datadir can be saved in case of constantly failing servers
Remove datadirs each time all servers have been stopped thus limiting
the size of vardir.
The datadirs of failed tests have been saved and that should be
enough to diagnose any failures

into  pilot.mysql.com:/data/msvensson/mysql/mysql-5.1-rpl

Ie don't treat cluster_config.mysqld.1.1 as a section specifying a cluster

directory part
 

into  pilot.mysql.com:/data/msvensson/mysql/mysql-5.1-rpl

into  mysql.com:/home/ksm/mysql/WL3754/commit-mysql-5.1-new-rpl

into  pilot.mysql.com:/data/msvensson/mysql/mysql-5.1-rpl

we don't use Unix domin sockets for connection to server 

early versions of 5.1 and 5.0
Fix usage instruction of defaults_extra_file

moving the supressions from perl to SQL

Just treat var/tmp and $opt_tmndir that might be different or same. 

running on windows. Apparently Cygwin tries to emulate UNIX Socket and thus
fails to create one.

"eval use" it. Instead trap the error that occurs when creating the socket
on platforms that does not support it. 

Straigthen code to be more explicit

Also make it possible to specify a path to the suite, for example:
./mtr --suite=/some/path/to/suitedir 

since it's used by the check

Don't allow unix socket path to be truncated 

Fix warnings

the server dissapears
Check return code of 'mysql_ping'
Add "shutdown", call 'mysql_shutdown' on the current connection