->ha_delete_row(m_table->record[0]); + } + return error; +} + +#endif /* !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) */ + + +#ifdef MYSQL_CLIENT +void Delete_rows_log_event_old::print(FILE *file, + PRINT_EVENT_INFO* print_event_info) +{ + Old_rows_log_event::print_helper(file, print_event_info, "Delete_rows_old"); +} +#endif + + +/************************************************************************** + Update_rows_log_event member functions +**************************************************************************/ + +/* + Constructor used to build an event for writing to the binary log. + */ +#if !defined(MYSQL_CLIENT) +Update_rows_log_event_old::Update_rows_log_event_old(THD *thd_arg, + TABLE *tbl_arg, + ulong tid, + MY_BITMAP const *cols, + bool is_transactional) + : Old_rows_log_event(thd_arg, tbl_arg, tid, cols, is_transactional), + m_after_image(NULL), m_memory(NULL) +{ + + // This constructor should not be reached. + assert(0); +} +#endif /* !defined(MYSQL_CLIENT) */ + + +/* + Constructor used by slave to read the event from the binary log. + */ +#ifdef HAVE_REPLICATION +Update_rows_log_event_old::Update_rows_log_event_old(const char *buf, + uint event_len, + const + Format_description_log_event + *description_event) + : Old_rows_log_event(buf, event_len, PRE_GA_UPDATE_ROWS_EVENT, + description_event), + m_after_image(NULL), m_memory(NULL) +{ +} +#endif + + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + +int +Update_rows_log_event_old::do_before_row_operations(const Slave_reporting_capability *const) +{ + if (m_table->s->keys > 0) + { + // Allocate buffer for key searches + m_key= (uchar*)my_malloc(m_table->key_info->key_length, MYF(MY_WME)); + if (!m_key) + return HA_ERR_OUT_OF_MEM; + } + + m_table->timestamp_field_type= TIMESTAMP_NO_AUTO_SET; + + return 0; +} + + +int +Update_rows_log_event_old::do_after_row_operations(const Slave_reporting_capability *const, + int error) +{ + /*error= ToDo:find out what this should really be, this triggers close_scan in nbd, returning error?*/ + m_table->file->ha_index_or_rnd_end(); + my_free(m_key, MYF(MY_ALLOW_ZERO_PTR)); // Free for multi_malloc + m_key= NULL; + + return error; +} + + +int +Update_rows_log_event_old::do_exec_row(const Relay_log_info *const rli) +{ + DBUG_ASSERT(m_table != NULL); + + int error= find_row(rli); + if (error) + { + /* + We need to read the second image in the event of error to be + able to skip to the next pair of updates + */ + m_curr_row= m_curr_row_end; + unpack_current_row(rli); + return error; + } + + /* + This is the situation after locating BI: + + ===|=== before image ====|=== after image ===|=== + ^ ^ + m_curr_row m_curr_row_end + + BI found in the table is stored in record[0]. We copy it to record[1] + and unpack AI to record[0]. + */ + + store_record(m_table,record[1]); + + m_curr_row= m_curr_row_end; + error= unpack_current_row(rli); // this also updates m_curr_row_end + + /* + Now we have the right row to update. The old row (the one we're + looking for) is in record[1] and the new row is in record[0]. + */ +#ifndef HAVE_purify + /* + Don't print debug messages when running valgrind since they can + trigger false warnings. + */ + DBUG_PRINT("info",("Updating row in table")); + DBUG_DUMP("old record", m_table->record[1], m_table->s->reclength); + DBUG_DUMP("new values", m_table->record[0], m_table->s->reclength); +#endif + + error= m_table->file->ha_update_row(m_table->record[1], m_table->record[0]); + if (error == HA_ERR_RECORD_IS_THE_SAME) + error= 0; + + return error; +} + +#endif /* !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) */ + + +#ifdef MYSQL_CLIENT +void Update_rows_log_event_old::print(FILE *file, + PRINT_EVENT_INFO* print_event_info) +{ + Old_rows_log_event::print_helper(file, print_event_info, "Update_rows_old"); +} +#endif diff --git a/sql/log_event_old.h b/sql/log_event_old.h index 4ae0b00aeac..719802a80fb 100644 --- a/sql/log_event_old.h +++ b/sql/log_event_old.h @@ -20,18 +20,261 @@ Need to include this file at the proper position of log_event.h */ + +/** + @file + + @brief This file contains classes handling old formats of row-based + binlog events. +*/ +/* + Around 2007-10-31, I made these classes completely separated from + the new classes (before, there was a complex class hierarchy + involving multiple inheritance; see BUG#31581), by simply copying + and pasting the entire contents of Rows_log_event into + Old_rows_log_event and the entire contents of + {Write|Update|Delete}_rows_log_event into + {Write|Update|Delete}_rows_log_event_old. For clarity, I will keep + the comments marking which code was cut-and-pasted for some time. + With the classes collapsed into one, there is probably some + redundancy (maybe some methods can be simplified and/or removed), + but we keep them this way for now. /Sven +*/ + + +/** + @class Old_rows_log_event -class Old_rows_log_event + Base class for the three types of row-based events + {Write|Update|Delete}_row_log_event_old, with event type codes + PRE_GA_{WRITE|UPDATE|DELETE}_ROWS_EVENT. These events are never + created any more, except when reading a relay log created by an old + server. +*/ +class Old_rows_log_event : public Log_event { - public: - - virtual ~Old_rows_log_event() {} + /********** BEGIN CUT & PASTE FROM Rows_log_event **********/ +public: + /** + Enumeration of the errors that can be returned. + */ + enum enum_error + { + ERR_OPEN_FAILURE = -1, /**< Failure to open table */ + ERR_OK = 0, /**< No error */ + ERR_TABLE_LIMIT_EXCEEDED = 1, /**< No more room for tables */ + ERR_OUT_OF_MEM = 2, /**< Out of memory */ + ERR_BAD_TABLE_DEF = 3, /**< Table definition does not match */ + ERR_RBR_TO_SBR = 4 /**< daisy-chanining RBR to SBR not allowed */ + }; + + /* + These definitions allow you to combine the flags into an + appropriate flag set using the normal bitwise operators. The + implicit conversion from an enum-constant to an integer is + accepted by the compiler, which is then used to set the real set + of flags. + */ + enum enum_flag + { + /* Last event of a statement */ + STMT_END_F = (1U << 0), + + /* Value of the OPTION_NO_FOREIGN_KEY_CHECKS flag in thd->options */ + NO_FOREIGN_KEY_CHECKS_F = (1U << 1), + + /* Value of the OPTION_RELAXED_UNIQUE_CHECKS flag in thd->options */ + RELAXED_UNIQUE_CHECKS_F = (1U << 2), + + /** + Indicates that rows in this event are complete, that is contain + values for all columns of the table. + */ + COMPLETE_ROWS_F = (1U << 3) + }; + + typedef uint16 flag_set; + + /* Special constants representing sets of flags */ + enum + { + RLE_NO_FLAGS = 0U + }; + + virtual ~Old_rows_log_event(); + + void set_flags(flag_set flags_arg) { m_flags |= flags_arg; } + void clear_flags(flag_set flags_arg) { m_flags &= ~flags_arg; } + flag_set get_flags(flag_set flags_arg) const { return m_flags & flags_arg; } + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + virtual void pack_info(Protocol *protocol); +#endif + +#ifdef MYSQL_CLIENT + /* not for direct call, each derived has its own ::print() */ + virtual void print(FILE *file, PRINT_EVENT_INFO *print_event_info)= 0; +#endif + +#ifndef MYSQL_CLIENT + int add_row_data(uchar *data, size_t length) + { + return do_add_row_data(data,length); + } +#endif + + /* Member functions to implement superclass interface */ + virtual int get_data_size(); + + MY_BITMAP const *get_cols() const { return &m_cols; } + size_t get_width() const { return m_width; } + ulong get_table_id() const { return m_table_id; } + +#ifndef MYSQL_CLIENT + virtual bool write_data_header(IO_CACHE *file); + virtual bool write_data_body(IO_CACHE *file); + virtual const char *get_db() { return m_table->s->db.str; } +#endif + /* + Check that malloc() succeeded in allocating memory for the rows + buffer and the COLS vector. Checking that an Update_rows_log_event_old + is valid is done in the Update_rows_log_event_old::is_valid() + function. + */ + virtual bool is_valid() const + { + return m_rows_buf && m_cols.bitmap; + } + + uint m_row_count; /* The number of rows added to the event */ + +protected: + /* + The constructors are protected since you're supposed to inherit + this class, not create instances of this class. + */ +#ifndef MYSQL_CLIENT + Old_rows_log_event(THD*, TABLE*, ulong table_id, + MY_BITMAP const *cols, bool is_transactional); +#endif + Old_rows_log_event(const char *row_data, uint event_len, + Log_event_type event_type, + const Format_description_log_event *description_event); + +#ifdef MYSQL_CLIENT + void print_helper(FILE *, PRINT_EVENT_INFO *, char const *const name); +#endif + +#ifndef MYSQL_CLIENT + virtual int do_add_row_data(uchar *data, size_t length); +#endif + +#ifndef MYSQL_CLIENT + TABLE *m_table; /* The table the rows belong to */ +#endif + ulong m_table_id; /* Table ID */ + MY_BITMAP m_cols; /* Bitmap denoting columns available */ + ulong m_width; /* The width of the columns bitmap */ + + ulong m_master_reclength; /* Length of record on master side */ + + /* Bit buffers in the same memory as the class */ + uint32 m_bitbuf[128/(sizeof(uint32)*8)]; + uint32 m_bitbuf_ai[128/(sizeof(uint32)*8)]; + + uchar *m_rows_buf; /* The rows in packed format */ + uchar *m_rows_cur; /* One-after the end of the data */ + uchar *m_rows_end; /* One-after the end of the allocated space */ + + flag_set m_flags; /* Flags for row-level events */ + + /* helper functions */ + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + const uchar *m_curr_row; /* Start of the row being processed */ + const uchar *m_curr_row_end; /* One-after the end of the current row */ + uchar *m_key; /* Buffer to keep key value during searches */ + + int find_row(const Relay_log_info *const); + int write_row(const Relay_log_info *const, const bool); + + // Unpack the current row into m_table->record[0] + int unpack_current_row(const Relay_log_info *const rli) + { + DBUG_ASSERT(m_table); + ASSERT_OR_RETURN_ERROR(m_curr_row < m_rows_end, HA_ERR_CORRUPT_EVENT); + int const result= ::unpack_row(rli, m_table, m_width, m_curr_row, &m_cols, + &m_curr_row_end, &m_master_reclength); + ASSERT_OR_RETURN_ERROR(m_curr_row_end <= m_rows_end, HA_ERR_CORRUPT_EVENT); + return result; + } +#endif + +private: + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + virtual int do_apply_event(Relay_log_info const *rli); + virtual int do_update_pos(Relay_log_info *rli); + virtual enum_skip_reason do_shall_skip(Relay_log_info *rli); + + /* + Primitive to prepare for a sequence of row executions. + + DESCRIPTION + + Before doing a sequence of do_prepare_row() and do_exec_row() + calls, this member function should be called to prepare for the + entire sequence. Typically, this member function will allocate + space for any buffers that are needed for the two member + functions mentioned above. + + RETURN VALUE + + The member function will return 0 if all went OK, or a non-zero + error code otherwise. + */ + virtual + int do_before_row_operations(const Slave_reporting_capability *const log) = 0; + + /* + Primitive to clean up after a sequence of row executions. + + DESCRIPTION + + After doing a sequence of do_prepare_row() and do_exec_row(), + this member function should be called to clean up and release + any allocated buffers. + + The error argument, if non-zero, indicates an error which happened during + row processing before this function was called. In this case, even if + function is successful, it should return the error code given in the argument. + */ + virtual + int do_after_row_operations(const Slave_reporting_capability *const log, + int error) = 0; + + /* + Primitive to do the actual execution necessary for a row. + + DESCRIPTION + The member function will do the actual execution needed to handle a row. + The row is located at m_curr_row. When the function returns, + m_curr_row_end should point at the next row (one byte after the end + of the current row). + + RETURN VALUE + 0 if execution succeeded, 1 if execution failed. + + */ + virtual int do_exec_row(const Relay_log_info *const rli) = 0; +#endif /* !