reversed gcov to work the old way - the worst thing that can happen
if slave and master execute the same code at the same time is that
we get a lesser count on the code branch execution, but in any case,
it will be at least counted once, which is really all we care.
Besides, the source tracking is most likely not thread safe, but this
is ok for the same reason.
it's yet another mysqld hidden feature
--exit-info=256 is like --core-file but prints info at exit.
--exit-info itself implies also one-thread (undocumented)
myrg_info.c merge + insert OK
mymrgdef.h merge + insert OK
mrg000001.test test for the above
mysql-test-run compatibility (FreeBSD) changes from mysql-4.0 backpropagated
cleanup - remove unneeded daeamon restarts
sql/log_event.cc
fixed bug - do not confuse confused slaves even further if they try to
read less than LOG_EVENT_HEADER_LEN before the end of file
sql/mysqlbinlog.cc
give error if constructor fails
sql/mysqld.cc
--disconnect-slave-event-count for replication debugging
sql/slave.cc
--disconnect-slave-event-count
cleanup to remove unneeded sleeps when things go wrong - sleep only
when we have to
sql/slave.h
--disconnect-slave-event-count
test cases for ALTER TABLE ORDER BY and replication offset off by 22
confusion mystery bug (both pass)
