commit 05247a26019ce5416e13ef80a12c307bdc790ee4
author Nick Vatamaniuc <vatamane@gmail.com> Tue Jul 26 15:56:42 2016 -0400
committer Nick Vatamaniuc <vatamane@gmail.com> Tue Jul 26 16:00:17 2016 -0400
tree eefaea7472b584958644cceec24f75e9483431e7
parent c661e74d2de99007804b6cb60489421d68fe406f

Replication manager's rep_start_pids now contains only {Tag, Pid} items

Previously the local change feed was added to rep_start_pids as Pid only. So if
replication manager stopped and terminate/2 was called before that change
feed died, then

```
foreach(fun({_Tag, Pid}) -> ... end, [StartPids])
```

would crash with a function clause error.

Make sure add the replicator db name to the changes feed.

Jira: COUCHDB-3082

src/couch_replicator_manager.erl

diff --git a/src/couch_replicator_manager.erl b/src/couch_replicator_manager.erl
index dca1154..0b7eeca 100644
--- a/src/couch_replicator_manager.erl
+++ b/src/couch_replicator_manager.erl
@@ -183,7 +183,7 @@
         scan_pid = ScanPid,
         max_retries = retries_value(
             config:get("replicator", "max_replication_retry_count", "10")),
-        rep_start_pids = [Pid],
+        rep_start_pids = [{?REPLICATOR_DB, Pid}],
         live = Live,
         epoch = Epoch
     }}.