MariaDB loop with MyISAM tables marked as crashed

[Sergei Golubchik]

On several servers (3) MariaDB 5.1 crashed after an upgrade from from MySQL 5.0.
Log shows:
===
Version: '5.1.50-MariaDB-mariadb84-log' socket: '/var/run/mysqld/mysqld.sock'
port: 3306 (MariaDB - http://mariadb.com/)
101008 14:35:17 [ERROR] mysqld: Table './database123/table123' is marked as
crashed and should be repaired
101008 14:35:17 [Warning] Checking table: './database123/table123'
101008 14:35:17 mysqld_safe Number of processes running now: 0
101008 14:35:17 mysqld_safe mysqld restarted
===

This is looping, so things only get worse every time with more crashed tables.
Using myisamchk to repair all MYI tables actually "fixes" the issue, but I guess
it can come back after an unclean shutdown. This appears to be a major bug. I have
a copy of such corrupt database available on request. (see WL#155)

[Dreas van Donselaar (dreas)]

Monty found that this occurred because thread_stack was set to 128K instead of 256K which is required for MariaDB 5.1. Still seems that should be fixed though (or at least MariaDB should refuse to start with a clear error message).