applied percona patches to InnoDB as of 5.5.15

btr/btr0cur.c

@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1994, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2011, Oracle and/or its affiliates. All Rights Reserved.
 Copyright (c) 2008, Google Inc.
 
 Portions of this file contain modifications contributed and copyrighted by
@@ -2067,6 +2067,9 @@ btr_cur_optimistic_update(
 
 	heap = mem_heap_create(1024);
 	offsets = rec_get_offsets(rec, index, NULL, ULINT_UNDEFINED, &heap);
+#ifdef UNIV_BLOB_NULL_DEBUG
+	ut_a(!rec_offs_any_null_extern(rec, offsets));
+#endif /* UNIV_BLOB_NULL_DEBUG */
 
 #ifdef UNIV_DEBUG
 	if (btr_cur_print_record_ops && thr) {
@@ -3347,9 +3350,14 @@ btr_estimate_n_rows_in_range_on_level(
 
 		mtr_start(&mtr);
 
-		/* fetch the page */
-		block = buf_page_get(space, zip_size, page_no, RW_S_LATCH,
-				     &mtr);
+		/* Fetch the page. Because we are not holding the
+		index->lock, the tree may have changed and we may be
+		attempting to read a page that is no longer part of
+		the B-tree. We pass BUF_GET_POSSIBLY_FREED in order to
+		silence a debug assertion about this. */
+		block = buf_page_get_gen(space, zip_size, page_no, RW_S_LATCH,
+					 NULL, BUF_GET_POSSIBLY_FREED,
+					 __FILE__, __LINE__, &mtr);
 
 		page = buf_block_get_frame(block);
 
@@ -3368,6 +3376,13 @@ btr_estimate_n_rows_in_range_on_level(
 			goto inexact;
 		}
 
+		/* It is possible but highly unlikely that the page was
+		originally written by an old version of InnoDB that did
+		not initialize FIL_PAGE_TYPE on other than B-tree pages.
+		For example, this could be an almost-empty BLOB page
+		that happens to contain the magic values in the fields
+		that we checked above. */
+
 		n_pages_read++;
 
 		if (page_no != slot1->page_no) {

buf/buf0buf.c

@@ -2984,6 +2984,7 @@ buf_page_get_gen(
 	case BUF_GET_IF_IN_POOL:
 	case BUF_PEEK_IF_IN_POOL:
 	case BUF_GET_IF_IN_POOL_OR_WATCH:
+	case BUF_GET_POSSIBLY_FREED:
 		break;
 	default:
 		ut_error;
@@ -3359,7 +3360,10 @@ wait_until_unfixed:
 #endif /* UNIV_DEBUG || UNIV_IBUF_DEBUG */
 
 	buf_block_buf_fix_inc(block, file, line);
-
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
+	ut_a(mode == BUF_GET_POSSIBLY_FREED
+	     || !block->page.file_page_was_freed);
+#endif
 	//mutex_exit(&block->mutex);
 
 	/* Check if this is the first access to the page */
@@ -3373,10 +3377,6 @@ wait_until_unfixed:
 		buf_page_set_accessed_make_young(&block->page, access_time);
 	}
 
-#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
-	ut_a(!block->page.file_page_was_freed);
-#endif
-
 #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG
 	ut_a(++buf_dbg_counter % 5771 || buf_validate());
 	ut_a(block->page.buf_fix_count > 0);

buf/buf0flu.c

@@ -867,7 +867,7 @@ corrupted_page:
 flush:
 	/* Now flush the doublewrite buffer data to disk */
 
-	fil_flush(srv_doublewrite_file ? TRX_DOUBLEWRITE_SPACE : TRX_SYS_SPACE);
+	fil_flush(srv_doublewrite_file ? TRX_DOUBLEWRITE_SPACE : TRX_SYS_SPACE, FALSE);
 
 	/* We know that the writes have been flushed to disk now
 	and in recovery we will find them in the doublewrite buffer

data/data0data.c

@@ -585,7 +585,8 @@ dtuple_convert_big_rec(
 
 	if (dict_table_get_format(index->table) < DICT_TF_FORMAT_ZIP) {
 		/* up to MySQL 5.1: store a 768-byte prefix locally */
-		local_len = BTR_EXTERN_FIELD_REF_SIZE + DICT_MAX_INDEX_COL_LEN;
+		local_len = BTR_EXTERN_FIELD_REF_SIZE
+			+ DICT_ANTELOPE_MAX_INDEX_COL_LEN;
 	} else {
 		/* new-format table: do not store any BLOB prefix locally */
 		local_len = BTR_EXTERN_FIELD_REF_SIZE;
@@ -757,7 +758,10 @@ dtuple_convert_back_big_rec(
 
 		local_len -= BTR_EXTERN_FIELD_REF_SIZE;
 
-		ut_ad(local_len <= DICT_MAX_INDEX_COL_LEN);
+		/* Only in REDUNDANT and COMPACT format, we store
+		up to DICT_ANTELOPE_MAX_INDEX_COL_LEN (768) bytes
+		locally */
+		ut_ad(local_len <= DICT_ANTELOPE_MAX_INDEX_COL_LEN);
 
 		dfield_set_data(dfield,
 				(char*) b->data - local_len,

dict/dict0crea.c

@@ -730,9 +730,9 @@ dict_create_index_tree_step(
 	/* printf("Created a new index tree in space %lu root page %lu\n",
 	index->space, index->page_no); */
 
-	page_rec_write_index_page_no(btr_pcur_get_rec(&pcur),
-				     DICT_SYS_INDEXES_PAGE_NO_FIELD,
-				     node->page_no, &mtr);
+	page_rec_write_field(btr_pcur_get_rec(&pcur),
+			     DICT_SYS_INDEXES_PAGE_NO_FIELD,
+			     node->page_no, &mtr);
 	btr_pcur_close(&pcur);
 	mtr_commit(&mtr);
 
@@ -802,9 +802,8 @@ dict_drop_index_tree(
 	root_page_no); */
 	btr_free_root(space, zip_size, root_page_no, mtr);
 
-	page_rec_write_index_page_no(rec,
-				     DICT_SYS_INDEXES_PAGE_NO_FIELD,
-				     FIL_NULL, mtr);
+	page_rec_write_field(rec, DICT_SYS_INDEXES_PAGE_NO_FIELD,
+			     FIL_NULL, mtr);
 }
 
 /*******************************************************************//**
@@ -907,8 +906,8 @@ create:
 	in SYS_INDEXES, so that the database will not get into an
 	inconsistent state in case it crashes between the mtr_commit()
 	below and the following mtr_commit() call. */
-	page_rec_write_index_page_no(rec, DICT_SYS_INDEXES_PAGE_NO_FIELD,
-				     FIL_NULL, mtr);
+	page_rec_write_field(rec, DICT_SYS_INDEXES_PAGE_NO_FIELD,
+			     FIL_NULL, mtr);
 
 	/* We will need to commit the mini-transaction in order to avoid
 	deadlocks in the btr_create() call, because otherwise we would

dict/dict0dict.c

@@ -1415,36 +1415,63 @@ dict_index_too_big_for_undo(
 		ulint			fixed_size
 			= dict_col_get_fixed_size(col,
 						  dict_table_is_comp(table));
+		ulint			max_prefix
+			= col->max_prefix;
 
 		if (fixed_size) {
 			/* Fixed-size columns are stored locally. */
 			max_size = fixed_size;
 		} else if (max_size <= BTR_EXTERN_FIELD_REF_SIZE * 2) {
 			/* Short columns are stored locally. */
-		} else if (!col->ord_part) {
+		} else if (!col->ord_part
+			   || (col->max_prefix
+			       < (ulint) DICT_MAX_FIELD_LEN_BY_FORMAT(table))) {
 			/* See if col->ord_part would be set
-			because of new_index. */
+			because of new_index. Also check if the new
+			index could have longer prefix on columns
+			that already had ord_part set  */
 			ulint	j;
 
 			for (j = 0; j < new_index->n_uniq; j++) {
 				if (dict_index_get_nth_col(
 					    new_index, j) == col) {
+					const dict_field_t*     field
+						= dict_index_get_nth_field(
+							new_index, j);
+
+					if (field->prefix_len
+					    > col->max_prefix) {
+						max_prefix =
+							 field->prefix_len;
+					}
 
 					goto is_ord_part;
 				}
 			}
 
+			if (col->ord_part) {
+				goto is_ord_part;
+			}
+
 			/* This is not an ordering column in any index.
 			Thus, it can be stored completely externally. */
 			max_size = BTR_EXTERN_FIELD_REF_SIZE;
 		} else {
+			ulint	max_field_len;
 is_ord_part:
+			max_field_len = DICT_MAX_FIELD_LEN_BY_FORMAT(table);
+
 			/* This is an ordering column in some index.
 			A long enough prefix must be written to the
 			undo log.  See trx_undo_page_fetch_ext(). */
+			max_size = ut_min(max_size, max_field_len);
+
+			/* We only store the needed prefix length in undo log */
+			if (max_prefix) {
+			     ut_ad(dict_table_get_format(table)
+				   >= DICT_TF_FORMAT_ZIP);
 
-			if (max_size > REC_MAX_INDEX_COL_LEN) {
-				max_size = REC_MAX_INDEX_COL_LEN;
+				max_size = ut_min(max_prefix, max_size);
 			}
 
 			max_size += BTR_EXTERN_FIELD_REF_SIZE;
@@ -1698,15 +1725,16 @@ too_big:
 		/* In dtuple_convert_big_rec(), variable-length columns
 		that are longer than BTR_EXTERN_FIELD_REF_SIZE * 2
 		may be chosen for external storage.  If the column appears
-		in an ordering column of an index, a longer prefix of
-		REC_MAX_INDEX_COL_LEN will be copied to the undo log
-		by trx_undo_page_report_modify() and
+		in an ordering column of an index, a longer prefix determined
+		by dict_max_field_len_store_undo() will be copied to the undo
+		log by trx_undo_page_report_modify() and
 		trx_undo_page_fetch_ext().  It suffices to check the
 		capacity of the undo log whenever new_index includes
 		a column prefix on a column that may be stored externally. */
 
 		if (field->prefix_len /* prefix index */
-		    && !col->ord_part /* not yet ordering column */
+		    && (!col->ord_part /* not yet ordering column */
+			|| field->prefix_len > col->max_prefix)
 		    && !dict_col_get_fixed_size(col, TRUE) /* variable-length */
 		    && dict_col_get_max_size(col)
 		    > BTR_EXTERN_FIELD_REF_SIZE * 2 /* long enough */) {
@@ -1723,11 +1751,17 @@ too_big:
 	}
 
 undo_size_ok:
-	/* Flag the ordering columns */
+	/* Flag the ordering columns and also set column max_prefix */
 
 	for (i = 0; i < n_ord; i++) {
+		const dict_field_t*	field
+			= dict_index_get_nth_field(new_index, i);
 
-		dict_index_get_nth_field(new_index, i)->col->ord_part = 1;
+		field->col->ord_part = 1;
+
+		if (field->prefix_len > field->col->max_prefix) {
+			field->col->max_prefix = field->prefix_len;
+		}
 	}
 
 	/* Add the new index as the last index for the table */
@@ -1935,14 +1969,14 @@ dict_index_add_col(
 	variable-length fields, so that the extern flag can be embedded in
 	the length word. */
 
-	if (field->fixed_len > DICT_MAX_INDEX_COL_LEN) {
+	if (field->fixed_len > DICT_MAX_FIXED_COL_LEN) {
 		field->fixed_len = 0;
 	}
-#if DICT_MAX_INDEX_COL_LEN != 768
+#if DICT_MAX_FIXED_COL_LEN != 768
 	/* The comparison limit above must be constant.  If it were
 	changed, the disk format of some fixed-length columns would
 	change, which would be a disaster. */
-# error "DICT_MAX_INDEX_COL_LEN != 768"
+# error "DICT_MAX_FIXED_COL_LEN != 768"
 #endif
 
 	if (!(col->prtype & DATA_NOT_NULL)) {

dict/dict0load.c

@@ -432,7 +432,7 @@ dict_process_sys_fields_rec(
 	mach_write_to_8(last_index_id, last_id);
 
 	err_msg = dict_load_field_low(buf, NULL, sys_field,
-				      pos, last_index_id, heap, rec);
+				      pos, last_index_id, heap, rec, NULL, 0);
 
 	*index_id = mach_read_from_8(buf);
 
@@ -1066,6 +1066,9 @@ dict_load_columns(
 /** Error message for a delete-marked record in dict_load_field_low() */
 static const char* dict_load_field_del = "delete-marked record in SYS_FIELDS";
 
+static const char* dict_load_field_too_big = "column prefix exceeds maximum"
+					     " limit";
+
 /********************************************************************//**
 Loads an index field definition from a SYS_FIELDS record to
 dict_index_t.
@@ -1087,7 +1090,12 @@ dict_load_field_low(
 	byte*		last_index_id,	/*!< in: last index id */
 	mem_heap_t*	heap,		/*!< in/out: memory heap
 					for temporary storage */
-	const rec_t*	rec)		/*!< in: SYS_FIELDS record */
+	const rec_t*	rec,		/*!< in: SYS_FIELDS record */
+	char*		addition_err_str,/*!< out: additional error message
+					that requires information to be
+					filled, or NULL */
+	ulint		err_str_len)	/*!< in: length of addition_err_str
+					in bytes */
 {
 	const byte*	field;
 	ulint		len;
@@ -1167,6 +1175,19 @@ err_len:
 		goto err_len;
 	}
 
+	if (prefix_len > REC_VERSION_56_MAX_INDEX_COL_LEN) {
+		if (addition_err_str) {
+			ut_snprintf(addition_err_str, err_str_len,
+				    "index field '%s' has a prefix length"
+				    " of %lu bytes",
+				    mem_heap_strdupl(
+						heap, (const char*) field, len),
+				    (ulong) prefix_len);
+		}
+
+		return(dict_load_field_too_big);
+	}
+
 	if (index) {
 		dict_mem_index_add_field(
 			index, mem_heap_strdupl(heap, (const char*) field, len),
@@ -1226,14 +1247,16 @@ dict_load_fields(
 	btr_pcur_open_on_user_rec(sys_index, tuple, PAGE_CUR_GE,
 				  BTR_SEARCH_LEAF, &pcur, &mtr);
 	for (i = 0; i < index->n_fields; i++) {
-		const char* err_msg;
+		const char*	err_msg;
+		char		addition_err_str[1024];
 
 		rec = btr_pcur_get_rec(&pcur);
 
 		ut_a(btr_pcur_is_on_user_rec(&pcur));
 
 		err_msg = dict_load_field_low(buf, index, NULL, NULL, NULL,
-					      heap, rec);
+					      heap, rec, addition_err_str,
+					      sizeof(addition_err_str));
 
 		if (err_msg == dict_load_field_del) {
 			/* There could be delete marked records in
@@ -1242,7 +1265,24 @@ dict_load_fields(
 
 			goto next_rec;
 		} else if (err_msg) {
-			fprintf(stderr, "InnoDB: %s\n", err_msg);
+			if (err_msg == dict_load_field_too_big) {
+				fprintf(stderr, "InnoDB: Error: load index"
+					" '%s' failed.\n"
+					"InnoDB: %s,\n"
+					"InnoDB: which exceeds the"
+					" maximum limit of %lu bytes.\n"
+					"InnoDB: Please use server that"
+					" supports long index prefix\n"
+					"InnoDB: or turn on"
+					" innodb_force_recovery to load"
+					" the table\n",
+					index->name, addition_err_str,
+					(ulong) (REC_VERSION_56_MAX_INDEX_COL_LEN));
+
+			} else {
+				fprintf(stderr, "InnoDB: %s\n", err_msg);
+			}
+
 			error = DB_CORRUPTION;
 			goto func_exit;
 		}
@@ -1518,7 +1558,26 @@ corrupted:
 			of the database server */
 			dict_mem_index_free(index);
 		} else {
-			dict_load_fields(index, heap);
+			error = dict_load_fields(index, heap);
+
+			if (error != DB_SUCCESS) {
+
+				fprintf(stderr, "InnoDB: Error: load index '%s'"
+					" for table '%s' failed\n",
+					index->name, table->name);
+
+				/* If the force recovery flag is set, and
+				if the failed index is not the primary index, we
+				will continue and open other indexes */
+				if (srv_force_recovery
+				    && !dict_index_is_clust(index)) {
+					error = DB_SUCCESS;
+					goto next_rec;
+				} else {
+					goto func_exit;
+				}
+			}
+
 			error = dict_index_add_to_cache(table, index,
 							index->page, FALSE);
 			/* The data dictionary tables should never contain
@@ -1843,9 +1902,18 @@ err_exit:
 		} else {
 			table->fk_max_recusive_level = 0;
 		}
-	} else if (!srv_force_recovery) {
-		dict_table_remove_from_cache(table);
-		table = NULL;
+	} else {
+		dict_index_t*	index;
+
+		/* Make sure that at least the clustered index was loaded.
+		Otherwise refuse to load the table */
+		index = dict_table_get_first_index(table);
+
+		if (!srv_force_recovery || !index
+		     || !dict_index_is_clust(index)) {
+			dict_table_remove_from_cache(table);
+			table = NULL;
+		}
 	}
 #if 0
 	if (err != DB_SUCCESS && table != NULL) {

dict/dict0mem.c

@@ -234,6 +234,7 @@ dict_mem_fill_column_struct(
 
 	column->ind = (unsigned int) col_pos;
 	column->ord_part = 0;
+	column->max_prefix = 0;
 	column->mtype = (unsigned int) mtype;
 	column->prtype = (unsigned int) prtype;
 	column->len = (unsigned int) col_len;

fil/fil0fil.c

@@ -866,7 +866,8 @@ fil_node_close_file(
 	ut_a(node->open);
 	ut_a(node->n_pending == 0 || node->space->is_being_deleted);
 	ut_a(node->n_pending_flushes == 0);
-	ut_a(node->modification_counter == node->flush_counter);
+	ut_a(node->modification_counter == node->flush_counter
+	     || srv_fast_shutdown == 2);
 
 	ret = os_file_close(node->handle);
 	ut_a(ret);
@@ -2628,7 +2629,7 @@ retry:
 
 		os_thread_sleep(20000);
 
