Fix regression from Bug#12845774 OPTIMISTIC INSERT/UPDATE USES WRONG

HEURISTICS FOR COMPRESSED PAGE SIZE

The fix of Bug#12845774 was supposed to skip known-to-fail
btr_cur_optimistic_insert() calls. There was only one such call, in
btr_cur_pessimistic_update(). All other callers of
btr_cur_pessimistic_insert() would release and reacquire the B-tree
page latch before attempting the pessimistic insert. This would allow
other threads to restructure the B-tree, allowing (and requiring) the
insert to succeed as an optimistic (single-page) operation.

Failure to attempt an optimistic insert before a pessimistic one would
trigger an attempt to split an empty page.

rb:1234 approved by Sunny Bains

storage/innodb_plugin/ibuf/ibuf0ibuf.c

@@ -2752,11 +2752,19 @@ ibuf_insert_low(
 
 		root = ibuf_tree_root_get(&mtr);
 
-		err = btr_cur_pessimistic_insert(BTR_NO_LOCKING_FLAG
+		err = btr_cur_optimistic_insert(
+			BTR_NO_LOCKING_FLAG | BTR_NO_UNDO_LOG_FLAG,
+			cursor, ibuf_entry, &ins_rec,
+			&dummy_big_rec, 0, thr, &mtr);
+
+		if (err == DB_FAIL) {
+			err = btr_cur_pessimistic_insert(
+				BTR_NO_LOCKING_FLAG
 				| BTR_NO_UNDO_LOG_FLAG,
-						 cursor,
-						 ibuf_entry, &ins_rec,
+				cursor, ibuf_entry, &ins_rec,
 				&dummy_big_rec, 0, thr, &mtr);
+		}
+
 		if (err == DB_SUCCESS) {
 			/* Update the page max trx id field */
 			page_update_max_trx_id(btr_cur_get_block(cursor), NULL,

storage/innodb_plugin/row/row0ins.c

@@ -2179,9 +2179,16 @@ row_ins_index_entry_low(
 
 				goto function_exit;
 			}
-			err = btr_cur_pessimistic_insert(
+
+			err = btr_cur_optimistic_insert(
 				0, &cursor, entry, &insert_rec, &big_rec,
 				n_ext, thr, &mtr);
+
+			if (err == DB_FAIL) {
+				err = btr_cur_pessimistic_insert(
+					0, &cursor, entry, &insert_rec,
+					&big_rec, n_ext, thr, &mtr);
+			}
 		}
 	}