Comment 7 for bug 1577712

Verified. This is tricky to test because it is a race condition, but I have verified that it appears that mysql_upgrade isn't being called twice, and I have also ran a test upgrade five times with no issues.

I examined /var/log/upstart/mysql.log before and after a test upgrade. Previously, we were printing "Checking for tables which need an upgrade, are corrupt or were not closed cleanly." there when running mysql_upgrade from the upstart job. After a test upgrade to xenial-proposed, we longer see this string, so it seems likely that mysql-start was updated at the correct time. I also manually checked /var/log/apt/term.log after a release upgrade to xenial-proposed and this appears to be consistent with this. mysqld is running (still on upstart) after an upgrade. After a reboot, systemd has started it as expected.