-		fil_flush(id);
+		fil_flush(id, TRUE);
 
 		goto retry;
 
@@ -2842,7 +2843,7 @@ error_exit2:
 		goto error_exit;
 	}
 
-	ret = os_file_flush(file);
+	ret = os_file_flush(file, TRUE);
 
 	if (!ret) {
 		fputs("InnoDB: Error: file flush of tablespace ", stderr);
@@ -3028,7 +3029,7 @@ fil_reset_too_high_lsns(
 		}
 	}
 
-	success = os_file_flush(file);
+	success = os_file_flush(file, TRUE);
 	if (!success) {
 
 		goto func_exit;
@@ -3050,7 +3051,7 @@ fil_reset_too_high_lsns(
 
 		goto func_exit;
 	}
-	success = os_file_flush(file);
+	success = os_file_flush(file, TRUE);
 func_exit:
 	os_file_close(file);
 	ut_free(buf2);
@@ -4838,7 +4839,7 @@ fil_extend_space_to_desired_size(
 	mutex_exit(&fil_system->mutex);
 	mutex_exit(&fil_system->file_extend_mutex);
 
-	fil_flush(space_id);
+	fil_flush(space_id, TRUE);
 
 	return(success);
 }
@@ -5550,8 +5551,9 @@ UNIV_INTERN
 void
 fil_flush(
 /*======*/
-	ulint	space_id)	/*!< in: file space id (this can be a group of
+	ulint	space_id,	/*!< in: file space id (this can be a group of
 				log files or a tablespace of the database) */
+	ibool	metadata)
 {
 	fil_space_t*	space;
 	fil_node_t*	node;
@@ -5622,7 +5624,7 @@ retry:
 			/* fprintf(stderr, "Flushing to file %s\n",
 			node->name); */
 
-			os_file_flush(file);
+			os_file_flush(file, metadata);
 
 			mutex_enter(&fil_system->mutex);
 
@@ -5705,7 +5707,7 @@ fil_flush_file_spaces(
 	a non-existing space id. */
 	for (i = 0; i < n_space_ids; i++) {
 
-		fil_flush(space_ids[i]);
+		fil_flush(space_ids[i], TRUE);
 	}
 
 	mem_free(space_ids);

handler/ha_innodb.cc

@@ -196,6 +196,7 @@ static my_bool	innobase_stats_on_metadata		= TRUE;
 static my_bool	innobase_use_sys_stats_table		= FALSE;
 static my_bool	innobase_buffer_pool_shm_checksum	= TRUE;
 static uint	innobase_buffer_pool_shm_key		= 0;
+static my_bool	innobase_large_prefix			= FALSE;
 
 
 static char*	internal_innobase_data_file_path	= NULL;
@@ -1056,7 +1057,7 @@ int
 convert_error_code_to_mysql(
 /*========================*/
 	int	error,	/*!< in: InnoDB error code */
-	ulint	flags,	/*!< in: InnoDB table flags, or 0 */
+	ulint	flags,  /*!< in: InnoDB table flags, or 0 */
 	THD*	thd)	/*!< in: user thread handle or NULL */
 {
 	switch (error) {
@@ -1160,6 +1161,11 @@ convert_error_code_to_mysql(
 						      & DICT_TF_COMPACT) / 2);
 		return(HA_ERR_TO_BIG_ROW);
 
+	case DB_TOO_BIG_INDEX_COL:
+		my_error(ER_INDEX_COLUMN_TOO_LONG, MYF(0),
+			 DICT_MAX_FIELD_LEN_BY_FORMAT_FLAG(flags));
+		return(HA_ERR_INDEX_COL_TOO_LONG);
+
 	case DB_NO_SAVEPOINT:
 		return(HA_ERR_NO_SAVEPOINT);
 
@@ -3040,8 +3046,10 @@ innobase_alter_table_flags(
 	uint	flags)
 {
 	return(HA_INPLACE_ADD_INDEX_NO_READ_WRITE
+		| HA_INPLACE_ADD_INDEX_NO_WRITE
 		| HA_INPLACE_DROP_INDEX_NO_READ_WRITE
 		| HA_INPLACE_ADD_UNIQUE_INDEX_NO_READ_WRITE
+		| HA_INPLACE_ADD_UNIQUE_INDEX_NO_WRITE
 		| HA_INPLACE_DROP_UNIQUE_INDEX_NO_READ_WRITE
 		| HA_INPLACE_ADD_PK_INDEX_NO_READ_WRITE);
 }
@@ -4403,7 +4411,11 @@ UNIV_INTERN
 uint
 ha_innobase::max_supported_key_part_length() const
 {
-	return(DICT_MAX_INDEX_COL_LEN - 1);
+	/* A table format specific index column length check will be performed
+	at ha_innobase::add_index() and row_create_index_for_mysql() */
+	return(innobase_large_prefix
+		? REC_VERSION_56_MAX_INDEX_COL_LEN
+		: REC_ANTELOPE_MAX_INDEX_COL_LEN - 1);
 }
 
 /******************************************************************//**
@@ -7533,8 +7545,8 @@ ha_innobase::create(
 
 		if (i != (uint) primary_key_no) {
 
-			if ((error = create_index(trx, form, flags, norm_name,
-						  i))) {
+			if ((error = create_index(trx, form, flags,
+						  norm_name, i))) {
 				goto cleanup;
 			}
 		}
@@ -11714,6 +11726,11 @@ static MYSQL_SYSVAR_STR(flush_method, innobase_file_flush_method,
   PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY,
   "With which method to flush data.", NULL, NULL, NULL);
 
+static MYSQL_SYSVAR_BOOL(large_prefix, innobase_large_prefix,
+  PLUGIN_VAR_NOCMDARG,
+  "Support large index prefix length of REC_VERSION_56_MAX_INDEX_COL_LEN (3072) bytes.",
+  NULL, NULL, FALSE);
+
 static MYSQL_SYSVAR_BOOL(locks_unsafe_for_binlog, innobase_locks_unsafe_for_binlog,
   PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY,
   "Force InnoDB to not use next-key locking, to use only row-level locking.",
@@ -12162,6 +12179,7 @@ static struct st_mysql_sys_var* innobase_system_variables[]= {
   MYSQL_SYSVAR(use_global_flush_log_at_trx_commit),
   MYSQL_SYSVAR(flush_method),
   MYSQL_SYSVAR(force_recovery),
+  MYSQL_SYSVAR(large_prefix),
   MYSQL_SYSVAR(locks_unsafe_for_binlog),
   MYSQL_SYSVAR(lock_wait_timeout),
 #ifdef UNIV_LOG_ARCHIVE

handler/ha_innodb.h

@@ -217,7 +217,9 @@ class ha_innobase: public handler
 	bool primary_key_is_clustered();
 	int cmp_ref(const uchar *ref1, const uchar *ref2);
 	/** Fast index creation (smart ALTER TABLE) @see handler0alter.cc @{ */
-	int add_index(TABLE *table_arg, KEY *key_info, uint num_of_keys);
+	int add_index(TABLE *table_arg, KEY *key_info, uint num_of_keys,
+		      handler_add_index **add);
+	int final_add_index(handler_add_index *add, bool commit);
 	int prepare_drop_index(TABLE *table_arg, uint *key_num,
 			       uint num_of_keys);
 	int final_drop_index(TABLE *table_arg);

handler/handler0alter.cc

@@ -539,7 +539,7 @@ innobase_create_key_def(
 	if (!new_primary && (key_info->flags & HA_NOSAME)
 	    && (!(key_info->flags & HA_KEY_HAS_PART_KEY_SEG))
 	    && row_table_got_default_clust_index(table)) {
-		uint    key_part = key_info->key_parts;
+		uint	key_part = key_info->key_parts;
 
 		new_primary = TRUE;
 
@@ -594,6 +594,27 @@ innobase_create_key_def(
 	DBUG_RETURN(indexdefs);
 }
 
+/*******************************************************************//**
+Check each index column size, make sure they do not exceed the max limit
+@return	HA_ERR_INDEX_COL_TOO_LONG if index column size exceeds limit */
+static
+int
+innobase_check_column_length(
+/*=========================*/
+	const dict_table_t*table,	/*!< in: table definition */
+	const KEY*	key_info)	/*!< in: Indexes to be created */
+{
+	ulint	max_col_len = DICT_MAX_FIELD_LEN_BY_FORMAT(table);
+
+	for (ulint key_part = 0; key_part < key_info->key_parts; key_part++) {
+		if (key_info->key_part[key_part].length > max_col_len) {
+			my_error(ER_INDEX_COLUMN_TOO_LONG, MYF(0), max_col_len);
+			return(HA_ERR_INDEX_COL_TOO_LONG);
+		}
+	}
+	return(0);
+}
+
 /*******************************************************************//**
 Create a temporary tablename using query id, thread id, and id
 @return	temporary tablename */
@@ -619,6 +640,18 @@ innobase_create_temporary_tablename(
 	return(name);
 }
 
+class ha_innobase_add_index : public handler_add_index
+{
+public:
+	/** table where the indexes are being created */
+	dict_table_t* indexed_table;
+	ha_innobase_add_index(TABLE* table, KEY* key_info, uint num_of_keys,
+			      dict_table_t* indexed_table_arg) :
+		handler_add_index(table, key_info, num_of_keys),
+		indexed_table (indexed_table_arg) {}
+	~ha_innobase_add_index() {}
+};
+
 /*******************************************************************//**
 Create indexes.
 @return	0 or error number */
@@ -626,12 +659,15 @@ UNIV_INTERN
 int
 ha_innobase::add_index(
 /*===================*/
-	TABLE*	table,		/*!< in: Table where indexes are created */
-	KEY*	key_info,	/*!< in: Indexes to be created */
-	uint	num_of_keys)	/*!< in: Number of indexes to be created */
+	TABLE*			table,		/*!< in: Table where indexes
+						are created */
+	KEY*			key_info,	/*!< in: Indexes
+						to be created */
+	uint			num_of_keys,	/*!< in: Number of indexes
+						to be created */
+	handler_add_index**	add)		/*!< out: context */
 {
 	dict_index_t**	index;		/*!< Index to be created */
-	dict_table_t*	innodb_table;	/*!< InnoDB table in dictionary */
 	dict_table_t*	indexed_table;	/*!< Table where indexes are created */
 	merge_index_def_t* index_defs;	/*!< Index definitions */
 	mem_heap_t*     heap;		/*!< Heap for index definitions */
@@ -647,6 +683,8 @@ ha_innobase::add_index(
 	ut_a(key_info);
 	ut_a(num_of_keys);
 
+	*add = NULL;
+
 	if (srv_created_new_raw || srv_force_recovery) {
 		DBUG_RETURN(HA_ERR_WRONG_COMMAND);
 	}
@@ -662,20 +700,32 @@ ha_innobase::add_index(
 		DBUG_RETURN(-1);
 	}
 
-	innodb_table = indexed_table
-		= dict_table_get(prebuilt->table->name, FALSE);
+	indexed_table = dict_table_get(prebuilt->table->name, FALSE);
 
-	if (UNIV_UNLIKELY(!innodb_table)) {
+	if (UNIV_UNLIKELY(!indexed_table)) {
 		DBUG_RETURN(HA_ERR_NO_SUCH_TABLE);
 	}
 
+	ut_a(indexed_table == prebuilt->table);
+
 	/* Check that index keys are sensible */
-	error = innobase_check_index_keys(key_info, num_of_keys, innodb_table);
+	error = innobase_check_index_keys(key_info, num_of_keys, prebuilt->table);
 
 	if (UNIV_UNLIKELY(error)) {
 		DBUG_RETURN(error);
 	}
 
+	/* Check each index's column length to make sure they do not
+	exceed limit */
+	for (ulint i = 0; i < num_of_keys; i++) {
+		error = innobase_check_column_length(prebuilt->table,
+						     &key_info[i]);
+
+		if (error) {
+			DBUG_RETURN(error);
+		}
+	}
+
 	heap = mem_heap_create(1024);
 	trx_start_if_not_started(prebuilt->trx);
 
@@ -691,7 +741,7 @@ ha_innobase::add_index(
 	num_of_idx = num_of_keys;
 
 	index_defs = innobase_create_key_def(
-		trx, innodb_table, heap, key_info, num_of_idx);
+		trx, prebuilt->table, heap, key_info, num_of_idx);
 
 	new_primary = DICT_CLUSTERED & index_defs[0].ind_type;
 
@@ -705,7 +755,7 @@ ha_innobase::add_index(
 	trx_set_dict_operation(trx, TRX_DICT_OP_INDEX);
 
 	/* Acquire a lock on the table before creating any indexes. */
-	error = row_merge_lock_table(prebuilt->trx, innodb_table,
+	error = row_merge_lock_table(prebuilt->trx, prebuilt->table,
 				     new_primary ? LOCK_X : LOCK_S);
 
 	if (UNIV_UNLIKELY(error != DB_SUCCESS)) {
@@ -719,7 +769,7 @@ ha_innobase::add_index(
 	row_mysql_lock_data_dictionary(trx);
 	dict_locked = TRUE;
 
-	ut_d(dict_table_check_for_dup_indexes(innodb_table, FALSE));
+	ut_d(dict_table_check_for_dup_indexes(prebuilt->table, FALSE));
 
 	/* If a new primary key is defined for the table we need
 	to drop the original table and rebuild all indexes. */
@@ -727,15 +777,15 @@ ha_innobase::add_index(
 	if (UNIV_UNLIKELY(new_primary)) {
 		/* This transaction should be the only one
 		operating on the table. */
-		ut_a(innodb_table->n_mysql_handles_opened == 1);
+		ut_a(prebuilt->table->n_mysql_handles_opened == 1);
 
 		char*	new_table_name = innobase_create_temporary_tablename(
-			heap, '1', innodb_table->name);
+			heap, '1', prebuilt->table->name);
 
 		/* Clone the table. */
 		trx_set_dict_operation(trx, TRX_DICT_OP_TABLE);
 		indexed_table = row_merge_create_temporary_table(
-			new_table_name, index_defs, innodb_table, trx);
+			new_table_name, index_defs, prebuilt->table, trx);
 
 		if (!indexed_table) {
 
@@ -749,11 +799,12 @@ ha_innobase::add_index(
 				break;
 			default:
 				error = convert_error_code_to_mysql(
-					trx->error_state, innodb_table->flags,
+					trx->error_state,
+					prebuilt->table->flags,
 					user_thd);
 			}
 
-			ut_d(dict_table_check_for_dup_indexes(innodb_table,
+			ut_d(dict_table_check_for_dup_indexes(prebuilt->table,
 							      FALSE));
 			mem_heap_free(heap);
 			trx_general_rollback_for_mysql(trx, NULL);
@@ -768,17 +819,15 @@ ha_innobase::add_index(
 
 	/* Create the indexes in SYS_INDEXES and load into dictionary. */
 
-	for (ulint i = 0; i < num_of_idx; i++) {
+	for (num_created = 0; num_created < num_of_idx; num_created++) {
 
-		index[i] = row_merge_create_index(trx, indexed_table,
-						  &index_defs[i]);
+		index[num_created] = row_merge_create_index(
+			trx, indexed_table, &index_defs[num_created]);
 
-		if (!index[i]) {
+		if (!index[num_created]) {
 			error = trx->error_state;
 			goto error_handling;
 		}
-
-		num_created++;
 	}
 
 	ut_ad(error == DB_SUCCESS);
@@ -800,7 +849,7 @@ ha_innobase::add_index(
 	if (UNIV_UNLIKELY(new_primary)) {
 		/* A primary key is to be built.  Acquire an exclusive
 		table lock also on the table that is being created. */
-		ut_ad(indexed_table != innodb_table);
+		ut_ad(indexed_table != prebuilt->table);
 
 		error = row_merge_lock_table(prebuilt->trx, indexed_table,
 					     LOCK_X);
@@ -814,7 +863,7 @@ ha_innobase::add_index(
 	/* Read the clustered index of the table and build indexes
 	based on this information using temporary files and merge sort. */
 	error = row_merge_build_indexes(prebuilt->trx,
-					innodb_table, indexed_table,
+					prebuilt->table, indexed_table,
 					index, num_of_idx, table);
 
 error_handling:
@@ -822,63 +871,15 @@ error_handling:
 	dictionary which were defined. */
 
 	switch (error) {
-		const char*	old_name;
-		char*		tmp_name;
 	case DB_SUCCESS:
 		ut_a(!dict_locked);
-		row_mysql_lock_data_dictionary(trx);
-		dict_locked = TRUE;
 
