Bug #22080 CSV Engine CHECK fails to identify some corruption

• Submitted: 7 Sep 2006 11:23
• Modified: 10 Oct 2006 18:43
• Reporter: MC Brown
• Status: Closed
• Category: MySQL Server: CSV
• Severity: S3 (Non-critical)
• Version: 5.1.11
• OS: Linux (Linux 2.6 (Gentoo))
• Assigned to: Antony Curtis
• CPU Architecture: Any

[7 Sep 2006 11:23] MC Brown

Description:
I created a simple table:

create table doctest (id int,string varchar(64)) Engine=CSV;

Inserted three rows:

insert into doctest values(1,'string');
insert into doctest values(2,'string');
insert into doctest values(3,'string');

Checked the table:

 check table doctest;
+--------------+-------+----------+----------+
| Table        | Op    | Msg_type | Msg_text |
+--------------+-------+----------+----------+
| test.doctest | check | status   | OK       |
+--------------+-------+----------+----------+
1 row in set (0.00 sec)

Corrupting a single row:

2","string

Produces a corrupt table:

check table doctest;
+--------------+-------+----------+----------+
| Table        | Op    | Msg_type | Msg_text |
+--------------+-------+----------+----------+
| test.doctest | check | error    | Corrupt  |
+--------------+-------+----------+----------+
1 row in set (0.01 sec)

Different corruption on the same initial table:

"2",string"

This is not identified as bad, and in fact reads bad data:

check table doctest;
+--------------+-------+----------+----------+
| Table        | Op    | Msg_type | Msg_text |
+--------------+-------+----------+----------+
| test.doctest | check | status   | OK       |
+--------------+-------+----------+----------+
1 row in set (0.00 sec)

mysql> select * from doctest;
+------+--------+
| id   | string |
+------+--------+
|    1 | string |
|    2 | 3      |
|    3 | string |
+------+--------+
3 rows in set (0.00 sec)

How to repeat:
See above. 

Edit the corresponding .CSV while MySQL is running to 'force' the corruption. 

Suggested fix:
CSV Engine should be checking for matching quots within fields

[8 Sep 2006 19:39] Sveta Smirnova

Thank you for the report.

Verified as described on Linux using last BK sources.

[8 Sep 2006 19:46] Sveta Smirnova

While I test this report, I found problem with REPAIR. I think the resolution of these 2 issues is likely to be the same, so I paste "how to reproduce" here:

-- use table doctest from the original testcase filled by original 3 records,
-- then corrupt a single row as in first example:

2","string

Try to repair it:

mysql> check table doctest;
+--------------+-------+----------+----------+
| Table        | Op    | Msg_type | Msg_text |
+--------------+-------+----------+----------+
| test.doctest | check | error    | Corrupt  |
+--------------+-------+----------+----------+
1 row in set (0.13 sec)

mysql> repair table doctest;
+--------------+--------+----------+----------+
| Table        | Op     | Msg_type | Msg_text |
+--------------+--------+----------+----------+
| test.doctest | repair | status   | OK       |
+--------------+--------+----------+----------+
1 row in set (0.13 sec)

mysql> select * from doctest;
+------+--------+
| id   | string |
+------+--------+
|    1 | string |
+------+--------+
1 row in set (0.09 sec)

[22 Sep 2006 18:36] Bugs System

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/12426

ChangeSet@1.2341, 2006-09-22 11:35:45-07:00, [e-mail] +5 -0
  Bug#22080
    "CHECK fails to identify some corruption"
    Amended code to recogise corruption as illustrated in bug report.
  Support NULL in CSV

[25 Sep 2006 14:10] Calvin Sun

Sent comment via email.

[25 Sep 2006 16:06] Bugs System

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/12486

ChangeSet@1.2323, 2006-09-25 09:04:19-07:00, [e-mail] +3 -0
  Bug#22080
    "CHECK fails to identify some corruption"
    change to allow it to detect more errors in data format

[3 Oct 2006 20:06] Antony Curtis

approved by Brian via email.

[4 Oct 2006 2:56] Antony Curtis

pushed to mysql-5.1-engines tree

[9 Oct 2006 11:17] Sergey Vojtovich

Fixed in 5.1.12.

[10 Oct 2006 18:43] Paul DuBois

Noted in 5.1.12 changelog.