1. place the following units in the systemd search path:
ExecStart=/usr/bin/bash -c "sleep 10; exit 255"
ExecStop=/usr/bin/bash -c "exit 1"
2. Run systemctl start foo.service. foo.service and bar.service will both enter the loaded/active state
3. After 10s, foo.service will enter the loaded/failed state, while bar.service will enter loaded/inactive. The the Restart policy of foo.service should be applied, but it is not, and foo.service remains inactive until a user intervenes.
If you reattempt the above steps after removing either the Requires or the ExecStop options from foo.service, however, the Restart policy will be applied successfully. I expected the Restart policy to be applied even with the ExecStop/Requires options.
$ systemctl --version
-PAM -AUDIT -SELINUX +IMA -APPARMOR +SMACK -SYSVINIT +UTMP +LIBCRYPTSETUP -GCRYPT -GNUTLS -ACL -XZ -LZ4 +SECCOMP +BLKID -ELFUTILS +KMOD -IDN
+1 for fix for this issue.
not working here two with a simpler sddm service (systemd 219). sddm is crashing and core dumping ramdonly due to a driver problem