+		ut_d(mutex_enter(&dict_sys->mutex));
 		ut_d(dict_table_check_for_dup_indexes(prebuilt->table, TRUE));
-
-		if (!new_primary) {
-			error = row_merge_rename_indexes(trx, indexed_table);
-
-			if (error != DB_SUCCESS) {
-				row_merge_drop_indexes(trx, indexed_table,
-						       index, num_created);
-			}
-
-			goto convert_error;
-		}
-
-		/* If a new primary key was defined for the table and
-		there was no error at this point, we can now rename
-		the old table as a temporary table, rename the new
-		temporary table as the old table and drop the old table. */
-		old_name = innodb_table->name;
-		tmp_name = innobase_create_temporary_tablename(heap, '2',
-							       old_name);
-
-		error = row_merge_rename_tables(innodb_table, indexed_table,
-						tmp_name, trx);
-
-		if (error != DB_SUCCESS) {
-
-			row_merge_drop_table(trx, indexed_table);
-
-			switch (error) {
-			case DB_TABLESPACE_ALREADY_EXISTS:
-			case DB_DUPLICATE_KEY:
-				innobase_convert_tablename(tmp_name);
-				my_error(HA_ERR_TABLE_EXIST, MYF(0), tmp_name);
-				error = HA_ERR_TABLE_EXIST;
-				break;
-			default:
-				goto convert_error;
-			}
-			break;
-		}
-
-		trx_commit_for_mysql(prebuilt->trx);
-		row_prebuilt_free(prebuilt, TRUE);
-		prebuilt = row_create_prebuilt(indexed_table);
-
-		indexed_table->n_mysql_handles_opened++;
-
-		error = row_merge_drop_table(trx, innodb_table);
-		innodb_table = indexed_table;
-		goto convert_error;
+		ut_d(mutex_exit(&dict_sys->mutex));
+                *add = new ha_innobase_add_index(table, key_info, num_of_keys,
+                                                 indexed_table);
+		break;
 
 	case DB_TOO_BIG_RECORD:
 		my_error(HA_ERR_TO_BIG_ROW, MYF(0));
@@ -894,7 +895,7 @@ error:
 		trx->error_state = DB_SUCCESS;
 
 		if (new_primary) {
-			if (indexed_table != innodb_table) {
+			if (indexed_table != prebuilt->table) {
 				row_merge_drop_table(trx, indexed_table);
 			}
 		} else {
@@ -906,38 +907,161 @@ error:
 			row_merge_drop_indexes(trx, indexed_table,
 					       index, num_created);
 		}
-
-convert_error:
-		if (error == DB_SUCCESS) {
-			/* Build index is successful. We will need to
-			rebuild index translation table.  Reset the
-			index entry count in the translation table
-			to zero, so that translation table will be rebuilt */
-			share->idx_trans_tbl.index_count = 0;
-		}
-
-		error = convert_error_code_to_mysql(error,
-						    innodb_table->flags,
-						    user_thd);
 	}
 
-	mem_heap_free(heap);
 	trx_commit_for_mysql(trx);
 	if (prebuilt->trx) {
 		trx_commit_for_mysql(prebuilt->trx);
 	}
 
 	if (dict_locked) {
-		ut_d(dict_table_check_for_dup_indexes(innodb_table, FALSE));
 		row_mysql_unlock_data_dictionary(trx);
 	}
 
 	trx_free_for_mysql(trx);
+	mem_heap_free(heap);
 
 	/* There might be work for utility threads.*/
 	srv_active_wake_master_thread();
 
-	DBUG_RETURN(error);
+	DBUG_RETURN(convert_error_code_to_mysql(error, prebuilt->table->flags,
+						user_thd));
+}
+
+/*******************************************************************//**
+Finalize or undo add_index().
+@return	0 or error number */
+UNIV_INTERN
+int
+ha_innobase::final_add_index(
+/*=========================*/
+	handler_add_index*	add_arg,/*!< in: context from add_index() */
+	bool			commit)	/*!< in: true=commit, false=rollback */
+{
+	ha_innobase_add_index*	add;
+	trx_t*			trx;
+	int			err	= 0;
+
+	DBUG_ENTER("ha_innobase::final_add_index");
+
+	ut_ad(add_arg);
+	add = static_cast<class ha_innobase_add_index*>(add_arg);
+
+	/* Create a background transaction for the operations on
+	the data dictionary tables. */
+	trx = innobase_trx_allocate(user_thd);
+	trx_start_if_not_started(trx);
+
+	/* Flag this transaction as a dictionary operation, so that
+	the data dictionary will be locked in crash recovery. */
+	trx_set_dict_operation(trx, TRX_DICT_OP_INDEX);
+
+	/* Latch the InnoDB data dictionary exclusively so that no deadlocks
+	or lock waits can happen in it during an index create operation. */
+	row_mysql_lock_data_dictionary(trx);
+
+	if (add->indexed_table != prebuilt->table) {
+		ulint	error;
+
+		/* We copied the table (new_primary). */
+		if (commit) {
+			mem_heap_t*	heap;
+			char*		tmp_name;
+
+			heap = mem_heap_create(1024);
+
+			/* A new primary key was defined for the table
+			and there was no error at this point. We can
+			now rename the old table as a temporary table,
+			rename the new temporary table as the old
+			table and drop the old table. */
+			tmp_name = innobase_create_temporary_tablename(
+				heap, '2', prebuilt->table->name);
+
+			error = row_merge_rename_tables(
+				prebuilt->table, add->indexed_table,
+				tmp_name, trx);
+
+			switch (error) {
+			case DB_TABLESPACE_ALREADY_EXISTS:
+			case DB_DUPLICATE_KEY:
+				innobase_convert_tablename(tmp_name);
+				my_error(HA_ERR_TABLE_EXIST, MYF(0), tmp_name);
+				err = HA_ERR_TABLE_EXIST;
+				break;
+			default:
+				err = convert_error_code_to_mysql(
+					error, prebuilt->table->flags,
+					user_thd);
+				break;
+			}
+
+			mem_heap_free(heap);
+		}
+
+		if (!commit || err) {
+			error = row_merge_drop_table(trx, add->indexed_table);
+			trx_commit_for_mysql(prebuilt->trx);
+		} else {
+			dict_table_t*	old_table = prebuilt->table;
+			trx_commit_for_mysql(prebuilt->trx);
+			row_prebuilt_free(prebuilt, TRUE);
+			error = row_merge_drop_table(trx, old_table);
+			add->indexed_table->n_mysql_handles_opened++;
+			prebuilt = row_create_prebuilt(add->indexed_table);
+		}
+
+		err = convert_error_code_to_mysql(
+			error, prebuilt->table->flags, user_thd);
+	} else {
+		/* We created secondary indexes (!new_primary). */
+
+		if (commit) {
+			err = convert_error_code_to_mysql(
+				row_merge_rename_indexes(trx, prebuilt->table),
+				prebuilt->table->flags, user_thd);
+		}
+
+		if (!commit || err) {
+			dict_index_t*	index;
+			dict_index_t*	next_index;
+
+			for (index = dict_table_get_first_index(
+				     prebuilt->table);
+			     index; index = next_index) {
+
+				next_index = dict_table_get_next_index(index);
+
+				if (*index->name == TEMP_INDEX_PREFIX) {
+					row_merge_drop_index(
+						index, prebuilt->table, trx);
+				}
+			}
+		}
+	}
+
+	/* If index is successfully built, we will need to rebuild index
+	translation table. Set valid index entry count in the translation
+	table to zero. */
+	if (err == 0 && commit) {
+		share->idx_trans_tbl.index_count = 0;
+	}
+
+	trx_commit_for_mysql(trx);
+	if (prebuilt->trx) {
+		trx_commit_for_mysql(prebuilt->trx);
+	}
+
+	ut_d(dict_table_check_for_dup_indexes(prebuilt->table, FALSE));
+	row_mysql_unlock_data_dictionary(trx);
+
+	trx_free_for_mysql(trx);
+
+	/* There might be work for utility threads.*/
+	srv_active_wake_master_thread();
+
+	delete add;
+	DBUG_RETURN(err);
 }
 
 /*******************************************************************//**

include/buf0buf.h

@@ -53,6 +53,9 @@ Created 11/5/1995 Heikki Tuuri
 					/*!< Get the page only if it's in the
 					buffer pool, if not then set a watch
 					on the page. */
+#define BUF_GET_POSSIBLY_FREED		16
+					/*!< Like BUF_GET, but do not mind
+					if the file page has been freed. */
 /* @} */
 /** @name Modes for buf_page_get_known_nowait */
 /* @{ */

include/db0err.h

@@ -110,6 +110,8 @@ enum db_err {
 	DB_PARENT_NO_INDEX,		/* the parent table does not
 					have an index that contains the
 					foreign keys as its prefix columns */
+	DB_TOO_BIG_INDEX_COL,		/* index column size exceeds maximum
+					limit */
 
 	/* The following are partial failure codes */
 	DB_FAIL = 1000,

include/dict0dict.h

@@ -136,6 +136,19 @@ dict_col_copy_type(
 /*===============*/
 	const dict_col_t*	col,	/*!< in: column */
 	dtype_t*		type);	/*!< out: data type */
+/**********************************************************************//**
+Determine bytes of column prefix to be stored in the undo log. Please
+note if the table format is UNIV_FORMAT_A (< DICT_TF_FORMAT_ZIP), no prefix
+needs to be stored in the undo log.
+@return bytes of column prefix to be stored in the undo log */
+UNIV_INLINE
+ulint
+dict_max_field_len_store_undo(
+/*==========================*/
+	dict_table_t*		table,	/*!< in: table */
+	const dict_col_t*	col);	/*!< in: column which index prefix
+					is based on */
+
 #endif /* !UNIV_HOTBACKUP */
 #ifdef UNIV_DEBUG
 /*********************************************************************//**

include/dict0dict.ic

@@ -924,4 +924,30 @@ dict_table_get_on_id_low(
 
 	return(table);
 }
+
+/**********************************************************************//**
+Determine bytes of column prefix to be stored in the undo log. Please
+note if the table format is UNIV_FORMAT_A (< DICT_TF_FORMAT_ZIP), no prefix
+needs to be stored in the undo log.
+@return bytes of column prefix to be stored in the undo log */
+UNIV_INLINE
+ulint
+dict_max_field_len_store_undo(
+/*==========================*/
+	dict_table_t*		table,	/*!< in: table */
+	const dict_col_t*	col)	/*!< in: column which index prefix
+					is based on */
+{
+	ulint   prefix_len = 0;
+
+	if (dict_table_get_format(table) >= DICT_TF_FORMAT_ZIP)
+	{
+		prefix_len = col->max_prefix
+			? col->max_prefix
+			: DICT_MAX_FIELD_LEN_BY_FORMAT(table);
+	}
+
+	return(prefix_len);
+}
+
 #endif /* !UNIV_HOTBACKUP */

include/dict0load.h

@@ -156,7 +156,12 @@ dict_load_field_low(
 	byte*		last_index_id,	/*!< in: last index id */
 	mem_heap_t*	heap,		/*!< in/out: memory heap
 					for temporary storage */
-	const rec_t*	rec);		/*!< in: SYS_FIELDS record */
+	const rec_t*	rec,		/*!< in: SYS_FIELDS record */
+	char*		addition_err_str,/*!< out: additional error message
+					that requires information to be
+					filled, or NULL */
+	ulint		err_str_len);	/*!< in: length of addition_err_str
+					in bytes */
 /********************************************************************//**
 Loads a table definition and also all its index definitions, and also
 the cluster definition if the table is a member in a cluster. Also loads

include/dict0mem.h

@@ -302,32 +302,58 @@ struct dict_col_struct{
 	unsigned	ord_part:1;	/*!< nonzero if this column
 					appears in the ordering fields
 					of an index */
+	unsigned	max_prefix:12;	/*!< maximum index prefix length on
+					this column. Our current max limit is
+					3072 for Barracuda table */
 };
 
-/** @brief DICT_MAX_INDEX_COL_LEN is measured in bytes and is the maximum
-indexed column length (or indexed prefix length).
+/** @brief DICT_ANTELOPE_MAX_INDEX_COL_LEN is measured in bytes and
+is the maximum indexed column length (or indexed prefix length) in
+ROW_FORMAT=REDUNDANT and ROW_FORMAT=COMPACT. Also, in any format,
+any fixed-length field that is longer than this will be encoded as
+a variable-length field.
 
 It is set to 3*256, so that one can create a column prefix index on
 256 characters of a TEXT or VARCHAR column also in the UTF-8
 charset. In that charset, a character may take at most 3 bytes.  This
 constant MUST NOT BE CHANGED, or the compatibility of InnoDB data
 files would be at risk! */
-#define DICT_MAX_INDEX_COL_LEN		REC_MAX_INDEX_COL_LEN
+#define DICT_ANTELOPE_MAX_INDEX_COL_LEN	REC_ANTELOPE_MAX_INDEX_COL_LEN
+
+/** Find out maximum indexed column length by its table format.
+For ROW_FORMAT=REDUNDANT and ROW_FORMAT=COMPACT, the maximum
+field length is REC_ANTELOPE_MAX_INDEX_COL_LEN - 1 (767). For new
+barracuda format, the length could be REC_VERSION_56_MAX_INDEX_COL_LEN
+(3072) bytes */
+#define DICT_MAX_FIELD_LEN_BY_FORMAT(table)				\
+		((dict_table_get_format(table) < DICT_TF_FORMAT_ZIP)	\
+			? (REC_ANTELOPE_MAX_INDEX_COL_LEN - 1)		\
+			: REC_VERSION_56_MAX_INDEX_COL_LEN)
+
+#define DICT_MAX_FIELD_LEN_BY_FORMAT_FLAG(flags)			\
+		((((flags & DICT_TF_FORMAT_MASK) >> DICT_TF_FORMAT_SHIFT)\
+		    < DICT_TF_FORMAT_ZIP)				\
+			? (REC_ANTELOPE_MAX_INDEX_COL_LEN - 1)		\
+			: REC_VERSION_56_MAX_INDEX_COL_LEN)
+
+/** Defines the maximum fixed length column size */
+#define DICT_MAX_FIXED_COL_LEN		DICT_ANTELOPE_MAX_INDEX_COL_LEN
 
 /** Data structure for a field in an index */
 struct dict_field_struct{
 	dict_col_t*	col;		/*!< pointer to the table column */
 	const char*	name;		/*!< name of the column */
-	unsigned	prefix_len:10;	/*!< 0 or the length of the column
+	unsigned	prefix_len:12;	/*!< 0 or the length of the column
 					prefix in bytes in a MySQL index of
 					type, e.g., INDEX (textcol(25));
 					must be smaller than
-					DICT_MAX_INDEX_COL_LEN; NOTE that
-					in the UTF-8 charset, MySQL sets this
-					to 3 * the prefix len in UTF-8 chars */
+					DICT_MAX_FIELD_LEN_BY_FORMAT;
+					NOTE that in the UTF-8 charset, MySQL
+					sets this to (mbmaxlen * the prefix len)
+					in UTF-8 chars */
 	unsigned	fixed_len:10;	/*!< 0 or the fixed length of the
 					column if smaller than
-					DICT_MAX_INDEX_COL_LEN */
+					DICT_ANTELOPE_MAX_INDEX_COL_LEN */
 };
 
 /** Data structure for an index.  Most fields will be

include/fil0fil.h

@@ -671,8 +671,9 @@ UNIV_INTERN
 void
 fil_flush(
 /*======*/
-	ulint	space_id);	/*!< in: file space id (this can be a group of
+	ulint	space_id,	/*!< in: file space id (this can be a group of
 				log files or a tablespace of the database) */
+	ibool	metadata);
 /**********************************************************************//**
 Flushes to disk writes in file spaces of the given type possibly cached by
 the OS. */

include/lock0lock.h

@@ -73,9 +73,10 @@ UNIV_INLINE
 trx_t*
 lock_clust_rec_some_has_impl(
 /*=========================*/
-	const rec_t*	rec,	/*!< in: user record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets);/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: user record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
+	__attribute__((nonnull, warn_unused_result));
 /*********************************************************************//**
 Gets the heap_no of the smallest user record on a page.
 @return	heap_no of smallest user record, or PAGE_HEAP_NO_SUPREMUM */

include/lock0lock.ic

@@ -75,9 +75,9 @@ UNIV_INLINE
 trx_t*
 lock_clust_rec_some_has_impl(
 /*=========================*/
-	const rec_t*	rec,	/*!< in: user record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets)/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: user record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
 {
 	trx_id_t	trx_id;
 

include/mtr0mtr.ic

@@ -37,6 +37,8 @@ mtr_start(
 /*======*/
 	mtr_t*	mtr)	/*!< out: mini-transaction */
 {
+	UNIV_MEM_INVALID(mtr, sizeof *mtr);
+
 	dyn_array_create(&(mtr->memo));
 	dyn_array_create(&(mtr->log));
 

include/os0file.h

@@ -290,8 +290,7 @@ The wrapper functions have the prefix of "innodb_". */
 			      __FILE__, __LINE__)
 
 # define os_file_read_trx(file, buf, offset, offset_high, n, trx)	\
-	pfs_os_file_read_func(file, buf, offset, offset_high, n, trx,	\
-			      __FILE__, __LINE__)
+	os_file_read_func(file, buf, offset, offset_high, n, trx)
 
 # define os_file_read_no_error_handling(file, buf, offset,		\
 					offset_high, n)			\
@@ -303,8 +302,8 @@ The wrapper functions have the prefix of "innodb_". */
 	pfs_os_file_write_func(name, file, buf, offset, offset_high,	\
 			       n, __FILE__, __LINE__)
 
-# define os_file_flush(file)						\
-	pfs_os_file_flush_func(file, __FILE__, __LINE__)
+# define os_file_flush(file, metadata)					\
+	pfs_os_file_flush_func(file, metadata, __FILE__, __LINE__)
 
 # define os_file_rename(key, oldpath, newpath)				\
 	pfs_os_file_rename_func(key, oldpath, newpath, __FILE__, __LINE__)
@@ -343,7 +342,7 @@ to original un-instrumented file I/O APIs */
 # define os_file_write(name, file, buf, offset, offset_high, n)		\
 	os_file_write_func(name, file, buf, offset, offset_high, n)
 
-# define os_file_flush(file)	os_file_flush_func(file)
+# define os_file_flush(file, metadata)	os_file_flush_func(file, metadata)
 
 # define os_file_rename(key, oldpath, newpath)				\
 	os_file_rename_func(oldpath, newpath)
@@ -794,6 +793,7 @@ ibool
 pfs_os_file_flush_func(
 /*===================*/
 	os_file_t	file,	/*!< in, own: handle to a file */
+	ibool		metadata,
 	const char*	src_file,/*!< in: file name where func invoked */
 	ulint		src_line);/*!< in: line where the func invoked */
 