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) */ + /********** END OF CUT & PASTE FROM Rows_log_event **********/ protected: #if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) - int do_apply_event(Rows_log_event*,const Relay_log_info*); + int do_apply_event(Old_rows_log_event*,const Relay_log_info*); /* Primitive to prepare for a sequence of row executions. @@ -100,33 +343,61 @@ class Old_rows_log_event }; -class Write_rows_log_event_old - : public Write_rows_log_event, public Old_rows_log_event -{ +/** + @class Write_rows_log_event_old + Old class for binlog events that write new rows to a table (event + type code PRE_GA_WRITE_ROWS_EVENT). Such events are never produced + by this version of the server, but they may be read from a relay log + created by an old server. New servers create events of class + Write_rows_log_event (event type code WRITE_ROWS_EVENT) instead. +*/ +class Write_rows_log_event_old : public Old_rows_log_event +{ + /********** BEGIN CUT & PASTE FROM Write_rows_log_event **********/ public: - enum - { - /* Support interface to THD::binlog_prepare_pending_rows_event */ - TYPE_CODE = PRE_GA_WRITE_ROWS_EVENT - }; - #if !defined(MYSQL_CLIENT) - Write_rows_log_event_old(THD *thd_arg, TABLE *table, ulong table_id, - MY_BITMAP const *cols, bool is_transactional) - : Write_rows_log_event(thd_arg, table, table_id, cols, is_transactional) - { - } + Write_rows_log_event_old(THD*, TABLE*, ulong table_id, + MY_BITMAP const *cols, bool is_transactional); #endif -#if defined(HAVE_REPLICATION) +#ifdef HAVE_REPLICATION Write_rows_log_event_old(const char *buf, uint event_len, - const Format_description_log_event *descr) - : Write_rows_log_event(buf, event_len, descr) + const Format_description_log_event *description_event); +#endif +#if !defined(MYSQL_CLIENT) + static bool binlog_row_logging_function(THD *thd, TABLE *table, + bool is_transactional, + MY_BITMAP *cols, + uint fields, + const uchar *before_record + __attribute__((unused)), + const uchar *after_record) { + return thd->binlog_write_row(table, is_transactional, + cols, fields, after_record); } #endif private: +#ifdef MYSQL_CLIENT + void print(FILE *file, PRINT_EVENT_INFO *print_event_info); +#endif + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + virtual int do_before_row_operations(const Slave_reporting_capability *const); + virtual int do_after_row_operations(const Slave_reporting_capability *const,int); + virtual int do_exec_row(const Relay_log_info *const); +#endif + /********** END OF CUT & PASTE FROM Write_rows_log_event **********/ + +public: + enum + { + /* Support interface to THD::binlog_prepare_pending_rows_event */ + TYPE_CODE = PRE_GA_WRITE_ROWS_EVENT + }; + +private: virtual Log_event_type get_type_code() { return (Log_event_type)TYPE_CODE; } #if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) @@ -145,9 +416,56 @@ private: }; -class Update_rows_log_event_old - : public Update_rows_log_event, public Old_rows_log_event +/** + @class Update_rows_log_event_old + + Old class for binlog events that modify existing rows to a table + (event type code PRE_GA_UPDATE_ROWS_EVENT). Such events are never + produced by this version of the server, but they may be read from a + relay log created by an old server. New servers create events of + class Update_rows_log_event (event type code UPDATE_ROWS_EVENT) + instead. +*/ +class Update_rows_log_event_old : public Old_rows_log_event { + /********** BEGIN CUT & PASTE FROM Update_rows_log_event **********/ +public: +#ifndef MYSQL_CLIENT + Update_rows_log_event_old(THD*, TABLE*, ulong table_id, + MY_BITMAP const *cols, + bool is_transactional); +#endif + +#ifdef HAVE_REPLICATION + Update_rows_log_event_old(const char *buf, uint event_len, + const Format_description_log_event *description_event); +#endif + +#if !defined(MYSQL_CLIENT) + static bool binlog_row_logging_function(THD *thd, TABLE *table, + bool is_transactional, + MY_BITMAP *cols, + uint fields, + const uchar *before_record, + const uchar *after_record) + { + return thd->binlog_update_row(table, is_transactional, + cols, fields, before_record, after_record); + } +#endif + +protected: +#ifdef MYSQL_CLIENT + void print(FILE *file, PRINT_EVENT_INFO *print_event_info); +#endif + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + virtual int do_before_row_operations(const Slave_reporting_capability *const); + virtual int do_after_row_operations(const Slave_reporting_capability *const,int); + virtual int do_exec_row(const Relay_log_info *const); +#endif /* !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) */ + /********** END OF CUT & PASTE FROM Update_rows_log_event **********/ + uchar *m_after_image, *m_memory; public: @@ -157,23 +475,6 @@ public: TYPE_CODE = PRE_GA_UPDATE_ROWS_EVENT }; -#if !defined(MYSQL_CLIENT) - Update_rows_log_event_old(THD *thd_arg, TABLE *table, ulong table_id, - MY_BITMAP const *cols, bool is_transactional) - : Update_rows_log_event(thd_arg, table, table_id, cols, is_transactional), - m_after_image(NULL), m_memory(NULL) - { - } -#endif -#if defined(HAVE_REPLICATION) - Update_rows_log_event_old(const char *buf, uint event_len, - const Format_description_log_event *descr) - : Update_rows_log_event(buf, event_len, descr), - m_after_image(NULL), m_memory(NULL) - { - } -#endif - private: virtual Log_event_type get_type_code() { return (Log_event_type)TYPE_CODE; } @@ -192,9 +493,54 @@ private: }; -class Delete_rows_log_event_old - : public Delete_rows_log_event, public Old_rows_log_event +/** + @class Delete_rows_log_event_old + + Old class for binlog events that delete existing rows from a table + (event type code PRE_GA_DELETE_ROWS_EVENT). Such events are never + produced by this version of the server, but they may be read from a + relay log created by an old server. New servers create events of + class Delete_rows_log_event (event type code DELETE_ROWS_EVENT) + instead. +*/ +class Delete_rows_log_event_old : public Old_rows_log_event { + /********** BEGIN CUT & PASTE FROM Update_rows_log_event **********/ +public: +#ifndef MYSQL_CLIENT + Delete_rows_log_event_old(THD*, TABLE*, ulong, + MY_BITMAP const *cols, bool is_transactional); +#endif +#ifdef HAVE_REPLICATION + Delete_rows_log_event_old(const char *buf, uint event_len, + const Format_description_log_event *description_event); +#endif +#if !defined(MYSQL_CLIENT) + static bool binlog_row_logging_function(THD *thd, TABLE *table, + bool is_transactional, + MY_BITMAP *cols, + uint fields, + const uchar *before_record, + const uchar *after_record + __attribute__((unused))) + { + return thd->binlog_delete_row(table, is_transactional, + cols, fields, before_record); + } +#endif + +protected: +#ifdef MYSQL_CLIENT + void print(FILE *file, PRINT_EVENT_INFO *print_event_info); +#endif + +#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) + virtual int do_before_row_operations(const Slave_reporting_capability *const); + virtual int do_after_row_operations(const Slave_reporting_capability *const,int); + virtual int do_exec_row(const Relay_log_info *const); +#endif + /********** END CUT & PASTE FROM Delete_rows_log_event **********/ + uchar *m_after_image, *m_memory; public: @@ -204,23 +550,6 @@ public: TYPE_CODE = PRE_GA_DELETE_ROWS_EVENT }; -#if !defined(MYSQL_CLIENT) - Delete_rows_log_event_old(THD *thd_arg, TABLE *table, ulong table_id, - MY_BITMAP const *cols, bool is_transactional) - : Delete_rows_log_event(thd_arg, table, table_id, cols, is_transactional), - m_after_image(NULL), m_memory(NULL) - { - } -#endif -#if defined(HAVE_REPLICATION) - Delete_rows_log_event_old(const char *buf, uint event_len, - const Format_description_log_event *descr) - : Delete_rows_log_event(buf, event_len, descr), - m_after_image(NULL), m_memory(NULL) - { - } -#endif - private: virtual Log_event_type get_type_code() { return (Log_event_type)TYPE_CODE; } @@ -240,4 +569,3 @@ private: #endif - diff --git a/sql/mysql_priv.h b/sql/mysql_priv.h index cb12e50e054..3f83d577707 100644 --- a/sql/mysql_priv.h +++ b/sql/mysql_priv.h @@ -1880,6 +1880,7 @@ extern uint volatile thread_count, thread_running, global_read_lock; extern my_bool opt_sql_bin_update, opt_safe_user_create, opt_no_mix_types; extern my_bool opt_safe_show_db, opt_local_infile, opt_myisam_use_mmap; extern my_bool opt_slave_compressed_protocol, use_temp_pool; +extern ulong slave_exec_mode_options; extern my_bool opt_readonly, lower_case_file_system; extern my_bool opt_enable_named_pipe, opt_sync_frm, opt_allow_suspicious_udfs; extern my_bool opt_secure_auth; diff --git a/sql/mysqld.cc b/sql/mysqld.cc index 56c5734146c..03783e56a9f 100644 --- a/sql/mysqld.cc +++ b/sql/mysqld.cc @@ -461,6 +461,8 @@ ulong what_to_log; ulong query_buff_size, slow_launch_time, slave_open_temp_tables; ulong open_files_limit, max_binlog_size, max_relay_log_size; ulong slave_net_timeout, slave_trans_retries; +ulong slave_exec_mode_options; +const char *slave_exec_mode_str= "STRICT"; ulong thread_cache_size=0, thread_pool_size= 0; ulong binlog_cache_size=0, max_binlog_cache_size=0; ulong query_cache_size=0; @@ -610,7 +612,10 @@ char *opt_logname, *opt_slow_logname; /* Static variables */ static bool kill_in_progress, segfaulted; -static my_bool opt_do_pstack, opt_bootstrap, opt_myisam_log; +#ifdef HAVE_STACK_TRACE_ON_SEGV +static my_bool opt_do_pstack; +#endif /* HAVE_STACK_TRACE_ON_SEGV */ +static my_bool opt_bootstrap, opt_myisam_log; static int cleanup_done; static ulong opt_specialflag, opt_myisam_block_size; static char *opt_update_logname, *opt_binlog_index_name; @@ -5335,7 +5340,8 @@ enum options_mysqld OPT_SECURE_FILE_PRIV, OPT_MIN_EXAMINED_ROW_LIMIT, OPT_LOG_SLOW_SLAVE_STATEMENTS, - OPT_DEBUG_CRC, OPT_DEBUG_ON, OPT_OLD_MODE + OPT_DEBUG_CRC, OPT_DEBUG_ON, OPT_OLD_MODE, + OPT_SLAVE_EXEC_MODE }; @@ -5514,9 +5520,11 @@ struct my_option my_long_options[] = (uchar**) &opt_enable_named_pipe, (uchar**) &opt_enable_named_pipe, 0, GET_BOOL, NO_ARG, 0, 0, 0, 0, 0, 0}, #endif +#ifdef HAVE_STACK_TRACE_ON_SEGV {"enable-pstack", OPT_DO_PSTACK, "Print a symbolic stack trace on failure.", (uchar**) &opt_do_pstack, (uchar**) &opt_do_pstack, 0, GET_BOOL, NO_ARG, 0, 0, 0, 0, 0, 0}, +#endif /* HAVE_STACK_TRACE_ON_SEGV */ {"engine-condition-pushdown", OPT_ENGINE_CONDITION_PUSHDOWN, "Push supported query conditions to the storage engine.", @@ -6044,8 +6052,11 @@ replicating a LOAD DATA INFILE command.", (uchar**) &slave_load_tmpdir, (uchar**) &slave_load_tmpdir, 0, GET_STR_ALLOC, REQUIRED_ARG, 0, 0, 0, 0, 0, 0}, {"slave-skip-errors", OPT_SLAVE_SKIP_ERRORS, - "Tells the slave thread to continue replication when a query returns an error from the provided list.", + "Tells the slave thread to continue replication when a query event returns an error from the provided list.", 0, 0, 0, GET_STR, REQUIRED_ARG, 0, 0, 0, 0, 0, 0}, + {"slave-exec-mode", OPT_SLAVE_EXEC_MODE, + "Modes for how replication events should be executed. Legal values are STRICT (default) and IDEMPOTENT. In IDEMPOTENT mode, replication will not stop for operations that are idempotent. In STRICT mode, replication will stop on any unexpected difference between the master and the slave.", + (uchar**) &slave_exec_mode_str, (uchar**) &slave_exec_mode_str, 0, GET_STR, REQUIRED_ARG, 0, 0, 0, 0, 0, 0}, #endif {"slow-query-log", OPT_SLOW_LOG, "Enable|disable slow query log", (uchar**) &opt_slow_log, @@ -7254,6 +7265,9 @@ static void mysql_init_variables(void) /* Things with default values that are not zero */ delay_key_write_options= (uint) DELAY_KEY_WRITE_ON; + slave_exec_mode_options= 0; + slave_exec_mode_options= (uint) + find_bit_type_or_exit(slave_exec_mode_str, &slave_exec_mode_typelib, NULL); opt_specialflag= SPECIAL_ENGLISH; unix_sock= ip_sock= INVALID_SOCKET; mysql_home_ptr= mysql_home; @@ -7467,6 +7481,10 @@ mysqld_get_one_option(int optid, case OPT_SLAVE_SKIP_ERRORS: init_slave_skip_errors(argument); break; + case OPT_SLAVE_EXEC_MODE: + slave_exec_mode_options= (uint) + find_bit_type_or_exit(argument, &slave_exec_mode_typelib, ""); + break; #endif case OPT_SAFEMALLOC_MEM_LIMIT: #if !defined(DBUG_OFF) && defined(SAFEMALLOC) @@ -8017,6 +8035,8 @@ static void get_options(int *argc,char **argv) } /* Set global MyISAM variables from delay_key_write_options */ fix_delay_key_write((THD*) 0, OPT_GLOBAL); + /* Set global slave_exec_mode from its option */ + fix_slave_exec_mode(OPT_GLOBAL); #ifndef EMBEDDED_LIBRARY if (mysqld_chroot) diff --git a/sql/repl_failsafe.cc b/sql/repl_failsafe.cc index d33c81b55e8..589ee8b2605 100644 --- a/sql/repl_failsafe.cc +++ b/sql/repl_failsafe.cc @@ -946,7 +946,7 @@ bool load_master_data(THD* thd) 0, (SLAVE_IO | SLAVE_SQL))) my_message(ER_MASTER_INFO, ER(ER_MASTER_INFO), MYF(0)); strmake(active_mi->master_log_name, row[0], - sizeof(active_mi->master_log_name)); + sizeof(active_mi->master_log_name) -1); active_mi->master_log_pos= my_strtoll10(row[1], (char**) 0, &error_2); /* at least in recent versions, the condition below should be false */ if (active_mi->master_log_pos < BIN_LOG_HEADER_SIZE) diff --git a/sql/rpl_rli.cc b/sql/rpl_rli.cc index 3467f6fd67c..3e9a484126a 100644 --- a/sql/rpl_rli.cc +++ b/sql/rpl_rli.cc @@ -33,6 +33,7 @@ Relay_log_info::Relay_log_info() :Slave_reporting_capability("SQL"), no_storage(FALSE), replicate_same_server_id(::replicate_same_server_id), info_fd(-1), cur_log_fd(-1), save_temporary_tables(0), + group_relay_log_pos(0), cur_log_old_open_count(0), group_master_log_pos(0), log_space_total(0), ignore_log_space_limit(0), last_master_timestamp(0), slave_skip_counter(0), abort_pos_wait(0), slave_run_id(0), sql_thd(0), @@ -1160,6 +1161,11 @@ void Relay_log_info::cleanup_context(THD *thd, bool error) close_thread_tables(thd); clear_tables_to_lock(); clear_flag(IN_STMT); + /* + Cleanup for the flags that have been set at do_apply_event. + */ + thd->options&= ~OPTION_NO_FOREIGN_KEY_CHECKS; + thd->options&= ~OPTION_RELAXED_UNIQUE_CHECKS; last_event_start_time= 0; DBUG_VOID_RETURN; } diff --git a/sql/rpl_utility.cc b/sql/rpl_utility.cc index b3ca26d4c2c..4f4083d9b8f 100644 --- a/sql/rpl_utility.cc +++ b/sql/rpl_utility.cc @@ -164,7 +164,7 @@ uint32 table_def::calc_field_size(uint col, uchar *master_data) const break; } default: - length= -1; + length= ~(uint32) 0; } return length; } diff --git a/sql/set_var.cc b/sql/set_var.cc index 144abbb9a2d..fcdd36dc0f6 100644 --- a/sql/set_var.cc +++ b/sql/set_var.cc @@ -93,6 +93,16 @@ TYPELIB delay_key_write_typelib= delay_key_write_type_names, NULL }; +const char *slave_exec_mode_names[]= +{ "STRICT", "IDEMPOTENT", NullS }; +static const unsigned int slave_exec_mode_names_len[]= +{ sizeof("STRICT") - 1, sizeof("IDEMPOTENT") - 1, 0 }; +TYPELIB slave_exec_mode_typelib= +{ + array_elements(slave_exec_mode_names)-1, "", + slave_exec_mode_names, (unsigned int *) slave_exec_mode_names_len +}; + static int sys_check_ftb_syntax(THD *thd, set_var *var); static bool sys_update_ftb_syntax(THD *thd, set_var * var); static void sys_default_ftb_syntax(THD *thd, enum_var_type type); @@ -417,6 +427,11 @@ static sys_var_const_str_ptr sys_secure_file_priv(&vars, "secure_file_priv", static sys_var_long_ptr sys_server_id(&vars, "server_id", &server_id, fix_server_id); static sys_var_bool_ptr sys_slave_compressed_protocol(&vars, "slave_compressed_protocol", &opt_slave_compressed_protocol); +static sys_var_set_slave_mode slave_exec_mode(&vars, + "slave_exec_mode", + &slave_exec_mode_options, + &slave_exec_mode_typelib, + 0); static sys_var_long_ptr sys_slow_launch_time(&vars, "slow_launch_time", &slow_launch_time); static sys_var_thd_ulong sys_sort_buffer(&vars, "sort_buffer_size", @@ -1003,6 +1018,79 @@ extern void fix_delay_key_write(THD *thd, enum_var_type type) } } +bool sys_var_set::update(THD *thd, set_var *var) +{ + *value= var->save_result.ulong_value; + return 0; +}; + +uchar *sys_var_set::value_ptr(THD *thd, enum_var_type type, + LEX_STRING *base) +{ + char buff[256]; + String tmp(buff, sizeof(buff), &my_charset_latin1); + ulong length; + ulong val= *value; + + tmp.length(0); + for (uint i= 0; val; val>>= 1, i++) + { + if (val & 1) + { + tmp.append(enum_names->type_names[i], + enum_names->type_lengths[i]); + tmp.append(','); + } + } + + if ((length= tmp.length())) + length--; + return (uchar*) thd->strmake(tmp.ptr(), length); +} + +void sys_var_set_slave_mode::set_default(THD *thd, enum_var_type type) +{ + slave_exec_mode_options= 0; + bit_do_set(slave_exec_mode_options, SLAVE_EXEC_MODE_STRICT); +} + +bool sys_var_set_slave_mode::check(THD *thd, set_var *var) +{ + bool rc= sys_var_set::check(thd, var); + if (!rc && + bit_is_set(var->save_result.ulong_value, SLAVE_EXEC_MODE_STRICT) == 1 && + bit_is_set(var->save_result.ulong_value, SLAVE_EXEC_MODE_IDEMPOTENT) == 1) + { + rc= true; + my_error(ER_SLAVE_AMBIGOUS_EXEC_MODE, MYF(0), ""); + } + return rc; +} + +bool sys_var_set_slave_mode::update(THD *thd, set_var *var) +{ + bool rc; + pthread_mutex_lock(&LOCK_global_system_variables); + rc= sys_var_set::update(thd, var); + pthread_mutex_unlock(&LOCK_global_system_variables); + return rc; +} + +void fix_slave_exec_mode(enum_var_type type) +{ + DBUG_ENTER("fix_slave_exec_mode"); + compile_time_assert(sizeof(slave_exec_mode_options) * CHAR_BIT + > SLAVE_EXEC_MODE_LAST_BIT - 1); + if (bit_is_set(slave_exec_mode_options, SLAVE_EXEC_MODE_STRICT) == 1 && + bit_is_set(slave_exec_mode_options, SLAVE_EXEC_MODE_IDEMPOTENT) == 1) + { + sql_print_error("Ambiguous slave modes combination." + " STRICT will be used"); + bit_do_clear(slave_exec_mode_options, SLAVE_EXEC_MODE_IDEMPOTENT); + } + if (bit_is_set(slave_exec_mode_options, SLAVE_EXEC_MODE_IDEMPOTENT) == 0) + bit_do_set(slave_exec_mode_options, SLAVE_EXEC_MODE_STRICT); +} bool sys_var_thd_binlog_format::is_readonly() const { @@ -1116,6 +1204,7 @@ static void fix_trans_mem_root(THD *thd, enum_var_type type) static void fix_server_id(THD *thd, enum_var_type type) { server_id_supplied = 1; + thd->server_id= server_id; } @@ -3280,7 +3369,18 @@ int set_var::light_check(THD *thd) return 0; } +/** + Update variable + @param thd thread handler + @returns 0|1 ok or ERROR + + @note ERROR can be only due to abnormal operations involving + the server's execution evironment such as + out of memory, hard disk failure or the computer blows up. + Consider set_var::check() method if there is a need to return + an error due to logics. +*/ int set_var::update(THD *thd) { if (!value) diff --git a/sql/set_var.h b/sql/set_var.h index 723b31eb188..634bb67760e 100644 --- a/sql/set_var.h +++ b/sql/set_var.h @@ -30,7 +30,8 @@ class sys_var_pluginvar; /* opaque */ typedef struct system_variables SV; typedef struct my_locale_st MY_LOCALE; -extern TYPELIB bool_typelib, delay_key_write_typelib, sql_mode_typelib; +extern TYPELIB bool_typelib, delay_key_write_typelib, sql_mode_typelib, + slave_exec_mode_typelib; typedef int (*sys_check_func)(THD *, set_var *); typedef bool (*sys_update_func)(THD *, set_var *); @@ -804,6 +805,42 @@ public: }; +class sys_var_set :public sys_var +{ +protected: + ulong *value; + TYPELIB *enum_names; +public: + sys_var_set(sys_var_chain *chain, const char *name_arg, ulong *value_arg, + TYPELIB *typelib, sys_after_update_func func) + :sys_var(name_arg, func), value(value_arg), enum_names(typelib) + { chain_sys_var(chain); } + virtual bool check(THD *thd, set_var *var) + { + return check_set(thd, var, enum_names); + } + virtual void set_default(THD *thd, enum_var_type type) + { + *value= 0; + } + bool update(THD *thd, set_var *var); + uchar *value_ptr(THD *thd, enum_var_type type, LEX_STRING *base); + bool check_update_type(Item_result type) { return 0; } + SHOW_TYPE show_type() { return SHOW_CHAR; } +}; + +class sys_var_set_slave_mode :public sys_var_set +{ +public: + sys_var_set_slave_mode(sys_var_chain *chain, const char *name_arg, + ulong *value_arg, + TYPELIB *typelib, sys_after_update_func func) : + sys_var_set(chain, name_arg, value_arg, typelib, func) {} + void set_default(THD *thd, enum_var_type type); + bool check(THD *thd, set_var *var); + bool update(THD *thd, set_var *var); +}; + class sys_var_log_output :public sys_var { ulong *value; @@ -1222,6 +1259,7 @@ sys_var *find_sys_var(THD *thd, const char *str, uint length=0); int sql_set_variables(THD *thd, List<set_var_base> *var_list); bool not_all_support_one_shot(List<set_var_base> *var_list); void fix_delay_key_write(THD *thd, enum_var_type type); +void fix_slave_exec_mode(enum_var_type type); ulong fix_sql_mode(ulong sql_mode); extern sys_var_const_str sys_charset_system; extern sys_var_str sys_init_connect; diff --git a/sql/share/errmsg.txt b/sql/share/errmsg.txt index 02e20ad7c5c..026a0023660 100644 --- a/sql/share/errmsg.txt +++ b/sql/share/errmsg.txt @@ -6114,3 +6114,8 @@ ER_TRG_CANT_OPEN_TABLE ER_CANT_CREATE_SROUTINE eng "Cannot create stored routine `%-.64s`. Check warnings" +ER_SLAVE_AMBIGOUS_EXEC_MODE + eng "Ambiguous slave modes combination. %s" + +ER_NO_FORMAT_DESCRIPTION_EVENT_BEFORE_BINLOG_STATEMENT + eng "The BINLOG statement of type `%s` was not preceded by a format description BINLOG statement." diff --git a/sql/slave.cc b/sql/slave.cc index 45e3d4da090..4ffc2023e85 100644 --- a/sql/slave.cc +++ b/sql/slave.cc @@ -13,6 +13,17 @@ along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ + +/** + @addtogroup Replication + @{ + + @file + + @brief Code to run the io thread and the sql thread on the + replication slave. +*/ + #include "mysql_priv.h" #include <mysql.h> @@ -33,10 +44,6 @@ #include "rpl_tblmap.h" -int queue_event(Master_info* mi,const char* buf,ulong event_len); -static Log_event* next_event(Relay_log_info* rli); - - #define FLAGSTR(V,F) ((V)&(F)?#F" ":"") #define MAX_SLAVE_RETRY_PAUSE 5 @@ -132,6 +139,7 @@ static int create_table_from_dump(THD* thd, MYSQL *mysql, const char* db, const char* table_name, bool overwrite); static int get_master_version_and_clock(MYSQL* mysql, Master_info* mi); static Log_event* next_event(Relay_log_info* rli); +static int queue_event(Master_info* mi,const char* buf,ulong event_len); static int terminate_slave_thread(THD *thd, pthread_mutex_t* term_lock, pthread_cond_t* term_cond, @@ -1775,6 +1783,175 @@ static int has_temporary_error(THD *thd) DBUG_RETURN(0); } + +/** + Applies the given event and advances the relay log position. + + In essence, this function does: + + @code + ev->apply_event(rli); + ev->update_pos(rli); + @endcode + + But it also does some maintainance, such as skipping events if + needed and reporting errors. + + If the @c skip flag is set, then it is tested whether the event + should be skipped, by looking at the slave_skip_counter and the + server id. The skip flag should be set when calling this from a + replication thread but not set when executing an explicit BINLOG + statement. + + @retval 0 OK. + + @retval 1 Error calling ev->apply_event(). + + @retval 2 No error calling ev->apply_event(), but error calling + ev->update_pos(). +*/ +int apply_event_and_update_pos(Log_event* ev, THD* thd, Relay_log_info* rli, + bool skip) +{ + int exec_res= 0; + + DBUG_ENTER("apply_event_and_update_pos"); + + DBUG_PRINT("exec_event",("%s(type_code: %d; server_id: %d)", + ev->get_type_str(), ev->get_type_code(), + ev->server_id)); + DBUG_PRINT("info", ("thd->options: %s%s; rli->last_event_start_time: %lu", + FLAGSTR(thd->options, OPTION_NOT_AUTOCOMMIT), + FLAGSTR(thd->options, OPTION_BEGIN), + rli->last_event_start_time)); + + /* + Execute the event to change the database and update the binary + log coordinates, but first we set some data that is needed for + the thread. + + The event will be executed unless it is supposed to be skipped. + + Queries originating from this server must be skipped. Low-level + events (Format_description_log_event, Rotate_log_event, + Stop_log_event) from this server must also be skipped. But for + those we don't want to modify 'group_master_log_pos', because + these events did not exist on the master. + Format_description_log_event is not completely skipped. + + Skip queries specified by the user in 'slave_skip_counter'. We + can't however skip events that has something to do with the log + files themselves. + + Filtering on own server id is extremely important, to ignore + execution of events created by the creation/rotation of the relay + log (remember that now the relay log starts with its Format_desc, + has a Rotate etc). + */ + + thd->server_id = ev->server_id; // use the original server id for logging + thd->set_time(); // time the query + thd->lex->current_select= 0; + if (!ev->when) + ev->when= my_time(0); + ev->thd = thd; // because up to this point, ev->thd == 0 + + if (skip) + { + int reason= ev->shall_skip(rli); + if (reason == Log_event::EVENT_SKIP_COUNT) + --rli->slave_skip_counter; + pthread_mutex_unlock(&rli->data_lock); + if (reason == Log_event::EVENT_SKIP_NOT) + exec_res= ev->apply_event(rli); +#ifndef DBUG_OFF + /* + This only prints information to the debug trace. + + TODO: Print an informational message to the error log? + */ + static const char *const explain[] = { + // EVENT_SKIP_NOT, + "not skipped", + // EVENT_SKIP_IGNORE, + "skipped because event should be ignored", + // EVENT_SKIP_COUNT + "skipped because event skip counter was non-zero" + }; + DBUG_PRINT("info", ("OPTION_BEGIN: %d; IN_STMT: %d", + thd->options & OPTION_BEGIN ? 