vtctlclient Validate claims slaves are not replicating

[slanning]

vtctlclient Validate claims that slaves aren't replicating although they are.

Running this command
$ vtctlclient -server :15999 Validate -ping-tablets

The output includes pairs of error lines like this for each slave:

E0117 13:36:01.812848    3926 main.go:60] E0117 13:36:01.812707 validator.go:52] slave hostname.example.com not in replication graph for shard ks1/80- (mysql instance without vttablet?)
E0117 13:36:01.814426    3926 main.go:60] E0117 13:36:01.814245 validator.go:52] slave cell1-0000001234 not replicating: 192.168.0.1 slave list: ["hostname.example.com" (others elided...)]

As far as I can tell, the code around https://github.com/vitessio/vitess/blob/850574e0f70/go/vt/wrangler/validator.go#L203 assumes that GetSlaves returns a list of IP addresses. It obtains those using FindSlaves in https://github.com/vitessio/vitess/blob/850574e0f70/go/vt/mysqlctl/replication.go#L274
Although that's documented as "FindSlaves gets IP addresses for all currently connected slaves", what it does is: run show processlist on the master, find rows whose Command is like 'Binlog Dump%', and strip off the port from the Host column. Poking around various databases here, the Host column generally has hostnames instead of IP addresses, so that GetSlaves returns hostnames. As a result, back in validator.go tabletIPMap does indeed map from IP addresses, so that when comparing if tabletIPMap[normalizeIP(slaveAddr)] == nil it's always nil, because normalizeIP(slaveAddr) is a hostname instead of an IP address. I guess that FindSlaves should be fixed to return IP addresses as it is documented.