@@ -873,7 +873,8 @@ UNIV_INTERN
 ibool
 os_file_flush_func(
 /*===============*/
-	os_file_t	file);	/*!< in, own: handle to a file */
+	os_file_t	file,	/*!< in, own: handle to a file */
+	ibool		metadata);
 /***********************************************************************//**
 Retrieves the last error number if an error occurs in a file io function.
 The number should be retrieved before any other OS calls (because they may

include/os0file.ic

@@ -372,6 +372,7 @@ ibool
 pfs_os_file_flush_func(
 /*===================*/
 	os_file_t	file,	/*!< in, own: handle to a file */
+	ibool		metadata,
 	const char*	src_file,/*!< in: file name where func invoked */
 	ulint		src_line)/*!< in: line where the func invoked */
 {
@@ -381,7 +382,7 @@ pfs_os_file_flush_func(
 
 	register_pfs_file_io_begin(&state, locker, file, 0, PSI_FILE_SYNC,
 				   src_file, src_line);
-	result = os_file_flush_func(file);
+	result = os_file_flush_func(file, metadata);
 
 	register_pfs_file_io_end(locker, 0);
 

include/page0page.h

@@ -618,18 +618,19 @@ rec_t*
 page_rec_find_owner_rec(
 /*====================*/
 	rec_t*	rec);	/*!< in: the physical record */
+#ifndef UNIV_HOTBACKUP
 /***********************************************************************//**
-This is a low-level operation which is used in a database index creation
-to update the page number of a created B-tree to a data dictionary
-record. */
-UNIV_INTERN
+Write a 32-bit field in a data dictionary record. */
+UNIV_INLINE
 void
-page_rec_write_index_page_no(
-/*=========================*/
-	rec_t*	rec,	/*!< in: record to update */
+page_rec_write_field(
+/*=================*/
+	rec_t*	rec,	/*!< in/out: record to update */
 	ulint	i,	/*!< in: index of the field to update */
-	ulint	page_no,/*!< in: value to write */
-	mtr_t*	mtr);	/*!< in: mtr */
+	ulint	val,	/*!< in: value to write */
+	mtr_t*	mtr)	/*!< in/out: mini-transaction */
+	__attribute__((nonnull));
+#endif /* !UNIV_HOTBACKUP */
 /************************************************************//**
 Returns the maximum combined size of records which can be inserted on top
 of record heap.

include/page0page.ic

@@ -962,6 +962,29 @@ page_get_free_space_of_empty(
 		       - 2 * PAGE_DIR_SLOT_SIZE));
 }
 
+#ifndef UNIV_HOTBACKUP
+/***********************************************************************//**
+Write a 32-bit field in a data dictionary record. */
+UNIV_INLINE
+void
+page_rec_write_field(
+/*=================*/
+	rec_t*	rec,	/*!< in/out: record to update */
+	ulint	i,	/*!< in: index of the field to update */
+	ulint	val,	/*!< in: value to write */
+	mtr_t*	mtr)	/*!< in/out: mini-transaction */
+{
+	byte*	data;
+	ulint	len;
+
+	data = rec_get_nth_field_old(rec, i, &len);
+
+	ut_ad(len == 4);
+
+	mlog_write_ulint(data, val, MLOG_4BYTES, mtr);
+}
+#endif /* !UNIV_HOTBACKUP */
+
 /************************************************************//**
 Each user record on a page, and also the deleted user records in the heap
 takes its size plus the fraction of the dir cell size /

include/rem0rec.h

@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1994, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -480,6 +480,18 @@ ulint
 rec_offs_any_extern(
 /*================*/
 	const ulint*	offsets);/*!< in: array returned by rec_get_offsets() */
+#ifdef UNIV_BLOB_NULL_DEBUG
+/******************************************************//**
+Determine if the offsets are for a record containing null BLOB pointers.
+@return	first field containing a null BLOB pointer, or NULL if none found */
+UNIV_INLINE
+const byte*
+rec_offs_any_null_extern(
+/*=====================*/
+	const rec_t*	rec,		/*!< in: record */
+	const ulint*	offsets)	/*!< in: rec_get_offsets(rec) */
+	__attribute__((nonnull, warn_unused_result));
+#endif /* UNIV_BLOB_NULL_DEBUG */
 /******************************************************//**
 Returns nonzero if the extern bit is set in nth field of rec.
 @return	nonzero if externally stored */

include/rem0rec.ic

@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1994, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1994, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -26,6 +26,7 @@ Created 5/30/1994 Heikki Tuuri
 #include "mach0data.h"
 #include "ut0byte.h"
 #include "dict0dict.h"
+#include "btr0types.h"
 
 /* Compact flag ORed to the extra size returned by rec_get_offsets() */
 #define REC_OFFS_COMPACT	((ulint) 1 << 31)
@@ -1087,6 +1088,44 @@ rec_offs_any_extern(
 	return(UNIV_UNLIKELY(*rec_offs_base(offsets) & REC_OFFS_EXTERNAL));
 }
 
+#ifdef UNIV_BLOB_NULL_DEBUG
+/******************************************************//**
+Determine if the offsets are for a record containing null BLOB pointers.
+@return	first field containing a null BLOB pointer, or NULL if none found */
+UNIV_INLINE
+const byte*
+rec_offs_any_null_extern(
+/*=====================*/
+	const rec_t*	rec,		/*!< in: record */
+	const ulint*	offsets)	/*!< in: rec_get_offsets(rec) */
+{
+	ulint	i;
+	ut_ad(rec_offs_validate(rec, NULL, offsets));
+
+	if (!rec_offs_any_extern(offsets)) {
+		return(NULL);
+	}
+
+	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
+		if (rec_offs_nth_extern(offsets, i)) {
+			ulint		len;
+			const byte*	field
+				= rec_get_nth_field(rec, offsets, i, &len);
+
+			ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);
+			if (!memcmp(field + len
+				    - BTR_EXTERN_FIELD_REF_SIZE,
+				    field_ref_zero,
+				    BTR_EXTERN_FIELD_REF_SIZE)) {
+				return(field);
+			}
+		}
+	}
+
+	return(NULL);
+}
+#endif /* UNIV_BLOB_NULL_DEBUG */
+
 /******************************************************//**
 Returns nonzero if the extern bit is set in nth field of rec.
 @return	nonzero if externally stored */

include/rem0types.h

@@ -34,13 +34,21 @@ typedef byte	rec_t;
 #define REC_MAX_HEAP_NO		(2 * 8192 - 1)
 #define REC_MAX_N_OWNED		(16 - 1)
 
-/* REC_MAX_INDEX_COL_LEN is measured in bytes and is the maximum
-indexed column length (or indexed prefix length). It is set to 3*256,
-so that one can create a column prefix index on 256 characters of a
-TEXT or VARCHAR column also in the UTF-8 charset. In that charset,
-a character may take at most 3 bytes.
+/* REC_ANTELOPE_MAX_INDEX_COL_LEN is measured in bytes and is the maximum
+indexed field length (or indexed prefix length) for indexes on tables of
+ROW_FORMAT=REDUNDANT and ROW_FORMAT=COMPACT format.
+Before we support UTF-8 encodings with mbmaxlen = 4, a UTF-8 character
+may take at most 3 bytes.  So the limit was set to 3*256, so that one
+can create a column prefix index on 256 characters of a TEXT or VARCHAR
+column also in the UTF-8 charset.
 This constant MUST NOT BE CHANGED, or the compatibility of InnoDB data
 files would be at risk! */
-#define REC_MAX_INDEX_COL_LEN	768
+#define REC_ANTELOPE_MAX_INDEX_COL_LEN		768
+
+/** Maximum indexed field length for table format DICT_TF_FORMAT_ZIP and
+beyond.
+This (3072) is the maximum index row length allowed, so we cannot create index
+prefix column longer than that. */
+#define REC_VERSION_56_MAX_INDEX_COL_LEN	3072
 
 #endif

include/row0ext.h

@@ -30,6 +30,7 @@ Created September 2006 Marko Makela
 #include "row0types.h"
 #include "data0types.h"
 #include "mem0mem.h"
+#include "dict0types.h"
 
 /********************************************************************//**
 Creates a cache of column prefixes of externally stored columns.
@@ -43,13 +44,13 @@ row_ext_create(
 				in the InnoDB table object, as reported by
 				dict_col_get_no(); NOT relative to the records
 				in the clustered index */
+	ulint		flags, /*!< in: table->flags */
 	const dtuple_t*	tuple,	/*!< in: data tuple containing the field
 				references of the externally stored
 				columns; must be indexed by col_no;
 				the clustered index record must be
 				covered by a lock or a page latch
 				to prevent deletion (rollback or purge). */
-	ulint		zip_size,/*!< compressed page size in bytes, or 0 */
 	mem_heap_t*	heap);	/*!< in: heap where created */
 
 /********************************************************************//**
@@ -63,7 +64,8 @@ row_ext_lookup_ith(
 	const row_ext_t*	ext,	/*!< in/out: column prefix cache */
 	ulint			i,	/*!< in: index of ext->ext[] */
 	ulint*			len);	/*!< out: length of prefix, in bytes,
-					at most REC_MAX_INDEX_COL_LEN */
+					at most the length determined by
+					DICT_MAX_FIELD_LEN_BY_FORMAT() */
 /********************************************************************//**
 Looks up a column prefix of an externally stored column.
 @return column prefix, or NULL if the column is not stored externally,
@@ -78,13 +80,18 @@ row_ext_lookup(
 					dict_col_get_no(); NOT relative to the
 					records in the clustered index */
 	ulint*			len);	/*!< out: length of prefix, in bytes,
-					at most REC_MAX_INDEX_COL_LEN */
+					at most the length determined by
+					DICT_MAX_FIELD_LEN_BY_FORMAT() */
 
 /** Prefixes of externally stored columns */
 struct row_ext_struct{
 	ulint		n_ext;	/*!< number of externally stored columns */
 	const ulint*	ext;	/*!< col_no's of externally stored columns */
 	byte*		buf;	/*!< backing store of the column prefix cache */
+	ulint		max_len;/*!< maximum prefix length, it could be
+				REC_ANTELOPE_MAX_INDEX_COL_LEN or
+				REC_VERSION_56_MAX_INDEX_COL_LEN depending
+				on row format */
 	ulint		len[1];	/*!< prefix lengths; 0 if not cached */
 };
 

include/row0ext.ic

@@ -37,7 +37,7 @@ row_ext_lookup_ith(
 	const row_ext_t*	ext,	/*!< in/out: column prefix cache */
 	ulint			i,	/*!< in: index of ext->ext[] */
 	ulint*			len)	/*!< out: length of prefix, in bytes,
-					at most REC_MAX_INDEX_COL_LEN */
+					at most ext->max_len */
 {
 	ut_ad(ext);
 	ut_ad(len);
@@ -45,11 +45,14 @@ row_ext_lookup_ith(
 
 	*len = ext->len[i];
 
+	ut_ad(*len <= ext->max_len);
+	ut_ad(ext->max_len > 0);
+
 	if (UNIV_UNLIKELY(*len == 0)) {
 		/* The BLOB could not be fetched to the cache. */
 		return(field_ref_zero);
 	} else {
-		return(ext->buf + i * REC_MAX_INDEX_COL_LEN);
+		return(ext->buf + i * ext->max_len);
 	}
 }
 
@@ -67,7 +70,7 @@ row_ext_lookup(
 					dict_col_get_no(); NOT relative to the
 					records in the clustered index */
 	ulint*			len)	/*!< out: length of prefix, in bytes,
-					at most REC_MAX_INDEX_COL_LEN */
+					at most ext->max_len */
 {
 	ulint	i;
 

include/row0row.h

@@ -41,13 +41,24 @@ Created 4/20/1996 Heikki Tuuri
 Gets the offset of the trx id field, in bytes relative to the origin of
 a clustered index record.
 @return	offset of DATA_TRX_ID */
-UNIV_INTERN
+UNIV_INLINE
 ulint
-row_get_trx_id_offset(
-/*==================*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets);/*!< in: rec_get_offsets(rec, index) */
+row_get_trx_id_offset_func(
+/*=======================*/
+#ifdef UNIV_DEBUG
+	const rec_t*		rec,	/*!< in: record */
+#endif /* UNIV_DEBUG */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
+	__attribute__((nonnull, warn_unused_result));
+#ifdef UNIV_DEBUG
+# define row_get_trx_id_offset(rec, index, offsets)	\
+	row_get_trx_id_offset_func(rec, index, offsets)
+#else /* UNIV_DEBUG */
+# define row_get_trx_id_offset(rec, index, offsets)	\
+	row_get_trx_id_offset_func(index, offsets)
+#endif /* UNIV_DEBUG */
+
 /*********************************************************************//**
 Reads the trx id field from a clustered index record.
 @return	value of the field */
@@ -55,9 +66,10 @@ UNIV_INLINE
 trx_id_t
 row_get_rec_trx_id(
 /*===============*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets);/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
+	__attribute__((nonnull, warn_unused_result));
 /*********************************************************************//**
 Reads the roll pointer field from a clustered index record.
 @return	value of the field */

include/row0row.ic

@@ -27,6 +27,36 @@ Created 4/20/1996 Heikki Tuuri
 #include "rem0rec.h"
 #include "trx0undo.h"
 
+/*********************************************************************//**
+Gets the offset of trx id field, in bytes relative to the origin of
+a clustered index record.
+@return	offset of DATA_TRX_ID */
+UNIV_INLINE
+ulint
+row_get_trx_id_offset_func(
+/*=======================*/
+#ifdef UNIV_DEBUG
+	const rec_t*		rec,	/*!< in: record */
+#endif /* UNIV_DEBUG */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
+{
+	ulint	pos;
+	ulint	offset;
+	ulint	len;
+
+	ut_ad(dict_index_is_clust(index));
+	ut_ad(rec_offs_validate(rec, index, offsets));
+
+	pos = dict_index_get_sys_col_pos(index, DATA_TRX_ID);
+
+	offset = rec_get_nth_field_offs(offsets, pos, &len);
+
+	ut_ad(len == DATA_TRX_ID_LEN);
+
+	return(offset);
+}
+
 /*********************************************************************//**
 Reads the trx id field from a clustered index record.
 @return	value of the field */
@@ -34,9 +64,9 @@ UNIV_INLINE
 trx_id_t
 row_get_rec_trx_id(
 /*===============*/
-	const rec_t*	rec,	/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets)/*!< in: rec_get_offsets(rec, index) */
+	const rec_t*		rec,	/*!< in: record */
+	const dict_index_t*	index,	/*!< in: clustered index */
+	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
 {
 	ulint	offset;
 

include/univ.i

@@ -51,7 +51,7 @@ Created 1/20/1994 Heikki Tuuri
 
 #define INNODB_VERSION_MAJOR	1
 #define INNODB_VERSION_MINOR	1
-#define INNODB_VERSION_BUGFIX	7
+#define INNODB_VERSION_BUGFIX	8
 
 #ifndef PERCONA_INNODB_VERSION
 #define PERCONA_INNODB_VERSION 20.1
@@ -192,6 +192,8 @@ command. Not tested on Windows. */
 						debugging without UNIV_DEBUG */
 #define UNIV_BLOB_LIGHT_DEBUG			/* Enable off-page column
 						debugging without UNIV_DEBUG */
+#define UNIV_BLOB_NULL_DEBUG			/* Enable deep off-page
+						column debugging */
 #define UNIV_DEBUG				/* Enable ut_ad() assertions
 						and disable UNIV_INLINE */
 #define UNIV_DEBUG_LOCK_VALIDATE		/* Enable

lock/lock0lock.c

@@ -359,10 +359,8 @@ static
 ibool
 lock_rec_validate_page(
 /*===================*/
-	ulint	space,	/*!< in: space id */
-	ulint	zip_size,/*!< in: compressed page size in bytes
-			or 0 for uncompressed pages */
-	ulint	page_no);/*!< in: page number */
+	const buf_block_t*	block)	/*!< in: buffer block */
+	__attribute__((nonnull, warn_unused_result));
 #endif /* UNIV_DEBUG */
 
 /* The lock system */
@@ -1101,10 +1099,10 @@ lock_rec_reset_nth_bit(
 Gets the first or next record lock on a page.
 @return	next lock, NULL if none exists */
 UNIV_INLINE
-lock_t*
-lock_rec_get_next_on_page(
-/*======================*/
-	lock_t*	lock)	/*!< in: a record lock */
+const lock_t*
+lock_rec_get_next_on_page_const(
+/*============================*/
+	const lock_t*	lock)	/*!< in: a record lock */
 {
 	ulint	space;
 	ulint	page_no;
@@ -1133,6 +1131,18 @@ lock_rec_get_next_on_page(
 	return(lock);
 }
 
+/*********************************************************************//**
+Gets the first or next record lock on a page.
+@return	next lock, NULL if none exists */
+UNIV_INLINE
+lock_t*
+lock_rec_get_next_on_page(
+/*======================*/
+	lock_t*	lock)	/*!< in: a record lock */
+{
+	return((lock_t*) lock_rec_get_next_on_page_const(lock));
+}
+
 /*********************************************************************//**
 Gets the first record lock on a page, where the page is identified by its
 file address.
@@ -2655,9 +2665,7 @@ lock_move_reorganize_page(
 	mem_heap_free(heap);
 
 #ifdef UNIV_DEBUG_LOCK_VALIDATE
-	ut_ad(lock_rec_validate_page(buf_block_get_space(block),
-				     buf_block_get_zip_size(block),
-				     buf_block_get_page_no(block)));
+	ut_ad(lock_rec_validate_page(block));
 #endif
 }
 