1 : 0, + rli->get_flag(Relay_log_info::IN_STMT))); + DBUG_PRINT("skip_event", ("%s event was %s", + ev->get_type_str(), explain[reason])); +#endif + } + else + exec_res= ev->apply_event(rli); + + DBUG_PRINT("info", ("apply_event error = %d", exec_res)); + if (exec_res == 0) + { + int error= ev->update_pos(rli); + char buf[22]; + DBUG_PRINT("info", ("update_pos error = %d", error)); + DBUG_PRINT("info", ("group %s %s", + llstr(rli->group_relay_log_pos, buf), + rli->group_relay_log_name)); + DBUG_PRINT("info", ("event %s %s", + llstr(rli->event_relay_log_pos, buf), + rli->event_relay_log_name)); + /* + The update should not fail, so print an error message and + return an error code. + + TODO: Replace this with a decent error message when merged + with BUG#24954 (which adds several new error message). + */ + if (error) + { + rli->report(ERROR_LEVEL, ER_UNKNOWN_ERROR, + "It was not possible to update the positions" + " of the relay log information: the slave may" + " be in an inconsistent state." + " Stopped in %s position %s", + rli->group_relay_log_name, + llstr(rli->group_relay_log_pos, buf)); + DBUG_RETURN(2); + } + } + + DBUG_RETURN(exec_res ? 1 : 0); +} + + +/** + Top-level function for executing the next event from the relay log. + + This function reads the event from the relay log, executes it, and + advances the relay log position. It also handles errors, etc. + + This function may fail to apply the event for the following reasons: + + - The position specfied by the UNTIL condition of the START SLAVE + command is reached. + + - It was not possible to read the event from the log. + + - The slave is killed. + + - An error occurred when applying the event, and the event has been + tried slave_trans_retries times. If the event has been retried + fewer times, 0 is returned. + + - init_master_info or init_relay_log_pos failed. (These are called + if a failure occurs when applying the event.)</li> + + - An error occurred when updating the binlog position. + + @retval 0 The event was applied. + + @retval 1 The event was not applied. +*/ static int exec_relay_log_event(THD* thd, Relay_log_info* rli) { DBUG_ENTER("exec_relay_log_event"); @@ -1820,117 +1997,26 @@ static int exec_relay_log_event(THD* thd, Relay_log_info* rli) } if (ev) { - int const type_code= ev->get_type_code(); - int exec_res= 0; - - DBUG_PRINT("exec_event",("%s(type_code: %d; server_id: %d)", - ev->get_type_str(), type_code, ev->server_id)); - DBUG_PRINT("info", ("thd->options: %s%s; rli->last_event_start_time: %lu", - FLAGSTR(thd->options, OPTION_NOT_AUTOCOMMIT), - FLAGSTR(thd->options, OPTION_BEGIN), - rli->last_event_start_time)); - + int exec_res= apply_event_and_update_pos(ev, thd, rli, TRUE); /* - Execute the event to change the database and update the binary - log coordinates, but first we set some data that is needed for - the thread. - - The event will be executed unless it is supposed to be skipped. - - Queries originating from this server must be skipped. Low-level - events (Format_description_log_event, Rotate_log_event, - Stop_log_event) from this server must also be skipped. But for - those we don't want to modify 'group_master_log_pos', because - these events did not exist on the master. - Format_description_log_event is not completely skipped. - - Skip queries specified by the user in 'slave_skip_counter'. We - can't however skip events that has something to do with the log - files themselves. - - Filtering on own server id is extremely important, to ignore - execution of events created by the creation/rotation of the relay - log (remember that now the relay log starts with its Format_desc, - has a Rotate etc). + Format_description_log_event should not be deleted because it will be + used to read info about the relay log's format; it will be deleted when + the SQL thread does not need it, i.e. when this thread terminates. */ - - thd->server_id = ev->server_id; // use the original server id for logging - thd->set_time(); // time the query - thd->lex->current_select= 0; - if (!ev->when) - ev->when= my_time(0); - ev->thd = thd; // because up to this point, ev->thd == 0 - - int reason= ev->shall_skip(rli); - if (reason == Log_event::EVENT_SKIP_COUNT) - --rli->slave_skip_counter; - pthread_mutex_unlock(&rli->data_lock); - if (reason == Log_event::EVENT_SKIP_NOT) - exec_res= ev->apply_event(rli); -#ifndef DBUG_OFF - /* - This only prints information to the debug trace. - - TODO: Print an informational message to the error log? - */ - static const char *const explain[] = { - // EVENT_SKIP_NOT, - "not skipped", - // EVENT_SKIP_IGNORE, - "skipped because event should be ignored", - // EVENT_SKIP_COUNT - "skipped because event skip counter was non-zero" - }; - DBUG_PRINT("info", ("OPTION_BEGIN: %d; IN_STMT: %d", - thd->options & OPTION_BEGIN ? 1 : 0, - rli->get_flag(Relay_log_info::IN_STMT))); - DBUG_PRINT("skip_event", ("%s event was %s", - ev->get_type_str(), explain[reason])); -#endif - - DBUG_PRINT("info", ("apply_event error = %d", exec_res)); - if (exec_res == 0) + if (ev->get_type_code() != FORMAT_DESCRIPTION_EVENT) { - int error= ev->update_pos(rli); - char buf[22]; - DBUG_PRINT("info", ("update_pos error = %d", error)); - DBUG_PRINT("info", ("group %s %s", - llstr(rli->group_relay_log_pos, buf), - rli->group_relay_log_name)); - DBUG_PRINT("info", ("event %s %s", - llstr(rli->event_relay_log_pos, buf), - rli->event_relay_log_name)); - /* - The update should not fail, so print an error message and - return an error code. - - TODO: Replace this with a decent error message when merged - with BUG#24954 (which adds several new error message). - */ - if (error) - { - rli->report(ERROR_LEVEL, ER_UNKNOWN_ERROR, - "It was not possible to update the positions" - " of the relay log information: the slave may" - " be in an inconsistent state." - " Stopped in %s position %s", - rli->group_relay_log_name, - llstr(rli->group_relay_log_pos, buf)); - DBUG_RETURN(1); - } + DBUG_PRINT("info", ("Deleting the event after it has been executed")); + delete ev; } /* - Format_description_log_event should not be deleted because it will be - used to read info about the relay log's format; it will be deleted when - the SQL thread does not need it, i.e. when this thread terminates. + update_log_pos failed: this should not happen, so we don't + retry. */ - if (type_code != FORMAT_DESCRIPTION_EVENT) - { - DBUG_PRINT("info", ("Deleting the event after it has been executed")); - delete ev; - } + if (exec_res == 2) + DBUG_RETURN(1); + if (slave_trans_retries) { int temp_err; @@ -2760,7 +2846,7 @@ static int process_io_create_file(Master_info* mi, Create_file_log_event* cev) } if (unlikely(cev_not_written)) { - cev->block = (char*)net->read_pos; + cev->block = net->read_pos; cev->block_len = num_bytes; if (unlikely(mi->rli.relay_log.append(cev))) { @@ -2774,7 +2860,7 @@ static int process_io_create_file(Master_info* mi, Create_file_log_event* cev) } else { - aev.block = (char*)net->read_pos; + aev.block = net->read_pos; aev.block_len = num_bytes; aev.log_pos = cev->log_pos; if (unlikely(mi->rli.relay_log.append(&aev))) @@ -3074,7 +3160,7 @@ static int queue_old_event(Master_info *mi, const char *buf, any >=5.0.0 format. */ -int queue_event(Master_info* mi,const char* buf, ulong event_len) +static int queue_event(Master_info* mi,const char* buf, ulong event_len) { int error= 0; ulong inc_pos; @@ -3960,4 +4046,8 @@ template class I_List_iterator<i_string>; template class I_List_iterator<i_string_pair>; #endif +/** + @} (end of group Replication) +*/ + #endif /* HAVE_REPLICATION */ diff --git a/sql/slave.h b/sql/slave.h index 2cd9ea352ba..f1772bbc1fc 100644 --- a/sql/slave.h +++ b/sql/slave.h @@ -187,6 +187,8 @@ int purge_relay_logs(Relay_log_info* rli, THD *thd, bool just_reset, void set_slave_thread_options(THD* thd); void set_slave_thread_default_charset(THD *thd, Relay_log_info const *rli); void rotate_relay_log(Master_info* mi); +int apply_event_and_update_pos(Log_event* ev, THD* thd, Relay_log_info* rli, + bool skip); pthread_handler_t handle_slave_io(void *arg); pthread_handler_t handle_slave_sql(void *arg); diff --git a/sql/sp_head.cc b/sql/sp_head.cc index 4a0e18129ad..0edb566ed9f 100644 --- a/sql/sp_head.cc +++ b/sql/sp_head.cc @@ -2120,11 +2120,17 @@ sp_head::backpatch(sp_label_t *lab) uint dest= instructions(); List_iterator_fast<bp_t> li(m_backpatch); + DBUG_ENTER("sp_head::backpatch"); while ((bp= li++)) { if (bp->lab == lab) + { + DBUG_PRINT("info", ("backpatch: (m_ip %d, label 0x%lx <%s>) to dest %d", + bp->instr->m_ip, (ulong) lab, lab->name, dest)); bp->instr->backpatch(dest, lab->ctx); + } } + DBUG_VOID_RETURN; } /** diff --git a/sql/sp_head.h b/sql/sp_head.h index 01a4cb73704..8d7062740c8 100644 --- a/sql/sp_head.h +++ b/sql/sp_head.h @@ -888,8 +888,9 @@ public: virtual void backpatch(uint dest, sp_pcontext *dst_ctx) { - if (m_dest == 0) // Don't reset - m_dest= dest; + /* Calling backpatch twice is a logic flaw in jump resolution. */ + DBUG_ASSERT(m_dest == 0); + m_dest= dest; } /** diff --git a/sql/sp_rcontext.cc b/sql/sp_rcontext.cc index 8395648689b..9b237b3e7cc 100644 --- a/sql/sp_rcontext.cc +++ b/sql/sp_rcontext.cc @@ -314,17 +314,91 @@ sp_rcontext::handle_error(uint sql_errno, void sp_rcontext::push_cursor(sp_lex_keeper *lex_keeper, sp_instr_cpush *i) { + DBUG_ENTER("sp_rcontext::push_cursor"); + DBUG_ASSERT(m_ccount < m_root_parsing_ctx->max_cursor_index()); m_cstack[m_ccount++]= new sp_cursor(lex_keeper, i); + DBUG_PRINT("info", ("m_ccount: %d", m_ccount)); + DBUG_VOID_RETURN; } - void sp_rcontext::pop_cursors(uint count) { + DBUG_ENTER("sp_rcontext::pop_cursors"); + DBUG_ASSERT(m_ccount >= count); while (count--) { delete m_cstack[--m_ccount]; } + DBUG_PRINT("info", ("m_ccount: %d", m_ccount)); + DBUG_VOID_RETURN; +} + +void +sp_rcontext::push_handler(struct sp_cond_type *cond, uint h, int type, uint f) +{ + DBUG_ENTER("sp_rcontext::push_handler"); + DBUG_ASSERT(m_hcount < m_root_parsing_ctx->max_handler_index()); + + m_handler[m_hcount].cond= cond; + m_handler[m_hcount].handler= h; + m_handler[m_hcount].type= type; + m_handler[m_hcount].foffset= f; + m_hcount+= 1; + + DBUG_PRINT("info", ("m_hcount: %d", m_hcount)); + DBUG_VOID_RETURN; +} + +void +sp_rcontext::pop_handlers(uint count) +{ + DBUG_ENTER("sp_rcontext::pop_handlers"); + DBUG_ASSERT(m_hcount >= count); + m_hcount-= count; + DBUG_PRINT("info", ("m_hcount: %d", m_hcount)); + DBUG_VOID_RETURN; +} + +void +sp_rcontext::push_hstack(uint h) +{ + DBUG_ENTER("sp_rcontext::push_hstack"); + DBUG_ASSERT(m_hsp < m_root_parsing_ctx->max_handler_index()); + m_hstack[m_hsp++]= h; + DBUG_PRINT("info", ("m_hsp: %d", m_hsp)); + DBUG_VOID_RETURN; +} + +uint +sp_rcontext::pop_hstack() +{ + uint handler; + DBUG_ENTER("sp_rcontext::pop_hstack"); + DBUG_ASSERT(m_hsp); + handler= m_hstack[--m_hsp]; + DBUG_PRINT("info", ("m_hsp: %d", m_hsp)); + DBUG_RETURN(handler); +} + +void +sp_rcontext::enter_handler(int hid) +{ + DBUG_ENTER("sp_rcontext::enter_handler"); + DBUG_ASSERT(m_ihsp < m_root_parsing_ctx->max_handler_index()); + m_in_handler[m_ihsp++]= hid; + DBUG_PRINT("info", ("m_ihsp: %d", m_ihsp)); + DBUG_VOID_RETURN; +} + +void +sp_rcontext::exit_handler() +{ + DBUG_ENTER("sp_rcontext::exit_handler"); + DBUG_ASSERT(m_ihsp); + m_ihsp-= 1; + DBUG_PRINT("info", ("m_ihsp: %d", m_ihsp)); + DBUG_VOID_RETURN; } diff --git a/sql/sp_rcontext.h b/sql/sp_rcontext.h index 43102cfeeb2..368a017da21 100644 --- a/sql/sp_rcontext.h +++ b/sql/sp_rcontext.h @@ -107,21 +107,9 @@ class sp_rcontext : public Sql_alloc return m_return_value_set; } - inline void - push_handler(struct sp_cond_type *cond, uint h, int type, uint f) - { - m_handler[m_hcount].cond= cond; - m_handler[m_hcount].handler= h; - m_handler[m_hcount].type= type; - m_handler[m_hcount].foffset= f; - m_hcount+= 1; - } + void push_handler(struct sp_cond_type *cond, uint h, int type, uint f); - inline void - pop_handlers(uint count) - { - m_hcount-= count; - } + void pop_handlers(uint count); // Returns 1 if a handler was found, 0 otherwise. bool @@ -158,29 +146,13 @@ class sp_rcontext : public Sql_alloc