Details
-
Bug
-
Status: Closed (View Workflow)
-
Minor
-
Resolution: Fixed
-
10.2(EOL), 10.3(EOL), 10.4(EOL), 10.5
Description
Table names are supposed to be case-sensitive on a case-sensitive file system. Alas, as noted in MDEV-25748, DROP DATABASE would seem drop InnoDB tables in a case-insensitive fashion, affecting tables in databases whose name only differs in the letter case:
diff --git a/mysql-test/suite/innodb/t/alter_foreign_crash.test b/mysql-test/suite/innodb/t/alter_foreign_crash.test
|
index 1952a1b30d4..b65dfa175c2 100644
|
--- a/mysql-test/suite/innodb/t/alter_foreign_crash.test
|
+++ b/mysql-test/suite/innodb/t/alter_foreign_crash.test
|
@@ -2,6 +2,7 @@
|
--source include/have_debug_sync.inc
|
# The embedded server does not support restarting.
|
--source include/not_embedded.inc
|
+--source include/have_case_sensitive_file_system.inc
|
|
--echo #
|
--echo # Bug #20476395 DICT_LOAD_FOREIGNS() FAILED IN
|
@@ -10,6 +11,11 @@
|
|
call mtr.add_suppression("InnoDB: Failed to load table");
|
|
+CREATE DATABASE Bug;
|
+CREATE TABLE Bug.parent(a SERIAL) ENGINE=INNODB;
|
+CREATE TABLE Bug.child(a SERIAL, FOREIGN KEY f(a) REFERENCES Bug.parent(a))
|
+ENGINE=INNODB;
|
+
|
create database bug;
|
use bug;
|
|
@@ -18,6 +24,7 @@ create table child(a serial, foreign key fk (a) references parent(a))engine=inno
|
|
insert into parent values(1);
|
insert into child values(1);
|
+drop database Bug;
|
|
connect (con1,localhost,root,,bug);
|
SET DEBUG_SYNC='innodb_rename_table_ready SIGNAL s1 WAIT_FOR s2 EXECUTE 2';
|
@@ -32,6 +39,9 @@ disconnect con1;
|
|
show tables;
|
alter table parent row_format=dynamic;
|
-
|
+select * from child;
|
+--error ER_ROW_IS_REFERENCED_2
|
+drop table parent;
|
+drop table child;
|
drop table parent;
|
drop database bug; |
The changed test would fail in an unexpected way (not a mere result difference):
|
10.2 629449172a5b0a6975663ca1ac420789e00b941d |
CURRENT_TEST: innodb.alter_foreign_crash
|
mysqltest: At line 42: query 'select * from child' failed: 1146: Table 'bug.child' doesn't exist
|
Possibly we could fix this by porting the fixes of MDEV-25691 and MDEV-25748 from MariaDB Server 10.6, but it might not be a good idea because the MDEV-25691 fix removes work-arounds for the lack of atomic DDL (MDEV-17567).
Hence, even though I usually treat InnoDB corruption bugs at high priority, I am only filing this bug for documenting its existence. I do not plan to fix this, because the fix could cause more damage, and because the bug has existed for years without anyone complaining.
Attachments
Issue Links
- relates to
-
-
- Closed
-
-
-
- Closed
-