Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Critical
-
Resolution: Fixed
-
Affects Version/s: 5.1.67, 5.2.14, 5.3.12, 10.1, 10.0, 5.5
-
Fix Version/s: 10.1.6
-
Component/s: Character Sets
-
Labels:
-
Environment:# mysql --version
mysql Ver 15.1 Distrib 5.5.39-MariaDB, for debian-linux-gnu (x86_64) using readline 5.1
-
Sprint:10.1.6-2
Description
It looks like that, the db server could not query if the value has a dash "-" inside, as far as I know, the affected charset is armscii8.
For more see the repo command as below, and the db dump is attached:
MariaDB [bugtest]> create table test(columnname varchar(64) CHARACTER SET armscii8);
Query OK, 0 rows affected (0.07 sec)
MariaDB [bugtest]> insert into test values ('abc-def');
Query OK, 1 row affected (0.04 sec)
MariaDB [bugtest]> select * from test where columnname = 'abc-def';
Empty set (0.00 sec)
MariaDB [bugtest]> select * from test where columnname like 'abc%';
+------------+
| columnname |
+------------+
| abc-def |
+------------+
1 row in set (0.00 sec)
Gliffy Diagrams
Attachments
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
It seems that the problem happens during utf8-to-armscii8 conversion because
the following ASCII characters have double encoding in the 8-bit range (0x80..0xFF):
0xA4 U+0029 RIGHT PARENTHESIS 0xA5 U+0028 LEFT PARENTHESIS 0xA9 U+002E FULL STOP 0xAB U+002C COMMA 0xAC U+002D HYPHEN-MINUS 0xFF U+0027 APOSTROPHE
So utf8 dash '-' is erroneously converted to armscii 0xAC instead of 0x2D:
MariaDB [test]> SELECT HEX(CONVERT(_utf8 0x2D USING armscii8)); +-----------------------------------------+ | HEX(CONVERT(_utf8 0x2D USING armscii8)) | +-----------------------------------------+ | AC | +-----------------------------------------+ 1 row in set (0.00 sec)
This should be fixed.
There is also a problem in the collation definition. It should probably sort the double coded characters as equal (e.g. armscii 0x2D should be equal to 0xAC).
@Elena Stepanova, I did not test or report to upstream MySQL. Do I need to do that? Or maybe you can help?
Yingyu Cheng,
No problem, if we decide it's worth trying, then either Alexander Barkov or I will do that.
Thanks for the report.
Same on MySQL 5.7, so if it's a bug, it's an upstream issue.
Alexander Barkov,
It does look like a bug to me, but I don't know how much this charset is supported.
If you decide it should be fixed, but prefer to treat it as an upstream bug, please report it at bugs.mysql.com (or maybe you know it has already been reported?). Alternatively, it can be fixed directly in MariaDB.