summaryrefslogtreecommitdiff
path: root/innobase/ibuf
diff options
context:
space:
mode:
authorheikki@hundin.mysql.fi <>2004-01-29 15:08:22 +0200
committerheikki@hundin.mysql.fi <>2004-01-29 15:08:22 +0200
commit49587c42441d8d636f1781b8e162c517c70619fd (patch)
treee8ef350d0b39b75fc32fd3507469f8b55c216482 /innobase/ibuf
parent85b00b7fb5288cf93360cc0ba5401eb37edc8ae8 (diff)
downloadmariadb-git-49587c42441d8d636f1781b8e162c517c70619fd.tar.gz
ibuf0ibuf.c:
Correct the comment about the ibuf record format in >= 4.1.1
Diffstat (limited to 'innobase/ibuf')
-rw-r--r--innobase/ibuf/ibuf0ibuf.c9
1 files changed, 6 insertions, 3 deletions
diff --git a/innobase/ibuf/ibuf0ibuf.c b/innobase/ibuf/ibuf0ibuf.c
index b4a28d5fcd8..805f08af361 100644
--- a/innobase/ibuf/ibuf0ibuf.c
+++ b/innobase/ibuf/ibuf0ibuf.c
@@ -48,9 +48,12 @@ insert buffer tree, and that is in the system tablespace of InnoDB.
1. The first field is the space id.
2. The second field is a one-byte marker which differentiates records from
the < 4.1.x storage format.
-3. The third field contains the type info, where we have also added 2 bytes to
- store the charset.
-4. The rest of the fields contain the fields of the actual index record.
+3. The third field is the page number.
+4. The fourth field contains the type info, where we have also added 2 bytes to
+ store the charset. In the compressed table format of 5.0.x we must add more
+ information here so that we can build a dummy 'index' struct which 5.0.x
+ can use in the binary search on the index page in the ibuf merge phase.
+5. The rest of the fields contain the fields of the actual index record.
*/