diff options
| author | Vladislav Vaintroub <wlad@montyprogram.com> | 2013-05-23 14:52:48 +0200 |
|---|---|---|
| committer | Vladislav Vaintroub <wlad@montyprogram.com> | 2013-05-23 14:52:48 +0200 |
| commit | 08ce9bfe057b6cd31e7fbca4e4e9e48edde242fb (patch) | |
| tree | 3882462528dc16c79cb55a3528d2998997e7a320 /mysql-test/r/connect.result | |
| parent | 52045d40d4fe0df0e60fa37eb35609fcb0fd857a (diff) | |
| download | mariadb-git-08ce9bfe057b6cd31e7fbca4e4e9e48edde242fb.tar.gz | |
MDEV-4566 : Failing DBUG_ASSERT() in SELECT SLEEP(), with threadpool.
This bug only happens with long sleep()s ( > 5 sec), and in debug version.
Analysis:
The assertion is caused by nested thd_wait_begin() calls, which is not an expected condition.
- "outer" thd_wait_begin()) , in Item_func_sleep::val_int()
- "inner" thd_wait_begin() in Interruptible_wait::wait(). This function periodically checks whether connection is still valid, via THD::is_connection(), which ends up calling vio_io_wait() with timeout parameter set to 0.
Fix is not to call thd wait callback in vio_io_wait(), if timeout parameter is 0. There is no "waiting" in this case.
Diffstat (limited to 'mysql-test/r/connect.result')
0 files changed, 0 insertions, 0 deletions