Description:
Start mysqld on our g5, with --interactive-timeout=3.
Open a 'mysql' command line client, do nothing for more than 3 seconds, then issue a command: your connection is still alive :(
Contrary to what happens on my Linux, where I get as expected:
MASTER> show tables;
ERROR 2006 (HY000): MySQL server has gone away
No connection. Trying to reconnect...
Connection id:    6
This apparently affects other timeouts: --slave-net-timeout too (slave never auto-breaks its connection whereas it should - which explains why rpl_auto_increment failed *only* on Mac OS X on Feb 22nd 2005).
I don't know if related, but mysqld often does not react to SIGTERM on our g5 (it is not always repeatable but quite frequent), while it reacts to mysqladmin shutdown. Maybe timeouts are implemented with alarm() and so SIGALRM?
How to repeat:
see description