@@ -2745,12 +2753,8 @@ lock_move_rec_list_end(
 	lock_mutex_exit_kernel();
 
 #ifdef UNIV_DEBUG_LOCK_VALIDATE
-	ut_ad(lock_rec_validate_page(buf_block_get_space(block),
-				     buf_block_get_zip_size(block),
-				     buf_block_get_page_no(block)));
-	ut_ad(lock_rec_validate_page(buf_block_get_space(new_block),
-				     buf_block_get_zip_size(block),
-				     buf_block_get_page_no(new_block)));
+	ut_ad(lock_rec_validate_page(block));
+	ut_ad(lock_rec_validate_page(new_block));
 #endif
 }
 
@@ -2858,9 +2862,7 @@ lock_move_rec_list_start(
 	lock_mutex_exit_kernel();
 
 #ifdef UNIV_DEBUG_LOCK_VALIDATE
-	ut_ad(lock_rec_validate_page(buf_block_get_space(block),
-				     buf_block_get_zip_size(block),
-				     buf_block_get_page_no(block)));
+	ut_ad(lock_rec_validate_page(block));
 #endif
 }
 
@@ -3850,17 +3852,18 @@ Checks if other transactions have an incompatible mode lock request in
 the lock queue.
 @return	lock or NULL */
 UNIV_INLINE
-lock_t*
+const lock_t*
 lock_table_other_has_incompatible(
 /*==============================*/
-	trx_t*		trx,	/*!< in: transaction, or NULL if all
-				transactions should be included */
-	ulint		wait,	/*!< in: LOCK_WAIT if also waiting locks are
-				taken into account, or 0 if not */
-	dict_table_t*	table,	/*!< in: table */
-	enum lock_mode	mode)	/*!< in: lock mode */
+	const trx_t*		trx,	/*!< in: transaction, or NULL if all
+					transactions should be included */
+	ulint			wait,	/*!< in: LOCK_WAIT if also
+					waiting locks are taken into
+					account, or 0 if not */
+	const dict_table_t*	table,	/*!< in: table */
+	enum lock_mode		mode)	/*!< in: lock mode */
 {
-	lock_t*	lock;
+	const lock_t*	lock;
 
 	ut_ad(mutex_own(&kernel_mutex));
 
@@ -3951,10 +3954,10 @@ static
 ibool
 lock_table_has_to_wait_in_queue(
 /*============================*/
-	lock_t*	wait_lock)	/*!< in: waiting table lock */
+	const lock_t*	wait_lock)	/*!< in: waiting table lock */
 {
-	dict_table_t*	table;
-	lock_t*		lock;
+	const dict_table_t*	table;
+	const lock_t*		lock;
 
 	ut_ad(mutex_own(&kernel_mutex));
 	ut_ad(lock_get_wait(wait_lock));
@@ -4696,9 +4699,9 @@ static
 ibool
 lock_table_queue_validate(
 /*======================*/
-	dict_table_t*	table)	/*!< in: table */
+	const dict_table_t*	table)	/*!< in: table */
 {
-	lock_t*	lock;
+	const lock_t*	lock;
 
 	ut_ad(mutex_own(&kernel_mutex));
 
@@ -4734,7 +4737,7 @@ lock_rec_queue_validate(
 /*====================*/
 	const buf_block_t*	block,	/*!< in: buffer block containing rec */
 	const rec_t*		rec,	/*!< in: record to look at */
-	dict_index_t*		index,	/*!< in: index, or NULL if not known */
+	const dict_index_t*	index,	/*!< in: index, or NULL if not known */
 	const ulint*		offsets)/*!< in: rec_get_offsets(rec, index) */
 {
 	trx_t*	impl_trx;
@@ -4883,46 +4886,37 @@ static
 ibool
 lock_rec_validate_page(
 /*===================*/
-	ulint	space,	/*!< in: space id */
-	ulint	zip_size,/*!< in: compressed page size in bytes
-			or 0 for uncompressed pages */
-	ulint	page_no)/*!< in: page number */
+	const buf_block_t*	block)	/*!< in: buffer block */
 {
-	dict_index_t*	index;
-	buf_block_t*	block;
-	const page_t*	page;
-	lock_t*		lock;
+	const lock_t*	lock;
 	const rec_t*	rec;
 	ulint		nth_lock	= 0;
 	ulint		nth_bit		= 0;
 	ulint		i;
-	mtr_t		mtr;
 	mem_heap_t*	heap		= NULL;
 	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
 	ulint*		offsets		= offsets_;
 	rec_offs_init(offsets_);
 
 	ut_ad(!mutex_own(&kernel_mutex));
-
-	mtr_start(&mtr);
-
-	ut_ad(zip_size != ULINT_UNDEFINED);
-	block = buf_page_get(space, zip_size, page_no, RW_X_LATCH, &mtr);
-	buf_block_dbg_add_level(block, SYNC_NO_ORDER_CHECK);
-
-	page = block->frame;
+	ut_ad(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
 
 	lock_mutex_enter_kernel();
 loop:
-	lock = lock_rec_get_first_on_page_addr(space, page_no);
+	lock = lock_rec_get_first_on_page_addr(buf_block_get_space(block),
+					       buf_block_get_page_no(block));
 
 	if (!lock) {
 		goto function_exit;
 	}
 
+#if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG
+	ut_a(!block->page.file_page_was_freed);
+#endif
+
 	for (i = 0; i < nth_lock; i++) {
 
-		lock = lock_rec_get_next_on_page(lock);
+		lock = lock_rec_get_next_on_page_const(lock);
 
 		if (!lock) {
 			goto function_exit;
@@ -4945,15 +4939,14 @@ loop:
 
 		if (i == 1 || lock_rec_get_nth_bit(lock, i)) {
 
-			index = lock->index;
-			rec = page_find_rec_with_heap_no(page, i);
+			rec = page_find_rec_with_heap_no(block->frame, i);
 			ut_a(rec);
-			offsets = rec_get_offsets(rec, index, offsets,
+			offsets = rec_get_offsets(rec, lock->index, offsets,
 						  ULINT_UNDEFINED, &heap);
 #if 0
 			fprintf(stderr,
-				"Validating %lu %lu\n",
-				(ulong) space, (ulong) page_no);
+				"Validating %u %u\n",
+				block->page.space, block->page.offset);
 #endif
 			lock_mutex_exit_kernel();
 
@@ -4962,7 +4955,8 @@ loop:
 			check WILL break the latching order and may
 			cause a deadlock of threads. */
 
-			lock_rec_queue_validate(block, rec, index, offsets);
+			lock_rec_queue_validate(block, rec, lock->index,
+						offsets);
 
 			lock_mutex_enter_kernel();
 
@@ -4980,8 +4974,6 @@ loop:
 function_exit:
 	lock_mutex_exit_kernel();
 
-	mtr_commit(&mtr);
-
 	if (UNIV_LIKELY_NULL(heap)) {
 		mem_heap_free(heap);
 	}
@@ -4996,11 +4988,8 @@ ibool
 lock_validate(void)
 /*===============*/
 {
-	lock_t*		lock;
-	trx_t*		trx;
-	ib_uint64_t	limit;
-	ulint		space;
-	ulint		page_no;
+	const lock_t*	lock;
+	const trx_t*	trx;
 	ulint		i;
 
 	lock_mutex_enter_kernel();
@@ -5025,9 +5014,14 @@ lock_validate(void)
 
 	for (i = 0; i < hash_get_n_cells(lock_sys->rec_hash); i++) {
 
-		limit = 0;
+		ulint		space;
+		ulint		page_no;
+		ib_uint64_t	limit	= 0;
 
 		for (;;) {
+			mtr_t		mtr;
+			buf_block_t*	block;
+
 			lock = HASH_GET_FIRST(lock_sys->rec_hash, i);
 
 			while (lock) {
@@ -5053,13 +5047,26 @@ lock_validate(void)
 
 			lock_mutex_exit_kernel();
 
-			lock_rec_validate_page(space,
-					       fil_space_get_zip_size(space),
-					       page_no);
+			/* The lock and the block that it is referring
+			to may be freed at this point. We pass
+			BUF_GET_POSSIBLY_FREED to skip a debug check.
+			If the lock exists in lock_rec_validate_page()
+			we assert !block->page.file_page_was_freed. */
 
-			lock_mutex_enter_kernel();
+			mtr_start(&mtr);
+			block = buf_page_get_gen(
+				space, fil_space_get_zip_size(space),
+				page_no, RW_X_LATCH, NULL,
+				BUF_GET_POSSIBLY_FREED,
+				__FILE__, __LINE__, &mtr);
+			buf_block_dbg_add_level(block, SYNC_NO_ORDER_CHECK);
+
+			ut_ad(lock_rec_validate_page(block));
+			mtr_commit(&mtr);
+
+			limit++;
 
-			limit = ut_ull_create(space, page_no + 1);
+			lock_mutex_enter_kernel();
 		}
 	}
 

log/log0log.c

@@ -1133,7 +1133,7 @@ log_io_complete(
 		    && srv_unix_file_flush_method != SRV_UNIX_ALL_O_DIRECT
 		    && srv_unix_file_flush_method != SRV_UNIX_NOSYNC) {
 
-			fil_flush(group->space_id);
+			fil_flush(group->space_id, FALSE);
 		}
 
 #ifdef UNIV_DEBUG
@@ -1156,7 +1156,7 @@ log_io_complete(
 	    && srv_unix_file_flush_method != SRV_UNIX_NOSYNC
 	    && thd_flush_log_at_trx_commit(NULL) != 2) {
 
-		fil_flush(group->space_id);
+		fil_flush(group->space_id, FALSE);
 	}
 
 	mutex_enter(&(log_sys->mutex));
@@ -1547,7 +1547,7 @@ loop:
 
 		group = UT_LIST_GET_FIRST(log_sys->log_groups);
 
-		fil_flush(group->space_id);
+		fil_flush(group->space_id, FALSE);
 		log_sys->flushed_to_disk_lsn = log_sys->write_lsn;
 	}
 
@@ -2644,7 +2644,7 @@ log_io_complete_archive(void)
 
 	mutex_exit(&(log_sys->mutex));
 
-	fil_flush(group->archive_space_id);
+	fil_flush(group->archive_space_id, TRUE);
 
 	mutex_enter(&(log_sys->mutex));
 

log/log0recv.c

@@ -3643,7 +3643,7 @@ recv_reset_log_files_for_backup(
 			exit(1);
 		}
 
-		os_file_flush(log_file);
+		os_file_flush(log_file, TRUE);
 		os_file_close(log_file);
 	}
 
@@ -3667,7 +3667,7 @@ recv_reset_log_files_for_backup(
 
 	os_file_write(name, log_file, buf, 0, 0,
 		      LOG_FILE_HDR_SIZE + OS_FILE_LOG_BLOCK_SIZE);
-	os_file_flush(log_file);
+	os_file_flush(log_file, TRUE);
 	os_file_close(log_file);
 
 	ut_free(buf);

mtr/mtr0mtr.c

@@ -265,9 +265,20 @@ mtr_commit(
 	mtr_memo_pop_all(mtr);
 #endif /* !UNIV_HOTBACKUP */
 
-	ut_d(mtr->state = MTR_COMMITTED);
 	dyn_array_free(&(mtr->memo));
 	dyn_array_free(&(mtr->log));
+#ifdef UNIV_DEBUG_VALGRIND
+	/* Declare everything uninitialized except
+	mtr->start_lsn, mtr->end_lsn and mtr->state. */
+	{
+		ib_uint64_t	start_lsn	= mtr->start_lsn;
+		ib_uint64_t	end_lsn		= mtr->end_lsn;
+		UNIV_MEM_INVALID(mtr, sizeof *mtr);
+		mtr->start_lsn = start_lsn;
+		mtr->end_lsn = end_lsn;
+	}
+#endif /* UNIV_DEBUG_VALGRIND */
+	ut_d(mtr->state = MTR_COMMITTED);
 }
 
 #ifndef UNIV_HOTBACKUP

os/os0file.c

@@ -2017,7 +2017,7 @@ os_file_set_size(
 
 	ut_free(buf2);
 
-	ret = os_file_flush(file);
+	ret = os_file_flush(file, TRUE);
 
 	if (ret) {
 		return(TRUE);
@@ -2055,7 +2055,8 @@ static
 int
 os_file_fsync(
 /*==========*/
-	os_file_t	file)	/*!< in: handle to a file */
+	os_file_t	file,	/*!< in: handle to a file */
+	ibool		metadata)
 {
 	int	ret;
 	int	failures;
@@ -2064,7 +2065,16 @@ os_file_fsync(
 	failures = 0;
 
 	do {
+#if defined(HAVE_FDATASYNC) && HAVE_DECL_FDATASYNC
+		if (metadata) {
+			ret = fsync(file);
+		} else {
+			ret = fdatasync(file);
+		}
+#else
+		(void) metadata;
 		ret = fsync(file);
+#endif
 
 		os_n_fsyncs++;
 
@@ -2104,7 +2114,8 @@ UNIV_INTERN
 ibool
 os_file_flush_func(
 /*===============*/
-	os_file_t	file)	/*!< in, own: handle to a file */
+	os_file_t	file,	/*!< in, own: handle to a file */
+	ibool		metadata)
 {
 #ifdef __WIN__
 	BOOL	ret;
@@ -2154,18 +2165,18 @@ os_file_flush_func(
 		/* If we are not on an operating system that supports this,
 		then fall back to a plain fsync. */
 
-		ret = os_file_fsync(file);
+		ret = os_file_fsync(file, metadata);
 	} else {
 		ret = fcntl(file, F_FULLFSYNC, NULL);
 
 		if (ret) {
 			/* If we are not on a file system that supports this,
 			then fall back to a plain fsync. */
-			ret = os_file_fsync(file);
+			ret = os_file_fsync(file, metadata);
 		}
 	}
 #else
-	ret = os_file_fsync(file);
+	ret = os_file_fsync(file, metadata);
 #endif
 
 	if (ret == 0) {
@@ -2411,7 +2422,7 @@ os_file_pwrite(
 		the OS crashes, a database page is only partially
 		physically written to disk. */
 
-		ut_a(TRUE == os_file_flush(file));
+		ut_a(TRUE == os_file_flush(file, TRUE));
 	}
 # endif /* UNIV_DO_FLUSH */
 
@@ -2463,7 +2474,7 @@ os_file_pwrite(
 			the OS crashes, a database page is only partially
 			physically written to disk. */
 
-			ut_a(TRUE == os_file_flush(file));
+			ut_a(TRUE == os_file_flush(file, TRUE));
 		}
 # endif /* UNIV_DO_FLUSH */
 
@@ -2836,7 +2847,7 @@ retry:
 
 # ifdef UNIV_DO_FLUSH
 	if (!os_do_not_call_flush_at_each_write) {
-		ut_a(TRUE == os_file_flush(file));
+		ut_a(TRUE == os_file_flush(file, TRUE));
 	}
 # endif /* UNIV_DO_FLUSH */
 
@@ -4141,7 +4152,13 @@ os_aio_func(
 		Windows async i/o, Windows does not allow us to use
 		ordinary synchronous os_file_read etc. on the same file,
 		therefore we have built a special mechanism for synchronous
-		wait in the Windows case. */
+		wait in the Windows case.
+		Also note that the Performance Schema instrumentation has
+		been performed by current os_aio_func()'s wrapper function
+		pfs_os_aio_func(). So we would no longer need to call
+		Performance Schema instrumented os_file_read() and
+		os_file_write(). Instead, we should use os_file_read_func()
+		and os_file_write_func() */
 
 		if (type == OS_FILE_READ) {
 			return(os_file_read_trx(file, buf, offset,
@@ -4150,7 +4167,8 @@ os_aio_func(
 
 		ut_a(type == OS_FILE_WRITE);
 
-		return(os_file_write(name, file, buf, offset, offset_high, n));
+		return(os_file_write_func(name, file, buf, offset,
+					  offset_high, n));
 	}
 
 try_again:
@@ -4398,7 +4416,7 @@ os_aio_windows_handle(
 #ifdef UNIV_DO_FLUSH
 		if (slot->type == OS_FILE_WRITE
 		    && !os_do_not_call_flush_at_each_write) {
-			if (!os_file_flush(slot->file)) {
+			if (!os_file_flush(slot->file, TRUE)) {
 				ut_error;
 			}
 		}
@@ -4701,7 +4719,7 @@ found:
 #ifdef UNIV_DO_FLUSH
 		if (slot->type == OS_FILE_WRITE
 		    && !os_do_not_call_flush_at_each_write)
-		    && !os_file_flush(slot->file) {
+		    && !os_file_flush(slot->file, TRUE) {
 			ut_error;
 		}
 #endif /* UNIV_DO_FLUSH */

page/page0page.c

@@ -1253,28 +1253,6 @@ page_move_rec_list_start(
 
 	return(TRUE);
 }
-
-/***********************************************************************//**
-This is a low-level operation which is used in a database index creation
-to update the page number of a created B-tree to a data dictionary record. */
-UNIV_INTERN
-void
-page_rec_write_index_page_no(
-/*=========================*/
-	rec_t*	rec,	/*!< in: record to update */
-	ulint	i,	/*!< in: index of the field to update */
-	ulint	page_no,/*!< in: value to write */
-	mtr_t*	mtr)	/*!< in: mtr */
-{
-	byte*	data;
-	ulint	len;
-
-	data = rec_get_nth_field_old(rec, i, &len);
-
-	ut_ad(len == 4);
-
-	mlog_write_ulint(data, page_no, MLOG_4BYTES, mtr);
-}
 #endif /* !UNIV_HOTBACKUP */
 
