Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.11
-
None
-
Notable changes
-
mariadbd --help output is flushed before exit to resolve an issue when used in scripting that where there is incomplete results of the output pipe of this command.
Description
Since commit 854869d69fd ("Fix a UBSAN runtime error on MacOSX", mysys/my_stack.c) the amd64-windows-packages builder fails on 10.11 before running any test:
perl mysql-test-run.pl --force --suite-timeout=120 --max-test-fail=10 --retry=3 --suite=vcol,gcol,perfschema,main,innodb,versioning,plugins,mariabackup,roles,auth_gssapi,rocksdb,sysschema
|
VS config: RelWithDebInfo
|
Checking supported features...
|
MariaDB Version 10.11.20-MariaDB
|
Could not parse mysqld --help: thread-
|
dojob: spawn returned error 255
|
The error comes from mysql-test/mariadb-test-run.pl (collect_mysqld_features()): a line of the "mysqld --help --verbose" variables list is just thread-, without a value. The output looks truncated in the middle of a variable name (thread-cache-size / thread-handling / thread-pool-* / thread-stack), which suggests mysqld dies while printing --help.
Builds (all builder amd64-windows-packages, id 239):
- 44116eca17c (
MDEV-36295): build 48319 - OK - 854869d69fd (Fix a UBSAN runtime error on MacOSX): build 48329 - fails (step "connect test")
- 6a063120d4d (
MDEV-23864): build 48334 - fails (step "nm test")
https://buildbot.mariadb.org/#/builders/239/builds/48329
https://buildbot.mariadb.org/#/builders/239/builds/48334
854869d69fd is the only commit between the last green and the first red build. It changes the fallback path of my_get_stack_bounds():
- *stack_end= (char *)(*stack_start) + stack_size * STACK_DIRECTION; |
+ *stack_end= (char *)(*stack_start) + ((ssize_t)stack_size) * STACK_DIRECTION; |
The plain amd64-windows builder (57066) is green on the same commit, so it is not reproducible on every Windows build.