m_hfound= -1; } - inline void - push_hstack(uint h) - { - m_hstack[m_hsp++]= h; - } + void push_hstack(uint h); - inline uint - pop_hstack() - { - return m_hstack[--m_hsp]; - } + uint pop_hstack(); - inline void - enter_handler(int hid) - { - m_in_handler[m_ihsp++]= hid; - } + void enter_handler(int hid); - inline void - exit_handler() - { - m_ihsp-= 1; - } + void exit_handler(); void push_cursor(sp_lex_keeper *lex_keeper, sp_instr_cpush *i); diff --git a/sql/sql_acl.cc b/sql/sql_acl.cc index d2d26da229a..0d563ab9051 100644 --- a/sql/sql_acl.cc +++ b/sql/sql_acl.cc @@ -5579,6 +5579,7 @@ bool mysql_create_user(THD *thd, List <LEX_USER> &list) LEX_USER *user_name, *tmp_user_name; List_iterator <LEX_USER> user_list(list); TABLE_LIST tables[GRANT_TABLES]; + bool some_users_created= FALSE; DBUG_ENTER("mysql_create_user"); /* @@ -5614,6 +5615,7 @@ bool mysql_create_user(THD *thd, List <LEX_USER> &list) continue; } + some_users_created= TRUE; sql_mode= thd->variables.sql_mode; if (replace_user_table(thd, tables[0].table, *user_name, 0, 0, 1, 0)) { @@ -5624,12 +5626,14 @@ bool mysql_create_user(THD *thd, List <LEX_USER> &list) VOID(pthread_mutex_unlock(&acl_cache->lock)); - write_bin_log(thd, FALSE, thd->query, thd->query_length); + if (result) + my_error(ER_CANNOT_USER, MYF(0), "CREATE USER", wrong_users.c_ptr_safe()); + + if (some_users_created) + write_bin_log(thd, FALSE, thd->query, thd->query_length); rw_unlock(&LOCK_grant); close_thread_tables(thd); - if (result) - my_error(ER_CANNOT_USER, MYF(0), "CREATE USER", wrong_users.c_ptr_safe()); DBUG_RETURN(result); } @@ -5654,6 +5658,7 @@ bool mysql_drop_user(THD *thd, List <LEX_USER> &list) LEX_USER *user_name, *tmp_user_name; List_iterator <LEX_USER> user_list(list); TABLE_LIST tables[GRANT_TABLES]; + bool some_users_deleted= FALSE; DBUG_ENTER("mysql_drop_user"); /* @@ -5682,7 +5687,9 @@ bool mysql_drop_user(THD *thd, List <LEX_USER> &list) { append_user(&wrong_users, user_name); result= TRUE; + continue; } + some_users_deleted= TRUE; } /* Rebuild 'acl_check_hosts' since 'acl_users' has been modified */ @@ -5693,7 +5700,8 @@ bool mysql_drop_user(THD *thd, List <LEX_USER> &list) if (result) my_error(ER_CANNOT_USER, MYF(0), "DROP USER", wrong_users.c_ptr_safe()); - write_bin_log(thd, FALSE, thd->query, thd->query_length); + if (some_users_deleted) + write_bin_log(thd, FALSE, thd->query, thd->query_length); rw_unlock(&LOCK_grant); close_thread_tables(thd); @@ -5722,6 +5730,7 @@ bool mysql_rename_user(THD *thd, List <LEX_USER> &list) LEX_USER *user_to, *tmp_user_to; List_iterator <LEX_USER> user_list(list); TABLE_LIST tables[GRANT_TABLES]; + bool some_users_renamed= FALSE; DBUG_ENTER("mysql_rename_user"); /* @@ -5762,7 +5771,9 @@ bool mysql_rename_user(THD *thd, List <LEX_USER> &list) { append_user(&wrong_users, user_from); result= TRUE; + continue; } + some_users_renamed= TRUE; } /* Rebuild 'acl_check_hosts' since 'acl_users' has been modified */ @@ -5770,12 +5781,14 @@ bool mysql_rename_user(THD *thd, List <LEX_USER> &list) VOID(pthread_mutex_unlock(&acl_cache->lock)); - write_bin_log(thd, FALSE, thd->query, thd->query_length); + if (result) + my_error(ER_CANNOT_USER, MYF(0), "RENAME USER", wrong_users.c_ptr_safe()); + + if (some_users_renamed && mysql_bin_log.is_open()) + write_bin_log(thd, FALSE, thd->query, thd->query_length); rw_unlock(&LOCK_grant); close_thread_tables(thd); - if (result) - my_error(ER_CANNOT_USER, MYF(0), "RENAME USER", wrong_users.c_ptr_safe()); DBUG_RETURN(result); } diff --git a/sql/sql_base.cc b/sql/sql_base.cc index 9e4dceff87d..799cb673a0e 100644 --- a/sql/sql_base.cc +++ b/sql/sql_base.cc @@ -3934,7 +3934,7 @@ retry: READ_KEYINFO | COMPUTE_TYPES | EXTRA_RECORD, ha_open_options | HA_OPEN_FOR_REPAIR, entry, FALSE) || ! entry->file || - (entry->file->is_crashed() && entry->file->check_and_repair(thd))) + (entry->file->is_crashed() && entry->file->ha_check_and_repair(thd))) { /* Give right error message */ thd->clear_error(); @@ -5400,7 +5400,7 @@ bool rm_temporary_table(handlerton *base, char *path) error=1; /* purecov: inspected */ *ext= 0; // remove extension file= get_new_handler((TABLE_SHARE*) 0, current_thd->mem_root, base); - if (file && file->delete_table(path)) + if (file && file->ha_delete_table(path)) { error=1; sql_print_warning("Could not remove temporary table: '%s', error: %d", @@ -6400,7 +6400,36 @@ find_item_in_list(Item *find, List<Item> &items, uint *counter, *resolution= RESOLVED_IGNORING_ALIAS; break; } - } + } + else if (table_name && item->type() == Item::REF_ITEM && + ((Item_ref *)item)->ref_type() == Item_ref::VIEW_REF) + { + /* + TODO:Here we process prefixed view references only. What we should + really do is process all types of Item_refs. But this will currently + lead to a clash with the way references to outer SELECTs (from the + HAVING clause) are handled in e.g. : + SELECT 1 FROM t1 AS t1_o GROUP BY a + HAVING (SELECT t1_o.a FROM t1 AS t1_i GROUP BY t1_i.a LIMIT 1). + Processing all Item_refs here will cause t1_o.a to resolve to itself. + We still need to process the special case of Item_direct_view_ref + because in the context of views they have the same meaning as + Item_field for tables. + */ + Item_ident *item_ref= (Item_ident *) item; + if (item_ref->name && item_ref->table_name && + !my_strcasecmp(system_charset_info, item_ref->name, field_name) && + !my_strcasecmp(table_alias_charset, item_ref->table_name, + table_name) && + (!db_name || (item_ref->db_name && + !strcmp (item_ref->db_name, db_name)))) + { + found= li.ref(); + *counter= i; + *resolution= RESOLVED_IGNORING_ALIAS; + break; + } + } } if (!found) { @@ -8070,7 +8099,7 @@ my_bool mysql_rm_tmp_tables(void) ((handler_file= get_new_handler(&share, thd->mem_root, share.db_type())))) { - handler_file->delete_table(filePathCopy); + handler_file->ha_delete_table(filePathCopy); delete handler_file; } free_table_share(&share); diff --git a/sql/sql_binlog.cc b/sql/sql_binlog.cc index 77c5155b41b..04f408453ea 100644 --- a/sql/sql_binlog.cc +++ b/sql/sql_binlog.cc @@ -17,15 +17,14 @@ #include "rpl_rli.h" #include "base64.h" -/* +/** Execute a BINLOG statement - TODO: This currently assumes a MySQL 5.x binlog. - When we'll have binlog with a different format, to execute the - BINLOG command properly the server will need to know which format - the BINLOG command's event is in. mysqlbinlog should then send - the Format_description_log_event of the binlog it reads and the - server thread should cache this format into + To execute the BINLOG command properly the server needs to know + which format the BINLOG command's event is in. Therefore, the first + BINLOG statement seen must be a base64 encoding of the + Format_description_log_event, as outputted by mysqlbinlog. This + Format_description_log_event is cached in rli->description_event_for_exec. */ @@ -47,11 +46,24 @@ void mysql_client_binlog_statement(THD* thd) /* Allocation */ + + /* + If we do not have a Format_description_event, we create a dummy + one here. In this case, the first event we read must be a + Format_description_event. + */ + my_bool have_fd_event= TRUE; if (!thd->rli_fake) + { thd->rli_fake= new Relay_log_info; - - const Format_description_log_event *desc= - new Format_description_log_event(4); + have_fd_event= FALSE; + } + if (thd->rli_fake && !thd->rli_fake->relay_log.description_event_for_exec) + { + thd->rli_fake->relay_log.description_event_for_exec= + new Format_description_log_event(4); + have_fd_event= FALSE; + } const char *error= 0; char *buf= (char *) my_malloc(decoded_len, MYF(MY_WME)); @@ -60,7 +72,9 @@ void mysql_client_binlog_statement(THD* thd) /* Out of memory check */ - if (!(thd->rli_fake && desc && buf)) + if (!(thd->rli_fake && + thd->rli_fake->relay_log.description_event_for_exec && + buf)) { my_error(ER_OUTOFMEMORY, MYF(0), 1); /* needed 1 bytes */ goto end; @@ -131,7 +145,28 @@ void mysql_client_binlog_statement(THD* thd) goto end; } - ev= Log_event::read_log_event(bufptr, event_len, &error, desc); + /* + If we have not seen any Format_description_event, then we must + see one; it is the only statement that can be read in base64 + without a prior Format_description_event. + */ + if (!have_fd_event) + { + if (bufptr[EVENT_TYPE_OFFSET] == FORMAT_DESCRIPTION_EVENT) + have_fd_event= TRUE; + else + { + my_error(ER_NO_FORMAT_DESCRIPTION_EVENT_BEFORE_BINLOG_STATEMENT, + MYF(0), + Log_event::get_type_str( + (Log_event_type)bufptr[EVENT_TYPE_OFFSET])); + goto end; + } + } + + ev= Log_event::read_log_event(bufptr, event_len, &error, + thd->rli_fake->relay_log. + description_event_for_exec); DBUG_PRINT("info",("binlog base64 err=%s", error)); if (!ev) @@ -167,11 +202,10 @@ void mysql_client_binlog_statement(THD* thd) Neither do we have to update the log positions, since that is not used at all: the rli_fake instance is used only for error reporting. - */ + */ #if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION) - if (IF_DBUG(int err= ) ev->apply_event(thd->rli_fake)) + if (apply_event_and_update_pos(ev, thd, thd->rli_fake, FALSE)) { - DBUG_PRINT("info", ("apply_event() returned: %d", err)); /* TODO: Maybe a better error message since the BINLOG statement now contains several events. @@ -181,7 +215,14 @@ void mysql_client_binlog_statement(THD* thd) } #endif - delete ev; + /* + Format_description_log_event should not be deleted because it + will be used to read info about the relay log's format; it + will be deleted when the SQL thread does not need it, + i.e. when this thread terminates. + */ + if (ev->get_type_code() != FORMAT_DESCRIPTION_EVENT) + delete ev; ev= 0; } } @@ -191,7 +232,6 @@ void mysql_client_binlog_statement(THD* thd) send_ok(thd); end: - delete desc; my_free(buf, MYF(MY_ALLOW_ZERO_PTR)); DBUG_VOID_RETURN; } diff --git a/sql/sql_cache.cc b/sql/sql_cache.cc index e51c53f644e..7e54f87fe14 100644 --- a/sql/sql_cache.cc +++ b/sql/sql_cache.cc @@ -371,6 +371,32 @@ TODO list: __LINE__,(ulong)(B)));B->query()->unlock_reading();} #define DUMP(C) DBUG_EXECUTE("qcache", {\ (C)->cache_dump(); (C)->queries_dump();(C)->tables_dump();}) + + +/** + Causes the thread to wait in a spin lock for a query kill signal. + This function is used by the test frame work to identify race conditions. + + The signal is caught and ignored and the thread is not killed. +*/ + +static void debug_wait_for_kill(const char *info) +{ + DBUG_ENTER("debug_wait_for_kill"); + const char *prev_info; + THD *thd; + thd= current_thd; + prev_info= thd->proc_info; + thd->proc_info= info; + sql_print_information(info); + while(!thd->killed) + my_sleep(1000); + thd->killed= THD::NOT_KILLED; + sql_print_information("Exit debug_wait_for_kill"); + thd->proc_info= prev_info; + DBUG_VOID_RETURN; +} + #else #define MUTEX_LOCK(M) pthread_mutex_lock(M) #define MUTEX_UNLOCK(M) pthread_mutex_unlock(M) @@ -647,13 +673,16 @@ void query_cache_insert(NET *net, const char *packet, ulong length) if (net->query_cache_query == 0) DBUG_VOID_RETURN; + DBUG_EXECUTE_IF("wait_in_query_cache_insert", + debug_wait_for_kill("wait_in_query_cache_insert"); ); + STRUCT_LOCK(&query_cache.structure_guard_mutex); bool interrupt; query_cache.wait_while_table_flush_is_in_progress(&interrupt); if (interrupt) { STRUCT_UNLOCK(&query_cache.structure_guard_mutex); - return; + DBUG_VOID_RETURN; } Query_cache_block *query_block= (Query_cache_block*)net->query_cache_query; @@ -667,11 +696,11 @@ void query_cache_insert(NET *net, const char *packet, ulong length) DBUG_VOID_RETURN; } + BLOCK_LOCK_WR(query_block); Query_cache_query *header= query_block->query(); Query_cache_block *result= header->result(); DUMP(&query_cache); - BLOCK_LOCK_WR(query_block); DBUG_PRINT("qcache", ("insert packet %lu bytes long",length)); /* @@ -687,6 +716,7 @@ void query_cache_insert(NET *net, const char *packet, ulong length) DBUG_PRINT("qcache", ("free query 0x%lx", (ulong) query_block)); // The following call will remove the lock on query_block query_cache.free_query(query_block); + query_cache.refused++; // append_result_data no success => we need unlock STRUCT_UNLOCK(&query_cache.structure_guard_mutex); DBUG_VOID_RETURN; @@ -886,6 +916,31 @@ ulong Query_cache::resize(ulong query_cache_size_arg) m_cache_status= Query_cache::FLUSH_IN_PROGRESS; STRUCT_UNLOCK(&structure_guard_mutex); + /* + Wait for all readers and writers to exit. When the list of all queries + is iterated