 /**************************************************************//**

page/page0zip.c

@@ -464,7 +464,7 @@ page_zip_fields_encode(
 
 			if (fixed_sum && UNIV_UNLIKELY
 			    (fixed_sum + field->fixed_len
-			     > DICT_MAX_INDEX_COL_LEN)) {
+			     > DICT_MAX_FIXED_COL_LEN)) {
 				/* Write out the length of the
 				preceding non-nullable fields,
 				to avoid exceeding the maximum

percona-suite/percona_mysqldump_innodb_optimize_keys.result

@@ -1,109 +0,0 @@
-#
-# Test the --innodb-optimize-keys option.
-#
-CREATE TABLE t1 (a INT NOT NULL PRIMARY KEY, b INT, KEY(b)) ENGINE=MyISAM;
-######################################
-
-/*!40101 SET @OLD_CHARACTER_SET_CLIENT=@@CHARACTER_SET_CLIENT */;
-/*!40101 SET @OLD_CHARACTER_SET_RESULTS=@@CHARACTER_SET_RESULTS */;
-/*!40101 SET @OLD_COLLATION_CONNECTION=@@COLLATION_CONNECTION */;
-/*!40101 SET NAMES utf8 */;
-/*!40103 SET @OLD_TIME_ZONE=@@TIME_ZONE */;
-/*!40103 SET TIME_ZONE='+00:00' */;
-/*!40014 SET @OLD_UNIQUE_CHECKS=@@UNIQUE_CHECKS, UNIQUE_CHECKS=0 */;
-/*!40014 SET @OLD_FOREIGN_KEY_CHECKS=@@FOREIGN_KEY_CHECKS, FOREIGN_KEY_CHECKS=0 */;
-/*!40101 SET @OLD_SQL_MODE=@@SQL_MODE, SQL_MODE='NO_AUTO_VALUE_ON_ZERO' */;
-/*!40111 SET @OLD_SQL_NOTES=@@SQL_NOTES, SQL_NOTES=0 */;
-DROP TABLE IF EXISTS `t1`;
-/*!40101 SET @saved_cs_client     = @@character_set_client */;
-/*!40101 SET character_set_client = utf8 */;
-CREATE TABLE `t1` (
-  `a` int(11) NOT NULL,
-  `b` int(11) DEFAULT NULL,
-  PRIMARY KEY (`a`),
-  KEY `b` (`b`)
-) ENGINE=MyISAM DEFAULT CHARSET=latin1;
-/*!40101 SET character_set_client = @saved_cs_client */;
-
-LOCK TABLES `t1` WRITE;
-/*!40000 ALTER TABLE `t1` DISABLE KEYS */;
-/*!40000 ALTER TABLE `t1` ENABLE KEYS */;
-UNLOCK TABLES;
-/*!40103 SET TIME_ZONE=@OLD_TIME_ZONE */;
-
-/*!40101 SET SQL_MODE=@OLD_SQL_MODE */;
-/*!40014 SET FOREIGN_KEY_CHECKS=@OLD_FOREIGN_KEY_CHECKS */;
-/*!40014 SET UNIQUE_CHECKS=@OLD_UNIQUE_CHECKS */;
-/*!40101 SET CHARACTER_SET_CLIENT=@OLD_CHARACTER_SET_CLIENT */;
-/*!40101 SET CHARACTER_SET_RESULTS=@OLD_CHARACTER_SET_RESULTS */;
-/*!40101 SET COLLATION_CONNECTION=@OLD_COLLATION_CONNECTION */;
-/*!40111 SET SQL_NOTES=@OLD_SQL_NOTES */;
-
-######################################
-DROP TABLE t1;
-CREATE TABLE t2 (a INT PRIMARY KEY) ENGINE=InnoDB;
-INSERT INTO t2 VALUES (0), (1), (2);
-CREATE TABLE t1 (
-id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
-a INT, b VARCHAR(255), c DECIMAL(10,3),
-KEY (b),
-UNIQUE KEY uniq(c,a),
-FOREIGN KEY (a) REFERENCES t2(a) ON DELETE CASCADE
-) ENGINE=InnoDB;
-INSERT INTO t1(a,b,c) VALUES (0, "0", 0.0), (1, "1", 1.1), (2, "2", 2.2);
-######################################
-
-/*!40101 SET @OLD_CHARACTER_SET_CLIENT=@@CHARACTER_SET_CLIENT */;
-/*!40101 SET @OLD_CHARACTER_SET_RESULTS=@@CHARACTER_SET_RESULTS */;
-/*!40101 SET @OLD_COLLATION_CONNECTION=@@COLLATION_CONNECTION */;
-/*!40101 SET NAMES utf8 */;
-/*!40103 SET @OLD_TIME_ZONE=@@TIME_ZONE */;
-/*!40103 SET TIME_ZONE='+00:00' */;
-/*!40014 SET @OLD_UNIQUE_CHECKS=@@UNIQUE_CHECKS, UNIQUE_CHECKS=0 */;
-/*!40014 SET @OLD_FOREIGN_KEY_CHECKS=@@FOREIGN_KEY_CHECKS, FOREIGN_KEY_CHECKS=0 */;
-/*!40101 SET @OLD_SQL_MODE=@@SQL_MODE, SQL_MODE='NO_AUTO_VALUE_ON_ZERO' */;
-/*!40111 SET @OLD_SQL_NOTES=@@SQL_NOTES, SQL_NOTES=0 */;
-DROP TABLE IF EXISTS `t1`;
-/*!40101 SET @saved_cs_client     = @@character_set_client */;
-/*!40101 SET character_set_client = utf8 */;
-CREATE TABLE `t1` (
-  `id` int(11) NOT NULL AUTO_INCREMENT,
-  `a` int(11) DEFAULT NULL,
-  `b` varchar(255) DEFAULT NULL,
-  `c` decimal(10,3) DEFAULT NULL,
-  PRIMARY KEY (`id`) 
-) ENGINE=InnoDB AUTO_INCREMENT=4 DEFAULT CHARSET=latin1;
-/*!40101 SET character_set_client = @saved_cs_client */;
-
-LOCK TABLES `t1` WRITE;
-/*!40000 ALTER TABLE `t1` DISABLE KEYS */;
-INSERT INTO `t1` VALUES (1,0,'0',0.000),(2,1,'1',1.100),(3,2,'2',2.200);
-ALTER TABLE `t1` ADD UNIQUE KEY `uniq` (`c`,`a`), ADD KEY `b` (`b`), ADD KEY `a` (`a`), ADD CONSTRAINT `t1_ibfk_1` FOREIGN KEY (`a`) REFERENCES `t2` (`a`) ON DELETE CASCADE;
-/*!40000 ALTER TABLE `t1` ENABLE KEYS */;
-UNLOCK TABLES;
-DROP TABLE IF EXISTS `t2`;
-/*!40101 SET @saved_cs_client     = @@character_set_client */;
-/*!40101 SET character_set_client = utf8 */;
-CREATE TABLE `t2` (
-  `a` int(11) NOT NULL,
-  PRIMARY KEY (`a`)
-) ENGINE=InnoDB DEFAULT CHARSET=latin1;
-/*!40101 SET character_set_client = @saved_cs_client */;
-
-LOCK TABLES `t2` WRITE;
-/*!40000 ALTER TABLE `t2` DISABLE KEYS */;
-INSERT INTO `t2` VALUES (0),(1),(2);
-/*!40000 ALTER TABLE `t2` ENABLE KEYS */;
-UNLOCK TABLES;
-/*!40103 SET TIME_ZONE=@OLD_TIME_ZONE */;
-
-/*!40101 SET SQL_MODE=@OLD_SQL_MODE */;
-/*!40014 SET FOREIGN_KEY_CHECKS=@OLD_FOREIGN_KEY_CHECKS */;
-/*!40014 SET UNIQUE_CHECKS=@OLD_UNIQUE_CHECKS */;
-/*!40101 SET CHARACTER_SET_CLIENT=@OLD_CHARACTER_SET_CLIENT */;
-/*!40101 SET CHARACTER_SET_RESULTS=@OLD_CHARACTER_SET_RESULTS */;
-/*!40101 SET COLLATION_CONNECTION=@OLD_COLLATION_CONNECTION */;
-/*!40111 SET SQL_NOTES=@OLD_SQL_NOTES */;
-
-######################################
-DROP TABLE t1, t2;

percona-suite/percona_mysqldump_innodb_optimize_keys.test

@@ -1,62 +0,0 @@
-# Embedded server doesn't support external clients
---source include/not_embedded.inc
-
-# Fast index creation is only available in InnoDB plugin
---source include/have_innodb.inc
-
-# Save the initial number of concurrent sessions
---source include/count_sessions.inc
-
---echo #
---echo # Test the --innodb-optimize-keys option.
---echo #
-
---let $file=$MYSQLTEST_VARDIR/tmp/t1.sql
-
-# First test that the option has no effect on non-InnoDB tables
-
-CREATE TABLE t1 (a INT NOT NULL PRIMARY KEY, b INT, KEY(b)) ENGINE=MyISAM;
-
---exec $MYSQL_DUMP --skip-comments --innodb-optimize-keys test t1 >$file
-
---echo ######################################
---cat_file $file
---echo ######################################
-
---remove_file $file
-
-DROP TABLE t1;
-
-
-# Check that for InnoDB tables secondary and foreign keys are created
-# after the data is dumped
-
-CREATE TABLE t2 (a INT PRIMARY KEY) ENGINE=InnoDB;
-INSERT INTO t2 VALUES (0), (1), (2);
-
-CREATE TABLE t1 (
-  id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
-  a INT, b VARCHAR(255), c DECIMAL(10,3),
-  KEY (b),
-  UNIQUE KEY uniq(c,a),
-  FOREIGN KEY (a) REFERENCES t2(a) ON DELETE CASCADE
-) ENGINE=InnoDB;
-
-INSERT INTO t1(a,b,c) VALUES (0, "0", 0.0), (1, "1", 1.1), (2, "2", 2.2);
-
---exec $MYSQL_DUMP --skip-comments --innodb-optimize-keys test t1 t2 >$file
-
---echo ######################################
---cat_file $file
---echo ######################################
-
-# Check that the resulting dump can be imported back
-
---exec $MYSQL test < $file
-
---remove_file $file
-
-DROP TABLE t1, t2;
-
-# Wait till we reached the initial number of concurrent sessions
---source include/wait_until_count_sessions.inc

percona-suite/percona_query_cache_with_comments.inc.backup

@@ -0,0 +1,88 @@
+--source include/percona_query_cache_with_comments_clear.inc
+let $query=/* with comment first */select * from t1;
+eval $query;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=# with comment first
+select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=-- with comment first
+select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=/* with comment first and "quote" */select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=# with comment first and "quote"
+select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=-- with comment first and "quote" 
+select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=
+    /* with comment and whitespaces first */select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query= 
+    # with comment and whitespaces first
+select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=
+    -- with comment and whitespaces first
+select * from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $internal=* internal comment *;
+
+let $query=select * /$internal/ from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+let $query=select */$internal/ from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+let $query=select */$internal/from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $internal=* internal comment with "quote" *;
+
+let $query=select * /$internal/ from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+let $query=select */$internal/ from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+let $query=select */$internal/from t1;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1
+;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1 ;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1	;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1
+/* comment in the end */;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1
+/* comment in the end */
+;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1 #comment in the end;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1 #comment in the end
+;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1 -- comment in the end;
+--source include/percona_query_cache_with_comments_eval.inc
+
+let $query=select * from t1 -- comment in the end
+;
+--source include/percona_query_cache_with_comments_eval.inc

percona-suite/percona_query_response_time-replication.result

@@ -0,0 +1,66 @@
+include/master-slave.inc
+[connection master]
+DROP TABLE IF EXISTS t;
+CREATE TABLE t(id INT);
+SELECT * from t;
+id
+SELECT * from t;
+id
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1;
+Warnings:
+Warning	1292	Truncated incorrect query_response_time_range_base value: '1'
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	10
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=ON;
+INSERT INTO t SELECT SLEEP(0.4);
+Warnings:
+Note	1592	Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT. Statement is unsafe because it uses a system function that may return a different value on the slave.
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+0
+INSERT INTO t SELECT SLEEP(0.4);
+Warnings:
+Note	1592	Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT. Statement is unsafe because it uses a system function that may return a different value on the slave.
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+0
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+2
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+3
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 2;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+FLUSH QUERY_RESPONSE_TIME;
+INSERT INTO t SELECT SLEEP(0.4);
+Warnings:
+Note	1592	Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT. Statement is unsafe because it uses a system function that may return a different value on the slave.
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+0
+INSERT INTO t SELECT SLEEP(0.4);
+Warnings:
+Note	1592	Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT. Statement is unsafe because it uses a system function that may return a different value on the slave.
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+0
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+2
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count)
+3
+DROP TABLE IF EXISTS t;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=OFF;
+STOP SLAVE;
+include/wait_for_slave_to_stop.inc

percona-suite/percona_query_response_time-replication.test

@@ -0,0 +1,61 @@
+--source include/have_response_time_distribution.inc
+--source include/master-slave.inc
+--source include/have_binlog_format_statement.inc
+--source include/have_debug.inc
+--disable_query_log
+call mtr.add_suppression("Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT. Statement is unsafe because it uses a system function that may return a different value on the slave. Statement:");
+--enable_query_log
+
+connection master;
+-- disable_warnings
+DROP TABLE IF EXISTS t;
+-- enable_warnings
+CREATE TABLE t(id INT);
+SELECT * from t;
+
+sync_slave_with_master;
+
+connection slave;
+SELECT * from t;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=ON;
+
+connection master;
+INSERT INTO t SELECT SLEEP(0.4);
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+INSERT INTO t SELECT SLEEP(0.4);
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+sync_slave_with_master;
+
+connection slave;
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 2;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+source include/percona_query_response_time_flush.inc;
+
+connection master;
+INSERT INTO t SELECT SLEEP(0.4);
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+INSERT INTO t SELECT SLEEP(0.4);
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+sync_slave_with_master;
+
+connection slave;
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SELECT SUM(INFORMATION_SCHEMA.QUERY_RESPONSE_TIME.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+
+connection master;
+DROP TABLE IF EXISTS t;
+sync_slave_with_master;
+connection slave;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=OFF;
+
+STOP SLAVE;
+-- source include/wait_for_slave_to_stop.inc

percona-suite/percona_query_response_time-stored.result

@@ -0,0 +1,306 @@
+SET GLOBAL debug="d,query_exec_time_debug";
+CREATE FUNCTION test_f()
+RETURNS CHAR(30) DETERMINISTIC
+BEGIN
+SET SESSION debug="+d,query_exec_time_1.1";
+RETURN 'Hello, world!';
+END/
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1;
+Warnings:
+Warning	1292	Truncated incorrect query_response_time_range_base value: '1'
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 2;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+FLUSH QUERY_RESPONSE_TIME;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+44
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000003
+      0.000007
+      0.000015
+      0.000030
+      0.000061
+      0.000122
+      0.000244
+      0.000488
+      0.000976
+      0.001953
+      0.003906
+      0.007812
+      0.015625
+      0.031250
+      0.062500
+      0.125000
+      0.250000
+      0.500000
+      1.000000
+      2.000000
+      4.000000
+      8.000000
+     16.000000
+     32.000000
+     64.000000
+    128.000000
+    256.000000
+    512.000000
+   1024.000000
+   2048.000000
+   4096.000000
+   8192.000000
+  16384.000000
+  32768.000000
+  65536.000000
+ 131072.000000
+ 262144.000000
+ 524288.000000
+ 1048576.00000
+ 2097152.00000
+ 4194304.00000
+ 8388608.00000
+TOO LONG
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+test_f()
+Hello, world!
+SELECT test_f();
+test_f()
+Hello, world!
+SELECT test_f();
+test_f()
+Hello, world!
+SELECT test_f();
+test_f()
+Hello, world!
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+4	4	4	1	44
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+44
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000003
+      0.000007
+      0.000015
+      0.000030
+      0.000061
+      0.000122
+      0.000244
+      0.000488
+      0.000976
+      0.001953
+      0.003906
+      0.007812
+      0.015625
+      0.031250
+      0.062500
+      0.125000
+      0.250000
+      0.500000
+      1.000000
+      2.000000
+      4.000000
+      8.000000
+     16.000000
+     32.000000
+     64.000000
+    128.000000
+    256.000000
+    512.000000
+   1024.000000
+   2048.000000
+   4096.000000
+   8192.000000
+  16384.000000
+  32768.000000
+  65536.000000
+ 131072.000000
+ 262144.000000
+ 524288.000000
+ 1048576.00000
+ 2097152.00000
+ 4194304.00000
+ 8388608.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	10
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+test_f()
+Hello, world!
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+1	1	1	1	14
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+14
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000010
+      0.000100
+      0.001000
+      0.010000
+      0.100000
+      1.000000
+     10.000000
+    100.000000
+   1000.000000
+  10000.000000
+ 100000.000000
+ 1000000.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	10
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 7;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	7
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+test_f()
+Hello, world!
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+1	1	1	1	17
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+17
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000008
+      0.000059
+      0.000416
+      0.002915
+      0.020408
+      0.142857
+      1.000000
+      7.000000
+     49.000000
+    343.000000
+   2401.000000
+  16807.000000
+ 117649.000000
+ 823543.000000
+ 5764801.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	7
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 156;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	156
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+test_f()
+Hello, world!
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+1	1	1	1	7
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+7
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000041
+      0.006410
+      1.000000
+    156.000000
+  24336.000000
+ 3796416.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	156
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1000;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	1000
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+test_f()
+Hello, world!
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+1	1	1	1	6
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+6
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.001000
+      1.000000
+   1000.000000
+ 1000000.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	1000
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1001;
+Warnings:
+Warning	1292	Truncated incorrect query_response_time_range_base value: '1001'
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	1000
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE  =10;
+DROP FUNCTION test_f;
+SET GLOBAL debug=default;

