Fix root split on very large append.

This commit fixes a bug where a root split on a very large insert would
cause an overflow of the root node. The problem was that the new root
was not split when it was created so a new root with more than 64K child
nodes would overflow the page.count (uint16).

3 files changed, 61 insertions, 22 deletions

diff --git a/bucket_test.go b/bucket_test.go
index 13e6a8a..99fdad8 100644
--- a/bucket_test.go
+++ b/bucket_test.go
@@ -107,6 +107,31 @@ func TestBucket_Put_Large(t *testing.T) {
 	})
 }
 
+// Ensure that a database can perform multiple large appends safely.
+func TestDB_Put_VeryLarge(t *testing.T) {
+	if testing.Short() {
+		t.Skip("skipping test in short mode.")
+	}
+
+	n, batchN := 400000, 200000
+	ksize, vsize := 8, 500
+
+	withOpenDB(func(db *DB, path string) {
+		for i := 0; i < n; i += batchN {
+			err := db.Update(func(tx *Tx) error {
+				b, _ := tx.CreateBucketIfNotExists([]byte("widgets"))
+				for j := 0; j < batchN; j++ {
+					k, v := make([]byte, ksize), make([]byte, vsize)
+					binary.BigEndian.PutUint32(k, uint32(i+j))
+					assert.NoError(t, b.Put(k, v))
+				}
+				return nil
+			})
+			assert.NoError(t, err)
+		}
+	})
+}
+
 // Ensure that a setting a value on a key with a bucket value returns an error.
 func TestBucket_Put_IncompatibleValue(t *testing.T) {
 	withOpenDB(func(db *DB, path string) {
diff --git a/db_test.go b/db_test.go
index 0809bcb..2063249 100644
--- a/db_test.go
+++ b/db_test.go
@@ -608,15 +608,36 @@ func withOpenDB(fn func(*DB, string)) {
 
 // mustCheck runs a consistency check on the database and panics if any errors are found.
 func mustCheck(db *DB) {
-	err := db.View(func(tx *Tx) error {
-		return <-tx.Check()
+	db.View(func(tx *Tx) error {
+		// Collect all the errors.
+		var errors []error
+		for err := range tx.Check() {
+			errors = append(errors, err)
+			if len(errors) > 10 {
+				break
+			}
+		}
+
+		// If errors occurred, copy the DB and print the errors.
+		if len(errors) > 0 {
+			var path = tempfile()
+			tx.CopyFile(path, 0600)
+
+			// Print errors.
+			fmt.Print("\n\n")
+			fmt.Printf("consistency check failed (%d errors)\n", len(errors))
+			for _, err := range errors {
+				fmt.Println(err)
+			}
+			fmt.Println("")
+			fmt.Println("db saved to:")
+			fmt.Println(path)
+			fmt.Print("\n\n")
+			os.Exit(-1)
+		}
+
+		return nil
 	})
-	if err != nil {
-		// Copy db off first.
-		var path = tempfile()
-		db.View(func(tx *Tx) error { return tx.CopyFile(path, 0600) })
-		panic("check failure: " + err.Error() + ": " + path)
-	}
 }
 
 // mustContainKeys checks that a bucket contains a given set of keys.
diff --git a/node.go b/node.go
index cda5201..75433ad 100644
--- a/node.go
+++ b/node.go
@@ -188,6 +188,8 @@ func (n *node) write(p *page) {
 	} else {
 		p.flags |= branchPageFlag
 	}
+
+	_assert(len(n.inodes) < 0xFFFF, "inode overflow: %d (pgid=%d)", len(n.inodes), p.id)
 	p.count = uint16(len(n.inodes))
 
 	// Loop over each item and write it to the page.
@@ -367,20 +369,11 @@ func (n *node) spill() error {
 		tx.stats.Spill++
 	}
 
-	// This is a special case where we need to write the parent if it is new
-	// and caused by a split in the root.
-	var parent = n.parent
-	if parent != nil && parent.pgid == 0 {
-		// Allocate contiguous space for the node.
-		p, err := tx.allocate((parent.size() / tx.db.pageSize) + 1)
-		if err != nil {
-			return err
-		}
-
-		// Write the new root.
-		_assert(p.id < tx.meta.pgid, "pgid (%d) above high water mark (%d)", p.id, tx.meta.pgid)
-		parent.pgid = p.id
-		parent.write(p)
+	// If the root node split and created a new root then we need to spill that
+	// as well. We'll clear out the children to make sure it doesn't try to respill.
+	if n.parent != nil && n.parent.pgid == 0 {
+		n.children = nil
+		return n.parent.spill()
 	}
 
 	return nil