over with a block level lock, we are done. + */ + Query_cache_block *block= queries_blocks; + if (block) + { + do + { + BLOCK_LOCK_WR(block); + Query_cache_query *query= block->query(); + if (query && query->writer()) + { + /* + Drop the writer; this will cancel any attempts to store + the processed statement associated with this writer. + */ + query->writer()->query_cache_query= 0; + query->writer(0); + refused++; + } + BLOCK_UNLOCK_WR(block); + block= block->next; + } while (block != queries_blocks); + } free_cache(); query_cache_size= query_cache_size_arg; diff --git a/sql/sql_class.h b/sql/sql_class.h index 2342f7326a2..bda8ae43917 100644 --- a/sql/sql_class.h +++ b/sql/sql_class.h @@ -38,6 +38,9 @@ enum enum_ha_read_modes { RFIRST, RNEXT, RPREV, RLAST, RKEY, RNEXT_SAME }; enum enum_duplicates { DUP_ERROR, DUP_REPLACE, DUP_UPDATE }; enum enum_delay_key_write { DELAY_KEY_WRITE_NONE, DELAY_KEY_WRITE_ON, DELAY_KEY_WRITE_ALL }; +enum enum_slave_exec_mode { SLAVE_EXEC_MODE_STRICT, + SLAVE_EXEC_MODE_IDEMPOTENT, + SLAVE_EXEC_MODE_LAST_BIT}; enum enum_mark_columns { MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE}; diff --git a/sql/sql_db.cc b/sql/sql_db.cc index f158b5488d0..b5f49b97ec9 100644 --- a/sql/sql_db.cc +++ b/sql/sql_db.cc @@ -1396,7 +1396,7 @@ static void backup_current_db_name(THD *thd, } else { - strmake(saved_db_name->str, thd->db, saved_db_name->length); + strmake(saved_db_name->str, thd->db, saved_db_name->length - 1); saved_db_name->length= thd->db_length; } } diff --git a/sql/sql_delete.cc b/sql/sql_delete.cc index 1f6216efb2e..adaf4d537f1 100644 --- a/sql/sql_delete.cc +++ b/sql/sql_delete.cc @@ -123,7 +123,7 @@ bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds, table->file->info(HA_STATUS_VARIABLE | HA_STATUS_NO_LOCK); ha_rows const maybe_deleted= table->file->stats.records; DBUG_PRINT("debug", ("Trying to use delete_all_rows()")); - if (!(error=table->file->delete_all_rows())) + if (!(error=table->file->ha_delete_all_rows())) { error= -1; // ok deleted= maybe_deleted; @@ -328,7 +328,7 @@ bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds, We're really doing a truncate and need to reset the table's auto-increment counter. */ - int error2= table->file->reset_auto_increment(0); + int error2= table->file->ha_reset_auto_increment(0); if (error2 && (error2 != HA_ERR_WRONG_COMMAND)) { diff --git a/sql/sql_insert.cc b/sql/sql_insert.cc index 1dc256bb7c3..cb0728490b7 100644 --- a/sql/sql_insert.cc +++ b/sql/sql_insert.cc @@ -619,7 +619,8 @@ bool mysql_insert(THD *thd,TABLE_LIST *table_list, if (mysql_prepare_insert(thd, table_list, table, fields, values, update_fields, update_values, duplic, &unused_conds, FALSE, - (fields.elements || !value_count), + (fields.elements || !value_count || + table_list->view != 0), !ignore && (thd->variables.sql_mode & (MODE_STRICT_TRANS_TABLES | MODE_STRICT_ALL_TABLES)))) @@ -1862,7 +1863,6 @@ bool delayed_get_table(THD *thd, TABLE_LIST *table_list) { if (!(di= new Delayed_insert())) { - my_error(ER_OUTOFMEMORY,MYF(0),sizeof(Delayed_insert)); thd->fatal_error(); goto end_create; } diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc index a739ddcc194..25584797549 100644 --- a/sql/sql_parse.cc +++ b/sql/sql_parse.cc @@ -1318,7 +1318,6 @@ bool dispatch_command(enum enum_server_command command, THD *thd, unregister_slave(thd,1,1); /* fake COM_QUIT -- if we get here, the thread needs to terminate */ error = TRUE; - net->error = 0; break; } #endif diff --git a/sql/sql_partition.cc b/sql/sql_partition.cc index ce70e177a85..eabf4526f7b 100644 --- a/sql/sql_partition.cc +++ b/sql/sql_partition.cc @@ -5108,9 +5108,9 @@ static bool mysql_change_partitions(ALTER_PARTITION_PARAM_TYPE *lpt) DBUG_ENTER("mysql_change_partitions"); build_table_filename(path, sizeof(path), lpt->db, lpt->table_name, "", 0); - if ((error= file->change_partitions(lpt->create_info, path, &lpt->copied, - &lpt->deleted, lpt->pack_frm_data, - lpt->pack_frm_len))) + if ((error= file->ha_change_partitions(lpt->create_info, path, &lpt->copied, + &lpt->deleted, lpt->pack_frm_data, + lpt->pack_frm_len))) { if (error != ER_OUTOFMEMORY) file->print_error(error, MYF(0)); @@ -5148,7 +5148,7 @@ static bool mysql_rename_partitions(ALTER_PARTITION_PARAM_TYPE *lpt) DBUG_ENTER("mysql_rename_partitions"); build_table_filename(path, sizeof(path), lpt->db, lpt->table_name, "", 0); - if ((error= lpt->table->file->rename_partitions(path))) + if ((error= lpt->table->file->ha_rename_partitions(path))) { if (error != 1) lpt->table->file->print_error(error, MYF(0)); @@ -5189,7 +5189,7 @@ static bool mysql_drop_partitions(ALTER_PARTITION_PARAM_TYPE *lpt) DBUG_ENTER("mysql_drop_partitions"); build_table_filename(path, sizeof(path), lpt->db, lpt->table_name, "", 0); - if ((error= lpt->table->file->drop_partitions(path))) + if ((error= lpt->table->file->ha_drop_partitions(path))) { lpt->table->file->print_error(error, MYF(0)); DBUG_RETURN(TRUE); @@ -6105,13 +6105,13 @@ uint fast_alter_partition_table(THD *thd, TABLE *table, int error; written_bin_log= FALSE; if (((alter_info->flags & ALTER_OPTIMIZE_PARTITION) && - (error= table->file->optimize_partitions(thd))) || + (error= table->file->ha_optimize_partitions(thd))) || ((alter_info->flags & ALTER_ANALYZE_PARTITION) && - (error= table->file->analyze_partitions(thd))) || + (error= table->file->ha_analyze_partitions(thd))) || ((alter_info->flags & ALTER_CHECK_PARTITION) && - (error= table->file->check_partitions(thd))) || + (error= table->file->ha_check_partitions(thd))) || ((alter_info->flags & ALTER_REPAIR_PARTITION) && - (error= table->file->repair_partitions(thd)))) + (error= table->file->ha_repair_partitions(thd)))) { table->file->print_error(error, MYF(0)); goto err; diff --git a/sql/sql_repl.cc b/sql/sql_repl.cc index 1e05e19c6f5..cdce2fa695b 100644 --- a/sql/sql_repl.cc +++ b/sql/sql_repl.cc @@ -1243,9 +1243,6 @@ bool change_master(THD* thd, Master_info* mi) DBUG_RETURN(TRUE); } } - mi->rli.group_master_log_pos = mi->master_log_pos; - DBUG_PRINT("info", ("master_log_pos: %lu", (ulong) mi->master_log_pos)); - /* Coordinates in rli were spoilt by the 'if (need_relay_log_purge)' block, so restore them to good values. If we left them to ''/0, that would work; @@ -1257,6 +1254,7 @@ bool change_master(THD* thd, Master_info* mi) That's why we always save good coords in rli. */ mi->rli.group_master_log_pos= mi->master_log_pos; + DBUG_PRINT("info", ("master_log_pos: %lu", (ulong) mi->master_log_pos)); strmake(mi->rli.group_master_log_name,mi->master_log_name, sizeof(mi->rli.group_master_log_name)-1); @@ -1376,6 +1374,11 @@ bool mysql_show_binlog_events(THD* thd) if ((file=open_binlog(&log, linfo.log_file_name, &errmsg)) < 0) goto err; + /* + to account binlog event header size + */ + thd->variables.max_allowed_packet += MAX_LOG_EVENT_HEADER; + pthread_mutex_lock(log_lock); /* @@ -1386,7 +1389,6 @@ bool mysql_show_binlog_events(THD* thd) This code will fail on a mixed relay log (one which has Format_desc then Rotate then Format_desc). */ - ev = Log_event::read_log_event(&log,(pthread_mutex_t*)0,description_event); if (ev) { @@ -1578,39 +1580,54 @@ err: DBUG_RETURN(TRUE); } - +/** + Load data's io cache specific hook to be executed + before a chunk of data is being read into the cache's buffer + The fuction instantianates and writes into the binlog + replication events along LOAD DATA processing. + + @param file pointer to io-cache + @return 0 +*/ int log_loaded_block(IO_CACHE* file) { + DBUG_ENTER("log_loaded_block"); LOAD_FILE_INFO *lf_info; - uint block_len ; - - /* file->request_pos contains position where we started last read */ - char* buffer = (char*) file->request_pos; - if (!(block_len = (char*) file->read_end - (char*) buffer)) - return 0; - lf_info = (LOAD_FILE_INFO*) file->arg; + uint block_len; + /* buffer contains position where we started last read */ + uchar* buffer= (uchar*) my_b_get_buffer_start(file); + uint max_event_size= current_thd->variables.max_allowed_packet; + lf_info= (LOAD_FILE_INFO*) file->arg; if (lf_info->thd->current_stmt_binlog_row_based) - return 0; + DBUG_RETURN(0); if (lf_info->last_pos_in_file != HA_POS_ERROR && - lf_info->last_pos_in_file >= file->pos_in_file) - return 0; - lf_info->last_pos_in_file = file->pos_in_file; - if (lf_info->wrote_create_file) - { - Append_block_log_event a(lf_info->thd, lf_info->thd->db, buffer, - block_len, lf_info->log_delayed); - mysql_bin_log.write(&a); - } - else + lf_info->last_pos_in_file >= my_b_get_pos_in_file(file)) + DBUG_RETURN(0); + + for (block_len= my_b_get_bytes_in_buffer(file); block_len > 0; + buffer += min(block_len, max_event_size), + block_len -= min(block_len, max_event_size)) { - Begin_load_query_log_event b(lf_info->thd, lf_info->thd->db, - buffer, block_len, - lf_info->log_delayed); - mysql_bin_log.write(&b); - lf_info->wrote_create_file = 1; - DBUG_SYNC_POINT("debug_lock.created_file_event",10); + lf_info->last_pos_in_file= my_b_get_pos_in_file(file); + if (lf_info->wrote_create_file) + { + Append_block_log_event a(lf_info->thd, lf_info->thd->db, buffer, + min(block_len, max_event_size), + lf_info->log_delayed); + mysql_bin_log.write(&a); + } + else + { + Begin_load_query_log_event b(lf_info->thd, lf_info->thd->db, + buffer, + min(block_len, max_event_size), + lf_info->log_delayed); + mysql_bin_log.write(&b); + lf_info->wrote_create_file= 1; + DBUG_SYNC_POINT("debug_lock.created_file_event",10); + } } - return 0; + DBUG_RETURN(0); } /* diff --git a/sql/sql_select.cc b/sql/sql_select.cc index 3a286a69e20..326fb39cc17 100644 --- a/sql/sql_select.cc +++ b/sql/sql_select.cc @@ -1579,14 +1579,14 @@ JOIN::reinit() if (exec_tmp_table1) { exec_tmp_table1->file->extra(HA_EXTRA_RESET_STATE); - exec_tmp_table1->file->delete_all_rows(); + exec_tmp_table1->file->ha_delete_all_rows(); free_io_cache(exec_tmp_table1); filesort_free_buffers(exec_tmp_table1,0); } if (exec_tmp_table2) { exec_tmp_table2->file->extra(HA_EXTRA_RESET_STATE); - exec_tmp_table2->file->delete_all_rows(); + exec_tmp_table2->file->ha_delete_all_rows(); free_io_cache(exec_tmp_table2); filesort_free_buffers(exec_tmp_table2,0); } @@ -5630,7 +5630,8 @@ get_store_key(THD *thd, KEYUSE *keyuse, table_map used_tables, (keyuse->val->type() == Item::REF_ITEM && ((Item_ref*)keyuse->val)->ref_type() == Item_ref::OUTER_REF && (*(Item_ref**)((Item_ref*)keyuse->val)->ref)->ref_type() == - Item_ref::DIRECT_REF) ) + Item_ref::DIRECT_REF && + keyuse->val->real_item()->type() == Item::FIELD_ITEM)) return new store_key_field(thd, key_part->field, key_buff + maybe_null, @@ -9353,6 +9354,8 @@ static Field *create_tmp_field_from_item(THD *thd, Item *item, TABLE *table, *((*copy_func)++) = item; // Save for copy_funcs if (modify_item) item->set_result_field(new_field); + if (item->type() == Item::NULL_ITEM) + new_field->is_created_from_null_item= TRUE; return new_field; } @@ -10724,7 +10727,7 @@ create_internal_tmp_table_from_heap2(THD *thd, TABLE *table, if (open_tmp_table(&new_table)) goto err1; if (table->file->indexes_are_disabled()) - new_table.file->disable_indexes(HA_KEY_SWITCH_ALL); + new_table.file->ha_disable_indexes(HA_KEY_SWITCH_ALL); table->file->ha_index_or_rnd_end(); table->file->ha_rnd_init(1); if (table->no_rows) @@ -10753,13 +10756,13 @@ create_internal_tmp_table_from_heap2(THD *thd, TABLE *table, */ while (!table->file->rnd_next(new_table.record[1])) { - write_err= new_table.file->write_row(new_table.record[1]); + write_err= new_table.file->ha_write_row(new_table.record[1]); DBUG_EXECUTE_IF("raise_error", write_err= HA_ERR_FOUND_DUPP_KEY ;); if (write_err) goto err; } /* copy row that filled HEAP table */ - if ((write_err=new_table.file->write_row(table->record[0]))) + if ((write_err=new_table.file->ha_write_row(table->record[0]))) { if (new_table.file->is_fatal_error(write_err, HA_CHECK_DUP) || !ignore_last_dupp_key_error) @@ -10790,7 +10793,7 @@ create_internal_tmp_table_from_heap2(THD *thd, TABLE *table, (void) table->file->ha_rnd_end(); (void) new_table.file->close(); err1: - new_table.file->delete_table(new_table.s->table_name.str); + new_table.file->ha_delete_table(new_table.s->table_name.str); err2: delete new_table.file; thd_proc_info(thd, save_proc_info); @@ -10813,9 +10816,9 @@ free_tmp_table(THD *thd, TABLE *entry) if (entry->file) { if (entry->db_stat) - entry->file->drop_table(entry->s->table_name.str); + entry->file->ha_drop_table(entry->s->table_name.str); else - entry->file->delete_table(entry->s->table_name.str); + entry->file->ha_delete_table(entry->s->table_name.str); delete entry->file; } @@ -12269,7 +12272,7 @@ end_write(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)), { int error; join->found_records++; - if ((error=table->file->write_row(table->record[0]))) + if ((error=table->file->ha_write_row(table->record[0]))) { if (!table->file->is_fatal_error(error, HA_CHECK_DUP)) goto end; @@ -12331,8 +12334,8 @@ end_update(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)), { /* Update old record */ restore_record(table,record[1]); update_tmptable_sum_func(join->sum_funcs,table); - if ((error=table->file->update_row(table->record[1], - table->record[0]))) + if ((error=table->file->ha_update_row(table->record[1], + table->record[0]))) { table->file->print_error(error,MYF(0)); /* purecov: inspected */ DBUG_RETURN(NESTED_LOOP_ERROR); /* purecov: inspected */ @@ -12355,7 +12358,7 @@ end_update(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)), } init_tmptable_sum_functions(join->sum_funcs); copy_funcs(join->tmp_table_param.items_to_copy); - if ((error=table->file->write_row(table->record[0]))) + if ((error=table->file->ha_write_row(table->record[0]))) { if (create_internal_tmp_table_from_heap(join->thd, table, &join->tmp_table_param, error, 0)) @@ -12391,7 +12394,7 @@ end_unique_update(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)), copy_fields(&join->tmp_table_param); // Groups are copied twice. copy_funcs(join->tmp_table_param.items_to_copy); - if (!