percona-suite/percona_query_response_time-stored.test

@@ -0,0 +1,85 @@
+--source include/have_response_time_distribution.inc
+--source include/have_debug.inc
+
+SET GLOBAL debug="d,query_exec_time_debug";
+
+delimiter /;
+CREATE FUNCTION test_f()
+RETURNS CHAR(30) DETERMINISTIC
+BEGIN
+	SET SESSION debug="+d,query_exec_time_1.1";
+	RETURN 'Hello, world!';
+END/
+delimiter ;/
+
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 2;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+source include/percona_query_response_time_show.inc;
+
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+SELECT test_f();
+SELECT test_f();
+SELECT test_f();
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 7;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 156;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1000;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SELECT test_f();
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1001;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE  =10;
+
+DROP FUNCTION test_f;
+
+SET GLOBAL debug=default;

percona-suite/percona_query_response_time.result

@@ -0,0 +1,377 @@
+SET GLOBAL debug="d,query_exec_time_debug";
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1;
+Warnings:
+Warning	1292	Truncated incorrect query_response_time_range_base value: '1'
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 2;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+FLUSH QUERY_RESPONSE_TIME;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+44
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000003
+      0.000007
+      0.000015
+      0.000030
+      0.000061
+      0.000122
+      0.000244
+      0.000488
+      0.000976
+      0.001953
+      0.003906
+      0.007812
+      0.015625
+      0.031250
+      0.062500
+      0.125000
+      0.250000
+      0.500000
+      1.000000
+      2.000000
+      4.000000
+      8.000000
+     16.000000
+     32.000000
+     64.000000
+    128.000000
+    256.000000
+    512.000000
+   1024.000000
+   2048.000000
+   4096.000000
+   8192.000000
+  16384.000000
+  32768.000000
+  65536.000000
+ 131072.000000
+ 262144.000000
+ 524288.000000
+ 1048576.00000
+ 2097152.00000
+ 4194304.00000
+ 8388608.00000
+TOO LONG
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SET SESSION debug="+d,query_exec_time_0.31";
+SET SESSION debug="+d,query_exec_time_0.32";
+SET SESSION debug="+d,query_exec_time_0.33";
+SET SESSION debug="+d,query_exec_time_0.34";
+SET SESSION debug="+d,query_exec_time_0.35";
+SET SESSION debug="+d,query_exec_time_0.36";
+SET SESSION debug="+d,query_exec_time_0.37";
+SET SESSION debug="+d,query_exec_time_0.38";
+SET SESSION debug="+d,query_exec_time_0.39";
+SET SESSION debug="+d,query_exec_time_0.4";
+SET SESSION debug="+d,query_exec_time_1.1";
+SET SESSION debug="+d,query_exec_time_1.2";
+SET SESSION debug="+d,query_exec_time_1.3";
+SET SESSION debug="+d,query_exec_time_1.5";
+SET SESSION debug="+d,query_exec_time_1.4";
+SET SESSION debug="+d,query_exec_time_0.5";
+SET SESSION debug="+d,query_exec_time_2.1";
+SET SESSION debug="+d,query_exec_time_2.3";
+SET SESSION debug="+d,query_exec_time_2.5";
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+10	19	15	4	44
+1	19	15	4	44
+5	19	15	4	44
+3	19	15	4	44
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+44
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000003
+      0.000007
+      0.000015
+      0.000030
+      0.000061
+      0.000122
+      0.000244
+      0.000488
+      0.000976
+      0.001953
+      0.003906
+      0.007812
+      0.015625
+      0.031250
+      0.062500
+      0.125000
+      0.250000
+      0.500000
+      1.000000
+      2.000000
+      4.000000
+      8.000000
+     16.000000
+     32.000000
+     64.000000
+    128.000000
+    256.000000
+    512.000000
+   1024.000000
+   2048.000000
+   4096.000000
+   8192.000000
+  16384.000000
+  32768.000000
+  65536.000000
+ 131072.000000
+ 262144.000000
+ 524288.000000
+ 1048576.00000
+ 2097152.00000
+ 4194304.00000
+ 8388608.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	2
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	10
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SET SESSION debug="+d,query_exec_time_0.31";
+SET SESSION debug="+d,query_exec_time_0.32";
+SET SESSION debug="+d,query_exec_time_0.33";
+SET SESSION debug="+d,query_exec_time_0.34";
+SET SESSION debug="+d,query_exec_time_0.35";
+SET SESSION debug="+d,query_exec_time_0.36";
+SET SESSION debug="+d,query_exec_time_0.37";
+SET SESSION debug="+d,query_exec_time_0.38";
+SET SESSION debug="+d,query_exec_time_0.39";
+SET SESSION debug="+d,query_exec_time_0.4";
+SET SESSION debug="+d,query_exec_time_1.1";
+SET SESSION debug="+d,query_exec_time_1.2";
+SET SESSION debug="+d,query_exec_time_1.3";
+SET SESSION debug="+d,query_exec_time_1.5";
+SET SESSION debug="+d,query_exec_time_1.4";
+SET SESSION debug="+d,query_exec_time_0.5";
+SET SESSION debug="+d,query_exec_time_2.1";
+SET SESSION debug="+d,query_exec_time_2.3";
+SET SESSION debug="+d,query_exec_time_2.5";
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+11	19	17	2	14
+8	19	17	2	14
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+14
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000010
+      0.000100
+      0.001000
+      0.010000
+      0.100000
+      1.000000
+     10.000000
+    100.000000
+   1000.000000
+  10000.000000
+ 100000.000000
+ 1000000.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	10
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 7;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	7
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SET SESSION debug="+d,query_exec_time_0.31";
+SET SESSION debug="+d,query_exec_time_0.32";
+SET SESSION debug="+d,query_exec_time_0.33";
+SET SESSION debug="+d,query_exec_time_0.34";
+SET SESSION debug="+d,query_exec_time_0.35";
+SET SESSION debug="+d,query_exec_time_0.36";
+SET SESSION debug="+d,query_exec_time_0.37";
+SET SESSION debug="+d,query_exec_time_0.38";
+SET SESSION debug="+d,query_exec_time_0.39";
+SET SESSION debug="+d,query_exec_time_0.4";
+SET SESSION debug="+d,query_exec_time_1.1";
+SET SESSION debug="+d,query_exec_time_1.2";
+SET SESSION debug="+d,query_exec_time_1.3";
+SET SESSION debug="+d,query_exec_time_1.5";
+SET SESSION debug="+d,query_exec_time_1.4";
+SET SESSION debug="+d,query_exec_time_0.5";
+SET SESSION debug="+d,query_exec_time_2.1";
+SET SESSION debug="+d,query_exec_time_2.3";
+SET SESSION debug="+d,query_exec_time_2.5";
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+11	19	17	2	17
+8	19	17	2	17
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+17
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.000008
+      0.000059
+      0.000416
+      0.002915
+      0.020408
+      0.142857
+      1.000000
+      7.000000
+     49.000000
+    343.000000
+   2401.000000
+  16807.000000
+ 117649.000000
+ 823543.000000
+ 5764801.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	7
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 156;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	156
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SET SESSION debug="+d,query_exec_time_0.31";
+SET SESSION debug="+d,query_exec_time_0.32";
+SET SESSION debug="+d,query_exec_time_0.33";
+SET SESSION debug="+d,query_exec_time_0.34";
+SET SESSION debug="+d,query_exec_time_0.35";
+SET SESSION debug="+d,query_exec_time_0.36";
+SET SESSION debug="+d,query_exec_time_0.37";
+SET SESSION debug="+d,query_exec_time_0.38";
+SET SESSION debug="+d,query_exec_time_0.39";
+SET SESSION debug="+d,query_exec_time_0.4";
+SET SESSION debug="+d,query_exec_time_1.1";
+SET SESSION debug="+d,query_exec_time_1.2";
+SET SESSION debug="+d,query_exec_time_1.3";
+SET SESSION debug="+d,query_exec_time_1.5";
+SET SESSION debug="+d,query_exec_time_1.4";
+SET SESSION debug="+d,query_exec_time_0.5";
+SET SESSION debug="+d,query_exec_time_2.1";
+SET SESSION debug="+d,query_exec_time_2.3";
+SET SESSION debug="+d,query_exec_time_2.5";
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+11	19	17	2	7
+8	19	17	2	7
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+7
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000041
+      0.006410
+      1.000000
+    156.000000
+  24336.000000
+ 3796416.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	156
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1000;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	1000
+FLUSH QUERY_RESPONSE_TIME;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+SET SESSION debug="+d,query_exec_time_0.31";
+SET SESSION debug="+d,query_exec_time_0.32";
+SET SESSION debug="+d,query_exec_time_0.33";
+SET SESSION debug="+d,query_exec_time_0.34";
+SET SESSION debug="+d,query_exec_time_0.35";
+SET SESSION debug="+d,query_exec_time_0.36";
+SET SESSION debug="+d,query_exec_time_0.37";
+SET SESSION debug="+d,query_exec_time_0.38";
+SET SESSION debug="+d,query_exec_time_0.39";
+SET SESSION debug="+d,query_exec_time_0.4";
+SET SESSION debug="+d,query_exec_time_1.1";
+SET SESSION debug="+d,query_exec_time_1.2";
+SET SESSION debug="+d,query_exec_time_1.3";
+SET SESSION debug="+d,query_exec_time_1.5";
+SET SESSION debug="+d,query_exec_time_1.4";
+SET SESSION debug="+d,query_exec_time_0.5";
+SET SESSION debug="+d,query_exec_time_2.1";
+SET SESSION debug="+d,query_exec_time_2.3";
+SET SESSION debug="+d,query_exec_time_2.5";
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+count	query_count	query_total	not_zero_region_count	region_count
+11	19	17	2	6
+8	19	17	2	6
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+region_count
+6
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+time
+      0.000001
+      0.001000
+      1.000000
+   1000.000000
+ 1000000.00000
+TOO LONG
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	1000
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1001;
+Warnings:
+Warning	1292	Truncated incorrect query_response_time_range_base value: '1001'
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+Variable_name	Value
+query_response_time_range_base	1000
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE  =10;
+SET GLOBAL debug=default;

percona-suite/percona_query_response_time.test

@@ -0,0 +1,71 @@
+--source include/have_response_time_distribution.inc
+--source include/have_debug.inc
+
+SET GLOBAL debug="d,query_exec_time_debug";
+
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 2;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+source include/percona_query_response_time_show.inc;
+
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+source include/percona_query_response_time_sleep.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 10;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+source include/percona_query_response_time_sleep.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 7;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+source include/percona_query_response_time_sleep.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 156;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+source include/percona_query_response_time_sleep.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1000;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+source include/percona_query_response_time_flush.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=1;
+source include/percona_query_response_time_sleep.inc;
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+
+source include/percona_query_response_time_show.inc;
+
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE = 1001;
+SHOW GLOBAL VARIABLES where Variable_name like 'QUERY_RESPONSE_TIME_RANGE_BASE';
+
+SET GLOBAL QUERY_RESPONSE_TIME_STATS=0;
+SET GLOBAL QUERY_RESPONSE_TIME_RANGE_BASE  =10;
+
+SET GLOBAL debug=default;

percona-suite/percona_query_response_time_flush.inc

@@ -0,0 +1 @@
+FLUSH QUERY_RESPONSE_TIME;

percona-suite/percona_query_response_time_show.inc

@@ -0,0 +1,8 @@
+SELECT d.count,
+(SELECT SUM(a.count) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as a WHERE a.count != 0) as query_count,
+(SELECT SUM((b.total * 1000000) DIV 1000000) FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as b WHERE b.count != 0) as query_total,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as c WHERE c.count != 0) as not_zero_region_count,
+(SELECT COUNT(*)     FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME) as region_count
+FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME as d WHERE d.count > 0;
+SELECT COUNT(*) as region_count FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;
+SELECT time FROM INFORMATION_SCHEMA.QUERY_RESPONSE_TIME;

percona-suite/percona_query_response_time_sleep.inc

@@ -0,0 +1,19 @@
+SET SESSION debug="+d,query_exec_time_0.31";
+SET SESSION debug="+d,query_exec_time_0.32";
+SET SESSION debug="+d,query_exec_time_0.33";
+SET SESSION debug="+d,query_exec_time_0.34";
+SET SESSION debug="+d,query_exec_time_0.35";
+SET SESSION debug="+d,query_exec_time_0.36";
+SET SESSION debug="+d,query_exec_time_0.37";
+SET SESSION debug="+d,query_exec_time_0.38";
+SET SESSION debug="+d,query_exec_time_0.39";
+SET SESSION debug="+d,query_exec_time_0.4";
+SET SESSION debug="+d,query_exec_time_1.1";
+SET SESSION debug="+d,query_exec_time_1.2";
+SET SESSION debug="+d,query_exec_time_1.3";
+SET SESSION debug="+d,query_exec_time_1.5";
+SET SESSION debug="+d,query_exec_time_1.4";
+SET SESSION debug="+d,query_exec_time_0.5";
+SET SESSION debug="+d,query_exec_time_2.1";
+SET SESSION debug="+d,query_exec_time_2.3";
+SET SESSION debug="+d,query_exec_time_2.5";

rem/rem0rec.c

@@ -1174,7 +1174,7 @@ rec_convert_dtuple_to_rec_comp(
 		} else if (dfield_is_ext(field)) {
 			ut_ad(ifield->col->len >= 256
 			      || ifield->col->mtype == DATA_BLOB);
-			ut_ad(len <= REC_MAX_INDEX_COL_LEN
+			ut_ad(len <= REC_ANTELOPE_MAX_INDEX_COL_LEN
 			      + BTR_EXTERN_FIELD_REF_SIZE);
 			*lens-- = (byte) (len >> 8) | 0xc0;
 			*lens-- = (byte) len;

row/row0ext.c

@@ -44,8 +44,9 @@ row_ext_cache_fill(
 {
 	const byte*	field	= dfield_get_data(dfield);
 	ulint		f_len	= dfield_get_len(dfield);
-	byte*		buf	= ext->buf + i * REC_MAX_INDEX_COL_LEN;
+	byte*		buf	= ext->buf + i * ext->max_len;
 
+	ut_ad(ext->max_len > 0);
 	ut_ad(i < ext->n_ext);
 	ut_ad(dfield_is_ext(dfield));
 	ut_a(f_len >= BTR_EXTERN_FIELD_REF_SIZE);
@@ -56,14 +57,14 @@ row_ext_cache_fill(
 		/* The BLOB pointer is not set: we cannot fetch it */
 		ext->len[i] = 0;
 	} else {
-		/* Fetch at most REC_MAX_INDEX_COL_LEN of the column.
+		/* Fetch at most ext->max_len of the column.
 		The column should be non-empty.  However,
 		trx_rollback_or_clean_all_recovered() may try to
 		access a half-deleted BLOB if the server previously
 		crashed during the execution of
 		btr_free_externally_stored_field(). */
 		ext->len[i] = btr_copy_externally_stored_field_prefix(
-			buf, REC_MAX_INDEX_COL_LEN, zip_size, field, f_len);
+			buf, ext->max_len, zip_size, field, f_len);
 	}
 }
 
@@ -79,16 +80,18 @@ row_ext_create(
 				in the InnoDB table object, as reported by
 				dict_col_get_no(); NOT relative to the records
 				in the clustered index */
+	ulint		flags,	/*!< in: table->flags */
 	const dtuple_t*	tuple,	/*!< in: data tuple containing the field
 				references of the externally stored
 				columns; must be indexed by col_no;
 				the clustered index record must be
 				covered by a lock or a page latch
 				to prevent deletion (rollback or purge). */
-	ulint		zip_size,/*!< compressed page size in bytes, or 0 */
 	mem_heap_t*	heap)	/*!< in: heap where created */
 {
 	ulint		i;
+	ulint		zip_size = dict_table_flags_to_zip_size(flags);
+
 	row_ext_t*	ret = mem_heap_alloc(heap, (sizeof *ret)
 					     + (n_ext - 1) * sizeof ret->len);
 
@@ -97,10 +100,12 @@ row_ext_create(
 
 	ret->n_ext = n_ext;
 	ret->ext = ext;
-	ret->buf = mem_heap_alloc(heap, n_ext * REC_MAX_INDEX_COL_LEN);
+	ret->max_len = DICT_MAX_FIELD_LEN_BY_FORMAT_FLAG(flags);
+
+	ret->buf = mem_heap_alloc(heap, n_ext * ret->max_len);
 #ifdef UNIV_DEBUG
-	memset(ret->buf, 0xaa, n_ext * REC_MAX_INDEX_COL_LEN);
-	UNIV_MEM_ALLOC(ret->buf, n_ext * REC_MAX_INDEX_COL_LEN);
+	memset(ret->buf, 0xaa, n_ext * ret->max_len);
+	UNIV_MEM_ALLOC(ret->buf, n_ext * ret->max_len);
 #endif
 
 	/* Fetch the BLOB prefixes */

row/row0mysql.c

@@ -2008,6 +2008,7 @@ row_create_index_for_mysql(
 	ulint		i;
 	ulint		len;
 	char*		table_name;
+	dict_table_t*	table;
 
 #ifdef UNIV_SYNC_DEBUG
 	ut_ad(rw_lock_own(&dict_operation_lock, RW_LOCK_EX));
@@ -2021,6 +2022,8 @@ row_create_index_for_mysql(
 	que_run_threads()) and thus index->table_name is not available. */
 	table_name = mem_strdup(index->table_name);
 
+	table = dict_table_get_low(table_name);
+
 	trx_start_if_not_started(trx);
 
 	/* Check that the same column does not appear twice in the index.
@@ -2053,7 +2056,7 @@ row_create_index_for_mysql(
 		}
 
 		/* Check also that prefix_len and actual length
-		< DICT_MAX_INDEX_COL_LEN */
+		is less than that from DICT_MAX_FIELD_LEN_BY_FORMAT() */
 
 		len = dict_index_get_nth_field(index, i)->prefix_len;
 
