Details
-
Bug
-
Status: Confirmed (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.4(EOL), 10.5, 10.6, 10.11, 11.0(EOL), 11.1(EOL), 11.2(EOL), 11.3(EOL), 11.4
Description
Recent versions of GCC and clang require much larger thread stack size when the server is built with -DWITH_ASAN=ON. This is by design, and a good thing, because it will allow more stack buffer overflows to be caught.
A number of tests in ./mtr --suite=perfschema are explicitly specifying thread_stack=655360, which usually is more than twice the DEFAULT_THREAD_STACK that is defined in include/my_pthread.h. However, the minimum for ASAN instrumented builds can be as large as 11 megabytes (more than 10 times this size).
The following patch would prevent server startup failures in a number of tests when I use cmake -DWITH_ASAN=ON on clang 16.0.6:
diff --git a/mysql-test/suite/perfschema/t/max_program_zero-master.opt b/mysql-test/suite/perfschema/t/max_program_zero-master.opt
|
index 7803f9c5e1b..7e5af1a2e3a 100644
|
--- a/mysql-test/suite/perfschema/t/max_program_zero-master.opt
|
+++ b/mysql-test/suite/perfschema/t/max_program_zero-master.opt
|
@@ -1,3 +1,2 @@
|
--loose-performance_schema_max_program_instances=0
|
--loose-performance_schema_max_statement_stack=1
|
---thread_stack=655360
|
diff --git a/mysql-test/suite/perfschema/t/start_server_low_digest_sql_length-master.opt b/mysql-test/suite/perfschema/t/start_server_low_digest_sql_length-master.opt
|
index e6483c360f9..0e7696ccaca 100644
|
--- a/mysql-test/suite/perfschema/t/start_server_low_digest_sql_length-master.opt
|
+++ b/mysql-test/suite/perfschema/t/start_server_low_digest_sql_length-master.opt
|
@@ -1,3 +1,2 @@
|
---thread_stack=655360
|
--max_digest_length=10
|
--loose-performance_schema_max_sql_text_length=10
|
diff --git a/mysql-test/suite/perfschema/t/start_server_zero_digest_sql_length-master.opt b/mysql-test/suite/perfschema/t/start_server_zero_digest_sql_length-master.opt
|
index 5ee2230e539..5ddf5d8caae 100644
|
--- a/mysql-test/suite/perfschema/t/start_server_zero_digest_sql_length-master.opt
|
+++ b/mysql-test/suite/perfschema/t/start_server_zero_digest_sql_length-master.opt
|
@@ -1,3 +1,2 @@
|
---thread_stack=655360
|
--max_digest_length=0
|
--loose-performance_schema_max_sql_text_length=0
|
diff --git a/mysql-test/suite/perfschema/t/statement_digest_long_query-master.opt b/mysql-test/suite/perfschema/t/statement_digest_long_query-master.opt
|
deleted file mode 100644
|
index 3f9d7900961..00000000000
|
--- a/mysql-test/suite/perfschema/t/statement_digest_long_query-master.opt
|
+++ /dev/null
|
@@ -1 +0,0 @@
|
---thread_stack=655360
|
diff --git a/mysql-test/suite/perfschema/t/statement_program_concurrency-master.opt b/mysql-test/suite/perfschema/t/statement_program_concurrency-master.opt
|
index 8c0dedeeb17..df5d7b1b706 100644
|
--- a/mysql-test/suite/perfschema/t/statement_program_concurrency-master.opt
|
+++ b/mysql-test/suite/perfschema/t/statement_program_concurrency-master.opt
|
@@ -1 +1 @@
|
---innodb_lock_wait_timeout=1 --thread_stack=655360
|
+--innodb_lock_wait_timeout=0
|
diff --git a/mysql-test/suite/perfschema/t/statement_program_lost_inst-master.opt b/mysql-test/suite/perfschema/t/statement_program_lost_inst-master.opt
|
index 699c91930ae..0744e09f8ae 100644
|
--- a/mysql-test/suite/perfschema/t/statement_program_lost_inst-master.opt
|
+++ b/mysql-test/suite/perfschema/t/statement_program_lost_inst-master.opt
|
@@ -1,3 +1,3 @@
|
--loose-performance_schema_max_program_instances=7
|
--loose-performance_schema_max_statement_stack=2
|
---thread_stack=655360
|
+--thread_stack=10485760
|
diff --git a/mysql-test/suite/perfschema/t/statement_program_lost_inst.test b/mysql-test/suite/perfschema/t/statement_program_lost_inst.test
|
index 023180b9d2b..fa70ec963e8 100644
|
--- a/mysql-test/suite/perfschema/t/statement_program_lost_inst.test
|
+++ b/mysql-test/suite/perfschema/t/statement_program_lost_inst.test
|
@@ -22,7 +22,7 @@ let $restart_file= $MYSQLTEST_VARDIR/tmp/mysqld.1.expect;
|
--exec echo "wait" > $restart_file
|
--shutdown_server
|
--source include/wait_until_disconnected.inc
|
---exec echo "restart:--performance_schema_max_program_instances=7 --performance_schema_max_statement_stack=2 --thread_stack=655360">$restart_file
|
+--exec echo "restart:--performance_schema_max_program_instances=7 --performance_schema_max_statement_stack=2">$restart_file
|
--enable_reconnect
|
--source include/wait_until_connected_again.inc
|
|
diff --git a/mysql-test/suite/perfschema/t/statement_program_nested-master.opt b/mysql-test/suite/perfschema/t/statement_program_nested-master.opt
|
deleted file mode 100644
|
index 3f9d7900961..00000000000
|
--- a/mysql-test/suite/perfschema/t/statement_program_nested-master.opt
|
+++ /dev/null
|
@@ -1 +0,0 @@
|
---thread_stack=655360
|
diff --git a/mysql-test/suite/perfschema/t/statement_program_nesting_event_check-master.opt b/mysql-test/suite/perfschema/t/statement_program_nesting_event_check-master.opt
|
deleted file mode 100644
|
index 3f9d7900961..00000000000
|
--- a/mysql-test/suite/perfschema/t/statement_program_nesting_event_check-master.opt
|
+++ /dev/null
|
@@ -1 +0,0 @@
|
---thread_stack=655360
|
diff --git a/mysql-test/suite/perfschema/t/statement_program_non_nested-master.opt b/mysql-test/suite/perfschema/t/statement_program_non_nested-master.opt
|
deleted file mode 100644
|
index 3f9d7900961..00000000000
|
--- a/mysql-test/suite/perfschema/t/statement_program_non_nested-master.opt
|
+++ /dev/null
|
@@ -1 +0,0 @@
|
---thread_stack=655360 |
I did not check what would happen in a build that would use the smaller DEFAULT_THREAD_STACK. The above patch is only indicating what is wrong and needs to be fixed in some way.
Attachments
Issue Links
- relates to
-
MDEV-34099 AddressSanitizer running out of memory regardless of stack_thread size
- Closed
-
MDEV-16660 Inadequate DEFAULT_THREAD_STACK size for AddressSanitizer
- Closed