(error=table->file->write_row(table->record[0]))) + if (!(error=table->file->ha_write_row(table->record[0]))) join->send_records++; // New group else { @@ -12407,8 +12410,8 @@ end_unique_update(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)), } restore_record(table,record[1]); update_tmptable_sum_func(join->sum_funcs,table); - if ((error=table->file->update_row(table->record[1], - table->record[0]))) + if ((error=table->file->ha_update_row(table->record[1], + table->record[0]))) { table->file->print_error(error,MYF(0)); /* purecov: inspected */ DBUG_RETURN(NESTED_LOOP_ERROR); /* purecov: inspected */ @@ -12451,7 +12454,7 @@ end_write_group(JOIN *join, JOIN_TAB *join_tab __attribute__((unused)), join->sum_funcs_end[send_group_parts]); if (!join->having || join->having->val_int()) { - int error= table->file->write_row(table->record[0]); + int error= table->file->ha_write_row(table->record[0]); if (error && create_internal_tmp_table_from_heap(join->thd, table, &join->tmp_table_param, error, 0)) @@ -13366,7 +13369,8 @@ check_reverse_order: select->quick=tmp; } } - else if (tab->ref.key >= 0 && tab->ref.key_parts <= used_key_parts) + else if (tab->type != JT_NEXT && + tab->ref.key >= 0 && tab->ref.key_parts <= used_key_parts) { /* SELECT * FROM t1 WHERE a=1 ORDER BY a DESC,b DESC @@ -13679,7 +13683,7 @@ static int remove_dup_with_compare(THD *thd, TABLE *table, Field **first_field, } if (having && !having->val_int()) { - if ((error=file->delete_row(record))) + if ((error=file->ha_delete_row(record))) goto err; error=file->rnd_next(record); continue; @@ -13706,7 +13710,7 @@ static int remove_dup_with_compare(THD *thd, TABLE *table, Field **first_field, } if (compare_record(table, first_field) == 0) { - if ((error=file->delete_row(record))) + if ((error=file->ha_delete_row(record))) goto err; } else if (!found) @@ -13806,7 +13810,7 @@ static int remove_dup_with_hash_index(THD *thd, TABLE *table, } if (having && !having->val_int()) { - if ((error=file->delete_row(record))) + if ((error=file->ha_delete_row(record))) goto err; continue; } @@ -13823,7 +13827,7 @@ static int remove_dup_with_hash_index(THD *thd, TABLE *table, if (hash_search(&hash, org_key_pos, key_length)) { /* Duplicated found ; Remove the row */ - if ((error=file->delete_row(record))) + if ((error=file->ha_delete_row(record))) goto err; } else @@ -15847,7 +15851,7 @@ int JOIN::rollup_write_data(uint idx, TABLE *table_arg) item->save_in_result_field(1); } copy_sum_funcs(sum_funcs_end[i+1], sum_funcs_end[i]); - if ((write_error= table_arg->file->write_row(table_arg->record[0]))) + if ((write_error= table_arg->file->ha_write_row(table_arg->record[0]))) { if (create_internal_tmp_table_from_heap(thd, table_arg, &tmp_table_param, write_error, 0)) diff --git a/sql/sql_show.cc b/sql/sql_show.cc index 8ba99e57ff1..4643aff95c4 100644 --- a/sql/sql_show.cc +++ b/sql/sql_show.cc @@ -5891,7 +5891,7 @@ bool get_schema_tables_result(JOIN *join, { table_list->table->file->extra(HA_EXTRA_NO_CACHE); table_list->table->file->extra(HA_EXTRA_RESET_STATE); - table_list->table->file->delete_all_rows(); + table_list->table->file->ha_delete_all_rows(); free_io_cache(table_list->table); filesort_free_buffers(table_list->table,1); table_list->table->null_row= 0; diff --git a/sql/sql_string.cc b/sql/sql_string.cc index 2e076af45eb..7fa3786c382 100644 --- a/sql/sql_string.cc +++ b/sql/sql_string.cc @@ -297,8 +297,8 @@ bool String::copy_aligned(const char *str,uint32 arg_length, uint32 offset, return TRUE; /* - Note, this is only safe for little-endian UCS-2. - If we add big-endian UCS-2 sometimes, this code + Note, this is only safe for big-endian UCS-2. + If we add little-endian UCS-2 sometimes, this code will be more complicated. But it's OK for now. */ bzero((char*) Ptr, offset); diff --git a/sql/sql_table.cc b/sql/sql_table.cc index c4897efd447..2438d8fc2ea 100644 --- a/sql/sql_table.cc +++ b/sql/sql_table.cc @@ -630,7 +630,7 @@ static int execute_ddl_log_action(THD *thd, DDL_LOG_ENTRY *ddl_log_entry) } else { - if ((error= file->delete_table(ddl_log_entry->name))) + if ((error= file->ha_delete_table(ddl_log_entry->name))) { if (error != ENOENT && error != HA_ERR_NO_SUCH_TABLE) break; @@ -667,8 +667,8 @@ static int execute_ddl_log_action(THD *thd, DDL_LOG_ENTRY *ddl_log_entry) } else { - if (file->rename_table(ddl_log_entry->from_name, - ddl_log_entry->name)) + if (file->ha_rename_table(ddl_log_entry->from_name, + ddl_log_entry->name)) break; } if ((deactivate_ddl_log_entry(ddl_log_entry->entry_pos))) @@ -1299,9 +1299,9 @@ bool mysql_write_frm(ALTER_PARTITION_PARAM_TYPE *lpt, uint flags) lpt->table_name, lpt->create_info, lpt->alter_info->create_list, lpt->key_count, lpt->key_info_buffer, lpt->table->file)) || - lpt->table->file->create_handler_files(shadow_path, NULL, - CHF_CREATE_FLAG, - lpt->create_info)) + lpt->table->file->ha_create_handler_files(shadow_path, NULL, + CHF_CREATE_FLAG, + lpt->create_info)) { my_delete(shadow_frm_name, MYF(0)); error= 1; @@ -1353,15 +1353,15 @@ bool mysql_write_frm(ALTER_PARTITION_PARAM_TYPE *lpt, uint flags) VOID(pthread_mutex_lock(&LOCK_open)); if (my_delete(frm_name, MYF(MY_WME)) || #ifdef WITH_PARTITION_STORAGE_ENGINE - lpt->table->file->create_handler_files(path, shadow_path, - CHF_DELETE_FLAG, NULL) || + lpt->table->file->ha_create_handler_files(path, shadow_path, + CHF_DELETE_FLAG, NULL) || deactivate_ddl_log_entry(part_info->frm_log_entry->entry_pos) || (sync_ddl_log(), FALSE) || #endif #ifdef WITH_PARTITION_STORAGE_ENGINE my_rename(shadow_frm_name, frm_name, MYF(MY_WME)) || - lpt->table->file->create_handler_files(path, shadow_path, - CHF_RENAME_FLAG, NULL)) + lpt->table->file->ha_create_handler_files(path, shadow_path, + CHF_RENAME_FLAG, NULL)) #else my_rename(shadow_frm_name, frm_name, MYF(MY_WME))) #endif @@ -3719,14 +3719,14 @@ mysql_rename_table(handlerton *base, const char *old_db, to_base= lc_to; } - if (!file || !(error=file->rename_table(from_base, to_base))) + if (!file || !(error=file->ha_rename_table(from_base, to_base))) { if (!(flags & NO_FRM_RENAME) && rename_file_ext(from,to,reg_ext)) { error=my_errno; /* Restore old file name */ if (file) - file->rename_table(to_base, from_base); + file->ha_rename_table(to_base, from_base); } } delete file; @@ -4379,7 +4379,7 @@ send_result_message: if (!result_code) // recreation went ok { if ((table->table= open_ltable(thd, table, lock_type, 0)) && - ((result_code= table->table->file->analyze(thd, check_opt)) > 0)) + ((result_code= table->table->file->ha_analyze(thd, check_opt)) > 0)) result_code= 0; // analyze went ok } if (result_code) // either mysql_recreate_table or analyze failed @@ -4489,7 +4489,7 @@ bool mysql_backup_table(THD* thd, TABLE_LIST* table_list) "MySQL Administrator (mysqldump, mysql)"); DBUG_RETURN(mysql_admin_table(thd, table_list, 0, "backup", TL_READ, 0, 0, 0, 0, - &handler::backup, 0)); + &handler::ha_backup, 0)); } @@ -4501,7 +4501,7 @@ bool mysql_restore_table(THD* thd, TABLE_LIST* table_list) DBUG_RETURN(mysql_admin_table(thd, table_list, 0, "restore", TL_WRITE, 1, 1, 0, &prepare_for_restore, - &handler::restore, 0)); + &handler::ha_restore, 0)); } @@ -4522,7 +4522,7 @@ bool mysql_optimize_table(THD* thd, TABLE_LIST* tables, HA_CHECK_OPT* check_opt) DBUG_ENTER("mysql_optimize_table"); DBUG_RETURN(mysql_admin_table(thd, tables, check_opt, "optimize", TL_WRITE, 1,0,0,0, - &handler::optimize, 0)); + &handler::ha_optimize, 0)); } @@ -4936,7 +4936,7 @@ bool mysql_analyze_table(THD* thd, TABLE_LIST* tables, HA_CHECK_OPT* check_opt) DBUG_ENTER("mysql_analyze_table"); DBUG_RETURN(mysql_admin_table(thd, tables, check_opt, "analyze", lock_type, 1, 0, 0, 0, - &handler::analyze, 0)); + &handler::ha_analyze, 0)); } @@ -4983,7 +4983,7 @@ mysql_discard_or_import_tablespace(THD *thd, DBUG_RETURN(-1); } - error=table->file->discard_or_import_tablespace(discard); + error= table->file->ha_discard_or_import_tablespace(discard); thd_proc_info(thd, "end"); @@ -5355,14 +5355,14 @@ bool alter_table_manage_keys(TABLE *table, int indexes_were_disabled, switch (keys_onoff) { case ENABLE: - error= table->file->enable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); + error= table->file->ha_enable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); break; case LEAVE_AS_IS: if (!indexes_were_disabled) break; /* fall-through: disabled indexes */ case DISABLE: - error= table->file->disable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); + error= table->file->ha_disable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); } if (error == HA_ERR_WRONG_COMMAND) @@ -6132,14 +6132,14 @@ view_err: wait_while_table_is_used(thd, table, HA_EXTRA_FORCE_REOPEN); VOID(pthread_mutex_unlock(&LOCK_open)); DBUG_EXECUTE_IF("sleep_alter_enable_indexes", my_sleep(6000000);); - error= table->file->enable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); + error= table->file->ha_enable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); /* COND_refresh will be signaled in close_thread_tables() */ break; case DISABLE: VOID(pthread_mutex_lock(&LOCK_open)); wait_while_table_is_used(thd, table, HA_EXTRA_FORCE_REOPEN); VOID(pthread_mutex_unlock(&LOCK_open)); - error=table->file->disable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); + error=table->file->ha_disable_indexes(HA_KEY_SWITCH_NONUNIQ_SAVE); /* COND_refresh will be signaled in close_thread_tables() */ break; default: @@ -6547,7 +6547,7 @@ view_err: KEY_PART_INFO *part_end; DBUG_PRINT("info", ("No new_table, checking add/drop index")); - table->file->prepare_for_alter(); + table->file->ha_prepare_for_alter(); if (index_add_count) { /* The add_index() method takes an array of KEY structs. */ @@ -6765,8 +6765,8 @@ view_err: t_table= table; } /* Tell the handler that a new frm file is in place. */ - if (t_table->file->create_handler_files(path, NULL, CHF_INDEX_FLAG, - create_info)) + if (t_table->file->ha_create_handler_files(path, NULL, CHF_INDEX_FLAG, + create_info)) goto err_with_placeholders; if (thd->locked_tables && new_name == table_name && new_db == db) { @@ -7064,7 +7064,7 @@ copy_data_between_tables(TABLE *from,TABLE *to, copy_ptr->do_copy(copy_ptr); } prev_insert_id= to->file->next_insert_id; - error=to->file->write_row(to->record[0]); + error=to->file->ha_write_row(to->record[0]); to->auto_increment_field_not_null= FALSE; if (error) { diff --git a/sql/sql_union.cc b/sql/sql_union.cc index 90d7b4dfc60..cee1c9d6bd4 100644 --- a/sql/sql_union.cc +++ b/sql/sql_union.cc @@ -440,10 +440,10 @@ bool st_select_lex_unit::exec() { item->assigned(0); // We will reinit & rexecute unit item->reset(); - table->file->delete_all_rows(); + table->file->ha_delete_all_rows(); } /* re-enabling indexes for next subselect iteration */ - if (union_distinct && table->file->enable_indexes(HA_KEY_SWITCH_ALL)) + if (union_distinct && table->file->ha_enable_indexes(HA_KEY_SWITCH_ALL)) { DBUG_ASSERT(0); } @@ -485,7 +485,7 @@ bool st_select_lex_unit::exec() sl->join->exec(); if (sl == union_distinct) { - if (table->file->disable_indexes(HA_KEY_SWITCH_ALL)) + if (table->file->ha_disable_indexes(HA_KEY_SWITCH_ALL)) DBUG_RETURN(TRUE); table->no_keyread=1; } diff --git a/sql/sql_update.cc b/sql/sql_update.cc index 6d0d5933971..eb44a13826c 100644 --- a/sql/sql_update.cc +++ b/sql/sql_update.cc @@ -526,7 +526,9 @@ int mysql_update(THD *thd, init_read_record(&info,thd,table,select,0,1); updated= found= 0; - thd->count_cuted_fields= CHECK_FIELD_WARN; /* calc cuted fields */ + /* Generate an error when trying to set a NOT NULL field to NULL. */ + thd->count_cuted_fields= ignore ? CHECK_FIELD_WARN + : CHECK_FIELD_ERROR_FOR_NULL; thd->cuted_fields=0L; thd_proc_info(thd, "Updating"); @@ -623,9 +625,9 @@ int mysql_update(THD *thd, call then it should be included in the count of dup_key_found and error should be set to 0 (only if these errors are ignored). */ - error= table->file->bulk_update_row(table->record[1], - table->record[0], - &dup_key_found); + error= table->file->ha_bulk_update_row(table->record[1], + table->record[0], + &dup_key_found); limit+= dup_key_found; updated-= dup_key_found; } diff --git a/sql/sql_view.cc b/sql/sql_view.cc index 79973b85181..da301b37484 100644 --- a/sql/sql_view.cc +++ b/sql/sql_view.cc @@ -1465,6 +1465,8 @@ bool mysql_drop_view(THD *thd, TABLE_LIST *views, enum_drop_mode drop_mode) char *wrong_object_db= NULL, *wrong_object_name= NULL; bool error= FALSE; enum legacy_db_type not_used; + bool some_views_deleted= FALSE; + bool something_wrong= FALSE; DBUG_ENTER("mysql_drop_view"); VOID(pthread_mutex_lock(&LOCK_open)); @@ -1506,6 +1508,8 @@ bool mysql_drop_view(THD *thd, TABLE_LIST *views, enum_drop_mode drop_mode) if (my_delete(path, MYF(MY_WME))) error= TRUE; + some_views_deleted= TRUE; + /* For a view, there is only one table_share object which should never be used outside of LOCK_open @@ -1523,29 +1527,32 @@ bool mysql_drop_view(THD *thd, TABLE_LIST *views, enum_drop_mode drop_mode) sp_cache_invalidate(); } - if (error) - { - VOID(pthread_mutex_unlock(&LOCK_open)); - DBUG_RETURN(TRUE); - } if (wrong_object_name) { - VOID(pthread_mutex_unlock(&LOCK_open)); my_error(ER_WRONG_OBJECT, MYF(0), wrong_object_db, wrong_object_name, "VIEW"); - DBUG_RETURN(TRUE); } if (non_existant_views.length()) { - VOID(pthread_mutex_unlock(&LOCK_open)); my_error(ER_BAD_TABLE_ERROR, MYF(0), non_existant_views.c_ptr()); - DBUG_RETURN(TRUE); } - write_bin_log(thd, TRUE, thd->query, thd->query_length); + something_wrong= error || wrong_object_name || non_existant_views.length(); + if (some_views_deleted || !something_wrong) + { + /* if something goes wrong, bin-log with possible error code, + otherwise bin-log with error code cleared. + */ + write_bin_log(thd, !something_wrong, thd->query, thd->query_length); + } - send_ok(thd); VOID(pthread_mutex_unlock(&LOCK_open)); + + if (something_wrong) + { + DBUG_RETURN(TRUE); + } + send_ok(thd); DBUG_RETURN(FALSE); } diff --git a/sql/sql_yacc.yy b/sql/sql_yacc.yy index 42fc5b6cbe1..c00394d96cf 100644 --- a/sql/sql_yacc.yy +++ b/sql/sql_yacc.yy @@ -509,10 +509,10 @@ bool my_yyoverflow(short **a, YYSTYPE **b, ulong *yystacksize); %pure_parser /* We have threads */ /* - Currently there are 177 shift/reduce conflicts. + Currently there are 169 shift/reduce conflicts. We should not introduce new conflicts any more. */ -%expect 177 +%expect 169 /* Comments for TOKENS. @@ -1209,7 +1209,7 @@ bool my_yyoverflow(short **a, YYSTYPE **b, ulong *yystacksize); %type <table_list> join_table_list join_table - table_factor table_ref + table_factor table_ref esc_table_ref select_derived derived_table_list %type <date_time_type> date_time_type; @@ -1300,7 +1300,9 @@ END_OF_INPUT %type <NONE> call sp_proc_stmts sp_proc_stmts1 sp_proc_stmt %type <NONE> sp_proc_stmt_statement sp_proc_stmt_return %type <NONE> sp_proc_stmt_if -%type <NONE> sp_labeled_control sp_proc_stmt_unlabeled sp_proc_stmt_leave +%type <NONE> sp_labeled_control sp_proc_stmt_unlabeled +%type <NONE> sp_labeled_block sp_unlabeled_block +%type <NONE> sp_proc_stmt_leave %type <NONE> sp_proc_stmt_iterate %type <NONE> sp_proc_stmt_open sp_proc_stmt_fetch sp_proc_stmt_close %type <NONE> case_stmt_specification simple_case_stmt searched_case_stmt @@ -1961,6 +1963,8 @@ ev_sql_stmt_inner: | sp_proc_stmt_return | sp_proc_stmt_if | case_stmt_specification + | sp_labeled_block + | sp_unlabeled_block | sp_labeled_control | sp_proc_stmt_unlabeled | sp_proc_stmt_leave @@ -2535,6 +2539,8 @@ sp_proc_stmt: | sp_proc_stmt_return | sp_proc_stmt_if | case_stmt_specification + | sp_labeled_block + | sp_unlabeled_block | sp_labeled_control | sp_proc_stmt_unlabeled | sp_proc_stmt_leave @@ -2661,14 +2667,35 @@ sp_proc_stmt_leave: sp_instr_jump *i; uint ip= sp->instructions(); uint n; + /* + When jumping to a BEGIN-END block end, the target jump + points to the block hpop/cpop cleanup instructions, + so we should exclude the block context here. + When jumping to something else (i.e., SP_LAB_ITER), + there are no hpop/cpop at the jump destination, + so we should include the block context here for cleanup. + */ + bool exclusive= (lab->type == SP_LAB_BEGIN); - n= ctx->diff_handlers(lab->ctx, TRUE); /* Exclusive the dest. */ + n= ctx->diff_handlers(lab->ctx, exclusive); if (n) - sp->add_instr(new sp_instr_hpop(ip++, ctx, n)); - n= ctx->diff_cursors(lab->ctx, TRUE); /* Exclusive the dest. */ + { + sp_instr_hpop *hpop= new sp_instr_hpop(ip++, ctx, n); + if (hpop == NULL) + MYSQL_YYABORT; + sp->add_instr(hpop); + } + n= ctx->diff_cursors(lab->ctx, exclusive); if (n) - sp->add_instr(new sp_instr_cpop(ip++, ctx, n)); + { + sp_instr_cpop *cpop= new sp_instr_cpop(ip++, ctx, n); + if (cpop == NULL) + MYSQL_YYABORT; + sp->add_instr(cpop); + } i= new sp_instr_jump(ip, ctx); + if (i == NULL) + MYSQL_YYABORT; sp->push_backpatch(i, lab); /* Jumping forward */ sp->add_instr(i); } @@ -2696,10 +2723,20 @@ sp_proc_stmt_iterate: n= ctx->diff_handlers(lab->ctx, FALSE); /* Inclusive the dest. */ if (n) - sp->add_instr(new sp_instr_hpop(ip++, ctx, n)); + { + sp_instr_hpop *hpop= new sp_instr_hpop(ip++, ctx, n); + if (hpop == NULL) + MYSQL_YYABORT; + sp->add_instr(hpop); + } n= ctx->diff_cursors(lab->ctx, FALSE); /* Inclusive the dest. */ if (n) - sp->add_instr(new sp_instr_cpop(ip++, ctx, n)); + { + sp_instr_cpop *cpop= new sp_instr_cpop(ip++, ctx, n); + if (cpop == NULL) + MYSQL_YYABORT; + sp->add_instr(cpop); + } i= new sp_instr_jump(ip, ctx, lab->ip); /* Jump back */ sp->add_instr(i); } @@ -2983,19 +3020,17 @@ sp_labeled_control: sp_unlabeled_control sp_opt_label { LEX *lex= Lex; + sp_label_t *lab= lex->spcont->pop_label(); if ($5.str) { - sp_label_t *lab= lex->spcont->find_label($5.str); - - if (!lab || - my_strcasecmp(system_charset_info, $5.str, lab->name) != 0) + if (my_strcasecmp(system_charset_info, $5.str, lab->name) != 0) { my_error(ER_SP_LABEL_MISMATCH, MYF(0), $5.str); MYSQL_YYABORT; } } - lex->sphead->backpatch(lex->spcont->pop_label()); + lex->sphead->backpatch(lab); } ; @@ -3004,15 +3039,59 @@ sp_opt_label: | label_ident { $$= $1; } ; -sp_unlabeled_control: +sp_labeled_block: + label_ident ':' + { + LEX *lex= Lex; + sp_pcontext *ctx= lex->spcont; + sp_label_t *lab= ctx->find_label($1.str); + + if (lab) + { + my_error(ER_SP_LABEL_REDEFINE, MYF(0), $1.str); + MYSQL_YYABORT; + } + + lab= lex->spcont->push_label($1.str, + lex->sphead->instructions()); + lab->type= SP_LAB_BEGIN; + } + sp_block_content sp_opt_label + { + LEX *lex= Lex; + sp_label_t *lab= lex->spcont->pop_label(); + + if ($5.str) + { + if (my_strcasecmp(system_charset_info, $5.str, lab->name) != 0) + { + my_error(ER_SP_LABEL_MISMATCH, MYF(0), $5.str); + MYSQL_YYABORT; + } + } + } + ; + +sp_unlabeled_block: + { /* Unlabeled blocks get a secret label. */ + LEX *lex= Lex; + uint ip= lex->sphead->instructions(); + sp_label_t *lab= lex->spcont->push_label((char *)"", ip); + lab->type= SP_LAB_BEGIN; + } + sp_block_content + { + LEX *lex= Lex; + lex->spcont->pop_label(); + } + ; + +sp_block_content: BEGIN_SYM { /* QQ This is just a dummy for grouping declarations and statements together. No [[NOT] ATOMIC] yet, and we need to figure out how make it coexist with the existing BEGIN COMMIT/ROLLBACK. */ LEX *lex= Lex; - sp_label_t *lab= lex->spcont->last_label(); - - lab->type= SP_LAB_BEGIN; lex->spcont= lex->spcont->push_context(LABEL_DEFAULT_SCOPE); } sp_decls @@ -3032,7 +3111,10 @@ sp_unlabeled_control: $3.curs)); lex->spcont= ctx->pop_context(); } - | LOOP_SYM + ; + +sp_unlabeled_control: + LOOP_SYM sp_proc_stmts1 END LOOP_SYM { LEX *lex= Lex; @@ -6176,6 +6258,14 @@ select_paren: my_parse_error(ER(ER_SYNTAX_ERROR)); MYSQL_YYABORT; } + if (sel->linkage == UNION_TYPE && + sel->olap != UNSPECIFIED_OLAP_TYPE && + sel->master_unit()->fake_select_lex) + { + my_error(ER_WRONG_USAGE, MYF(0), + "CUBE/ROLLUP", "ORDER BY"); + MYSQL_YYABORT; + } /* select in braces, can't contain global parameters */ if (sel->master_unit()->fake_select_lex) sel->master_unit()->global_parameters= @@ -7469,10 +7559,22 @@ join_table_list: derived_table_list { MYSQL_YYABORT_UNLESS($$=$1); } ; +/* + The ODBC escape syntax for Outer Join is: '{' OJ join_table '}' + The parser does not define OJ as a token, any ident is accepted + instead in $2 (ident). Also, all productions from table_ref can + be escaped, not only join_table. Both syntax extensions are safe + and are ignored. +*/ +esc_table_ref: + table_ref { $$=$1; } + | '{' ident table_ref '}' { $$=$3; } + ; + /* Warning - may return NULL in case of incomplete SELECT */ derived_table_list: - table_ref { $$=$1; } - | derived_table_list ',' table_ref + esc_table_ref { $$=$1; } + | derived_table_list ',' esc_table_ref { MYSQL_YYABORT_UNLESS($1 && ($$=$3)); } @@ -7637,25 +7739,6 @@ table_factor: MYSQL_YYABORT; Select->add_joined_table($$); } - | '{' ident table_ref LEFT OUTER JOIN_SYM table_ref - ON - { - /* Change the current name resolution context to a local context. */ - if (push_new_name_resolution_context(YYTHD, $3, $7)) - MYSQL_YYABORT; - - } - expr '}' - { - LEX *lex= Lex; - MYSQL_YYABORT_UNLESS($3 && $7); - add_join_on($7,$10); - Lex->pop_context(); - $7->outer_join|=JOIN_TYPE_LEFT; - $$=$7; - if (!($$= lex->current_select->nest_last_join(lex->thd))) - MYSQL_YYABORT; - } | select_derived_init get_select_lex select_derived2 { LEX *lex= Lex; @@ -8064,7 +8147,8 @@ order_clause: SELECT_LEX *sel= lex->current_select; SELECT_LEX_UNIT *unit= sel-> master_unit(); if (sel->linkage != GLOBAL_OPTIONS_TYPE && - sel->olap != UNSPECIFIED_OLAP_TYPE) + sel->olap != UNSPECIFIED_OLAP_TYPE && + (sel->linkage != UNION_TYPE || sel->braces)) { my_error(ER_WRONG_USAGE, MYF(0), "CUBE/ROLLUP", "ORDER BY"); diff --git a/sql/unireg.cc b/sql/unireg.cc index e5f230841f6..64a5b7d8910 100644 --- a/sql/unireg.cc +++ b/sql/unireg.cc @@ -227,6 +227,14 @@ bool mysql_create_frm(THD *thd, const char *file_name, strmake((char*) forminfo+47, create_info->comment.str ? create_info->comment.str : "", create_info->comment.length); forminfo[46]=(uchar) create_info->comment.length; +#ifdef EXTRA_DEBUG + /* + EXTRA_DEBUG causes strmake() to initialize its buffer behind the + payload with a magic value to detect wrong buffer-sizes. We + explicitly zero that segment again. + */ + memset((char*) forminfo+47 + forminfo[46], 0, 61 - forminfo[46]); +#endif #ifdef WITH_PARTITION_STORAGE_ENGINE if (part_info) { @@ -395,7 +403,7 @@ int rea_create_table(THD *thd, const char *path, DBUG_ASSERT(*fn_rext(frm_name)); if (thd->variables.keep_files_on_create) create_info->options|= HA_CREATE_KEEP_FILES; - if (file->create_handler_files(path, NULL, CHF_CREATE_FLAG, create_info)) + if (file->ha_create_handler_files(path, NULL, CHF_CREATE_FLAG, create_info)) goto err_handler; if (!create_info->frm_only && ha_create_table(thd, path, db, table_name, create_info,0)) @@ -403,7 +411,7 @@ int rea_create_table(THD *thd, const char *path, DBUG_RETURN(0); err_handler: - VOID(file->create_handler_files(path, NULL, CHF_DELETE_FLAG, create_info)); + VOID(file->ha_create_handler_files(path, NULL, CHF_DELETE_FLAG, create_info)); my_delete(frm_name, MYF(0)); DBUG_RETURN(1); } /* rea_create_table */ |