@@ -2061,8 +2064,9 @@ row_create_index_for_mysql(
 			len = ut_max(len, field_lengths[i]);
 		}
 
-		if (len >= DICT_MAX_INDEX_COL_LEN) {
-			err = DB_TOO_BIG_RECORD;
+		/* Column or prefix length exceeds maximum column length */
+		if (len > (ulint) DICT_MAX_FIELD_LEN_BY_FORMAT(table)) {
+			err = DB_TOO_BIG_INDEX_COL;
 
 			goto error_handling;
 		}
@@ -2087,6 +2091,7 @@ row_create_index_for_mysql(
 	que_graph_free((que_t*) que_node_get_parent(thr));
 
 error_handling:
+
 	if (err != DB_SUCCESS) {
 		/* We have special error handling here */
 
@@ -3027,7 +3032,7 @@ row_truncate_table_for_mysql(
 		rec = btr_pcur_get_rec(&pcur);
 
 		if (root_page_no != FIL_NULL) {
-			page_rec_write_index_page_no(
+			page_rec_write_field(
 				rec, DICT_SYS_INDEXES_PAGE_NO_FIELD,
 				root_page_no, &mtr);
 			/* We will need to commit and restart the

row/row0row.c

@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -47,35 +47,6 @@ Created 4/20/1996 Heikki Tuuri
 #include "read0read.h"
 #include "ut0mem.h"
 
-/*********************************************************************//**
-Gets the offset of trx id field, in bytes relative to the origin of
-a clustered index record.
-@return	offset of DATA_TRX_ID */
-UNIV_INTERN
-ulint
-row_get_trx_id_offset(
-/*==================*/
-	const rec_t*	rec __attribute__((unused)),
-				/*!< in: record */
-	dict_index_t*	index,	/*!< in: clustered index */
-	const ulint*	offsets)/*!< in: rec_get_offsets(rec, index) */
-{
-	ulint	pos;
-	ulint	offset;
-	ulint	len;
-
-	ut_ad(dict_index_is_clust(index));
-	ut_ad(rec_offs_validate(rec, index, offsets));
-
-	pos = dict_index_get_sys_col_pos(index, DATA_TRX_ID);
-
-	offset = rec_get_nth_field_offs(offsets, pos, &len);
-
-	ut_ad(len == DATA_TRX_ID_LEN);
-
-	return(offset);
-}
-
 /*****************************************************************//**
 When an insert or purge to a table is performed, this function builds
 the entry to be inserted into or purged from an index on the table.
@@ -151,8 +122,6 @@ row_build_index_entry(
 		} else if (dfield_is_ext(dfield)) {
 			ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);
 			len -= BTR_EXTERN_FIELD_REF_SIZE;
-			ut_a(ind_field->prefix_len <= len
-			     || dict_index_is_clust(index));
 		}
 
 		len = dtype_get_at_most_n_mbchars(
@@ -231,6 +200,14 @@ row_build(
 		ut_ad(rec_offs_validate(rec, index, offsets));
 	}
 
+#if 0 && defined UNIV_BLOB_NULL_DEBUG
+	/* This one can fail in trx_rollback_active() if
+	the server crashed during an insert before the
+	btr_store_big_rec_extern_fields() did mtr_commit()
+	all BLOB pointers to the clustered index record. */
+	ut_a(!rec_offs_any_null_extern(rec, offsets));
+#endif /* 0 && UNIV_BLOB_NULL_DEBUG */
+
 	if (type != ROW_COPY_POINTERS) {
 		/* Take a copy of rec to heap */
 		buf = mem_heap_alloc(heap, rec_offs_size(offsets));
@@ -301,8 +278,7 @@ row_build(
 		ut_ad(dict_table_get_format(index->table)
 		      < DICT_TF_FORMAT_ZIP);
 	} else if (j) {
-		*ext = row_ext_create(j, ext_cols, row,
-				      dict_table_zip_size(index->table),
+		*ext = row_ext_create(j, ext_cols, index->table->flags, row,
 				      heap);
 	} else {
 		*ext = NULL;
@@ -431,6 +407,10 @@ row_rec_to_index_entry(
 		rec = rec_copy(buf, rec, offsets);
 		/* Avoid a debug assertion in rec_offs_validate(). */
 		rec_offs_make_valid(rec, index, offsets);
+#ifdef UNIV_BLOB_NULL_DEBUG
+	} else {
+		ut_a(!rec_offs_any_null_extern(rec, offsets));
+#endif /* UNIV_BLOB_NULL_DEBUG */
 	}
 
 	entry = row_rec_to_index_entry_low(rec, index, offsets, n_ext, heap);

row/row0sel.c

@@ -99,10 +99,12 @@ row_sel_sec_rec_is_for_blob(
 	ulint		clust_len,	/*!< in: length of clust_field */
 	const byte*	sec_field,	/*!< in: column in secondary index */
 	ulint		sec_len,	/*!< in: length of sec_field */
-	ulint		zip_size)	/*!< in: compressed page size, or 0 */
+	dict_table_t*	table)		/*!< in: table */
 {
 	ulint	len;
-	byte	buf[DICT_MAX_INDEX_COL_LEN];
+	byte	buf[REC_VERSION_56_MAX_INDEX_COL_LEN];
+	ulint	zip_size = dict_table_flags_to_zip_size(table->flags);
+	ulint	max_prefix_len = DICT_MAX_FIELD_LEN_BY_FORMAT(table);
 
 	ut_a(clust_len >= BTR_EXTERN_FIELD_REF_SIZE);
 
@@ -116,7 +118,7 @@ row_sel_sec_rec_is_for_blob(
 		return(FALSE);
 	}
 
-	len = btr_copy_externally_stored_field_prefix(buf, sizeof buf,
+	len = btr_copy_externally_stored_field_prefix(buf, max_prefix_len,
 						      zip_size,
 						      clust_field, clust_len);
 
@@ -222,8 +224,7 @@ row_sel_sec_rec_is_for_clust_rec(
 					    col->mbminmaxlen,
 					    clust_field, clust_len,
 					    sec_field, sec_len,
-					    dict_table_zip_size(
-						    clust_index->table))) {
+					    clust_index->table)) {
 					goto inequal;
 				}
 

row/row0upd.c

@@ -1229,8 +1229,8 @@ row_upd_replace(
 	}
 
 	if (n_ext_cols) {
-		*ext = row_ext_create(n_ext_cols, ext_cols, row,
-				      dict_table_zip_size(table), heap);
+		*ext = row_ext_create(n_ext_cols, ext_cols, table->flags, row,
+				      heap);
 	} else {
 		*ext = NULL;
 	}

row/row0vers.c

@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1997, 2009, Innobase Oy. All Rights Reserved.
+Copyright (c) 1997, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -550,6 +550,11 @@ row_vers_build_for_consistent_read(
 				/* The view already sees this version: we can
 				copy it to in_heap and return */
 
+#ifdef UNIV_BLOB_NULL_DEBUG
+				ut_a(!rec_offs_any_null_extern(
+					     version, *offsets));
+#endif /* UNIV_BLOB_NULL_DEBUG */
+
 				buf = mem_heap_alloc(in_heap,
 						     rec_offs_size(*offsets));
 				*old_vers = rec_copy(buf, version, *offsets);
@@ -583,6 +588,10 @@ row_vers_build_for_consistent_read(
 		*offsets = rec_get_offsets(prev_version, index, *offsets,
 					   ULINT_UNDEFINED, offset_heap);
 
+#ifdef UNIV_BLOB_NULL_DEBUG
+		ut_a(!rec_offs_any_null_extern(prev_version, *offsets));
+#endif /* UNIV_BLOB_NULL_DEBUG */
+
 		trx_id = row_get_rec_trx_id(prev_version, index, *offsets);
 
 		if (read_view_sees_trx_id(view, trx_id)) {
@@ -682,6 +691,10 @@ row_vers_build_for_semi_consistent_read(
 			/* We found a version that belongs to a
 			committed transaction: return it. */
 
+#ifdef UNIV_BLOB_NULL_DEBUG
+			ut_a(!rec_offs_any_null_extern(version, *offsets));
+#endif /* UNIV_BLOB_NULL_DEBUG */
+
 			if (rec == version) {
 				*old_vers = rec;
 				err = DB_SUCCESS;
@@ -739,6 +752,9 @@ row_vers_build_for_semi_consistent_read(
 		version = prev_version;
 		*offsets = rec_get_offsets(version, index, *offsets,
 					   ULINT_UNDEFINED, offset_heap);
+#ifdef UNIV_BLOB_NULL_DEBUG
+		ut_a(!rec_offs_any_null_extern(version, *offsets));
+#endif /* UNIV_BLOB_NULL_DEBUG */
 	}/* for (;;) */
 
 	if (heap) {

srv/srv0srv.c

@@ -1406,7 +1406,7 @@ retry:
 
 	trx->op_info = "waiting in InnoDB queue";
 
-	thd_wait_begin(trx->mysql_thd, THD_WAIT_ROW_TABLE_LOCK);
+	thd_wait_begin(trx->mysql_thd, THD_WAIT_USER_LOCK);
 	os_event_wait(slot->event);
 	thd_wait_end(trx->mysql_thd);
 
@@ -1815,7 +1815,7 @@ srv_suspend_mysql_thread(
 
 	/* Suspend this thread and wait for the event. */
 
-	thd_wait_begin(trx->mysql_thd, THD_WAIT_ROW_TABLE_LOCK);
+	thd_wait_begin(trx->mysql_thd, THD_WAIT_ROW_LOCK);
 	os_event_wait(event);
 	thd_wait_end(trx->mysql_thd);
 

trx/trx0rec.c

@@ -1,6 +1,6 @@
 /*****************************************************************************
 
-Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved.
+Copyright (c) 1996, 2011, Oracle and/or its affiliates. All Rights Reserved.
 
 This program is free software; you can redistribute it and/or modify it under
 the terms of the GNU General Public License as published by the Free Software
@@ -351,10 +351,10 @@ trx_undo_rec_get_col_val(
 		ut_ad(*orig_len >= BTR_EXTERN_FIELD_REF_SIZE);
 		ut_ad(*len > *orig_len);
 		/* @see dtuple_convert_big_rec() */
-		ut_ad(*len >= BTR_EXTERN_FIELD_REF_SIZE * 2);
+		ut_ad(*len >= BTR_EXTERN_FIELD_REF_SIZE);
 		/* we do not have access to index->table here
 		ut_ad(dict_table_get_format(index->table) >= DICT_TF_FORMAT_ZIP
-		      || *len >= REC_MAX_INDEX_COL_LEN
+		      || *len >= col->max_prefix
 		      + BTR_EXTERN_FIELD_REF_SIZE);
 		*/
 
@@ -456,9 +456,10 @@ static
 byte*
 trx_undo_page_fetch_ext(
 /*====================*/
-	byte*		ext_buf,	/*!< in: a buffer of
-					REC_MAX_INDEX_COL_LEN
-					+ BTR_EXTERN_FIELD_REF_SIZE */
+	byte*		ext_buf,	/*!< in: buffer to hold the prefix
+					data and BLOB pointer */
+	ulint		prefix_len,	/*!< in: prefix size to store
+					in the undo log */
 	ulint		zip_size,	/*!< compressed page size in bytes,
 					or 0 for uncompressed BLOB  */
 	const byte*	field,		/*!< in: an externally stored column */
@@ -467,7 +468,7 @@ trx_undo_page_fetch_ext(
 {
 	/* Fetch the BLOB. */
 	ulint	ext_len = btr_copy_externally_stored_field_prefix(
-		ext_buf, REC_MAX_INDEX_COL_LEN, zip_size, field, *len);
+		ext_buf, prefix_len, zip_size, field, *len);
 	/* BLOBs should always be nonempty. */
 	ut_a(ext_len);
 	/* Append the BLOB pointer to the prefix. */
@@ -488,10 +489,11 @@ trx_undo_page_report_modify_ext(
 	byte*		ptr,		/*!< in: undo log position,
 					at least 15 bytes must be available */
 	byte*		ext_buf,	/*!< in: a buffer of
-					REC_MAX_INDEX_COL_LEN
-					+ BTR_EXTERN_FIELD_REF_SIZE,
+					DICT_MAX_FIELD_LEN_BY_FORMAT() size,
 					or NULL when should not fetch
 					a longer prefix */
+	ulint		prefix_len,	/*!< prefix size to store in the
+					undo log */
 	ulint		zip_size,	/*!< compressed page size in bytes,
 					or 0 for uncompressed BLOB  */
 	const byte**	field,		/*!< in/out: the locally stored part of
@@ -499,6 +501,8 @@ trx_undo_page_report_modify_ext(
 	ulint*		len)		/*!< in/out: length of field, in bytes */
 {
 	if (ext_buf) {
+		ut_a(prefix_len > 0);
+
 		/* If an ordering column is externally stored, we will
 		have to store a longer prefix of the field.  In this
 		case, write to the log a marker followed by the
@@ -507,7 +511,7 @@ trx_undo_page_report_modify_ext(
 
 		ptr += mach_write_compressed(ptr, *len);
 
-		*field = trx_undo_page_fetch_ext(ext_buf, zip_size,
+		*field = trx_undo_page_fetch_ext(ext_buf, prefix_len, zip_size,
 						 *field, len);
 
 		ptr += mach_write_compressed(ptr, *len);
@@ -553,7 +557,7 @@ trx_undo_page_report_modify(
 	ulint		i;
 	trx_id_t	trx_id;
 	ibool		ignore_prefix = FALSE;
-	byte		ext_buf[REC_MAX_INDEX_COL_LEN
+	byte		ext_buf[REC_VERSION_56_MAX_INDEX_COL_LEN
 				+ BTR_EXTERN_FIELD_REF_SIZE];
 
 	ut_a(dict_index_is_clust(index));
@@ -706,13 +710,21 @@ trx_undo_page_report_modify(
 			}
 
 			if (rec_offs_nth_extern(offsets, pos)) {
+				const dict_col_t*	col
+					= dict_index_get_nth_col(index, pos);
+				ulint			prefix_len
+					= dict_max_field_len_store_undo(
+						table, col);
+
+				ut_ad(prefix_len + BTR_EXTERN_FIELD_REF_SIZE
+				      <= sizeof ext_buf);
+
 				ptr = trx_undo_page_report_modify_ext(
 					ptr,
-					dict_index_get_nth_col(index, pos)
-					->ord_part
+					col->ord_part
 					&& !ignore_prefix
-					&& flen < REC_MAX_INDEX_COL_LEN
-					? ext_buf : NULL,
+					&& flen < REC_ANTELOPE_MAX_INDEX_COL_LEN
+					? ext_buf : NULL, prefix_len,
 					dict_table_zip_size(table),
 					&field, &flen);
 
@@ -791,11 +803,20 @@ trx_undo_page_report_modify(
 							  &flen);
 
 				if (rec_offs_nth_extern(offsets, pos)) {
+					const dict_col_t*	col =
+						dict_index_get_nth_col(
+							index, pos);
+					ulint			prefix_len =
+						dict_max_field_len_store_undo(
+							table, col);
+
+					ut_a(prefix_len < sizeof ext_buf);
+
 					ptr = trx_undo_page_report_modify_ext(
 						ptr,
-						flen < REC_MAX_INDEX_COL_LEN
+						flen < REC_ANTELOPE_MAX_INDEX_COL_LEN
 						&& !ignore_prefix
-						? ext_buf : NULL,
+						? ext_buf : NULL, prefix_len,
 						dict_table_zip_size(table),
 						&field, &flen);
 				} else {
@@ -1095,11 +1116,11 @@ trx_undo_rec_get_partial_row(
 			undo log record. */
 			if (!ignore_prefix && col->ord_part) {
 				ut_a(dfield_get_len(dfield)
-				     >= 2 * BTR_EXTERN_FIELD_REF_SIZE);
+				     >= BTR_EXTERN_FIELD_REF_SIZE);
 				ut_a(dict_table_get_format(index->table)
 				     >= DICT_TF_FORMAT_ZIP
 				     || dfield_get_len(dfield)
-				     >= REC_MAX_INDEX_COL_LEN
+				     >= REC_ANTELOPE_MAX_INDEX_COL_LEN 
 				     + BTR_EXTERN_FIELD_REF_SIZE);
 			}
 		}
@@ -1584,6 +1605,10 @@ trx_undo_prev_version_build(
 		return(DB_ERROR);
 	}
 
+# ifdef UNIV_BLOB_NULL_DEBUG
+	ut_a(!rec_offs_any_null_extern(rec, offsets));
+# endif /* UNIV_BLOB_NULL_DEBUG */
+
 	if (row_upd_changes_field_size_or_external(index, offsets, update)) {
 		ulint	n_ext;
 

trx/trx0undo.c

@@ -1985,8 +1985,6 @@ trx_undo_free_prepared(
 /*===================*/
 	trx_t*	trx)	/*!< in/out: PREPARED transaction */
 {
-	mutex_enter(&trx->rseg->mutex);
-
 	ut_ad(srv_shutdown_state == SRV_SHUTDOWN_EXIT_THREADS);
 
 	if (trx->update_undo) {
@@ -2001,6 +1999,5 @@ trx_undo_free_prepared(
 			       trx->insert_undo);
 		trx_undo_mem_free(trx->insert_undo);
 	}
-	mutex_exit(&trx->rseg->mutex);
 }
 #endif /* !UNIV_HOTBACKUP */

ut/ut0ut.c

@@ -662,6 +662,8 @@ ut_strerr(
 		return("Table is being used");
 	case DB_TOO_BIG_RECORD:
 		return("Record too big");
+	case DB_TOO_BIG_INDEX_COL:
+		return("Index columns size too big");
 	case DB_LOCK_WAIT_TIMEOUT:
 		return("Lock wait timeout");
 	case DB_NO_REFERENCED_ROW: