Metricbeat filters cause panic when removing events

[tsg]

• Version: master
• Operating System: darwin

Starting metricbeat master with this config:

metricbeat.modules:
- module: system
  enabled: true
  period: 1m
  metricsets:
    - filesystem
  filters:
    - drop_event.when.regexp.mount_point: '^/(sys|cgroup|proc|dev|etc|host|lib)($|/)'

output.elasticsearch.hosts: ["localhost:9200"]

Causes a panic in the master branch:

2017/05/16 12:35:21.928875 beat.go:262: INFO metricbeat start running.
panic: assignment to entry in nil map

goroutine 53 [running]:
github.com/elastic/beats/libbeat/publisher.(*client).annotateEvent(0xc42031b140, 0x0)
	/Users/tsg/src/github.com/elastic/beats/libbeat/publisher/client.go:180 +0xe0
github.com/elastic/beats/libbeat/publisher.(*client).PublishEvent(0xc42031b140, 0x0, 0x0, 0x0, 0x0, 0x1)
	/Users/tsg/src/github.com/elastic/beats/libbeat/publisher/client.go:99 +0x50
github.com/elastic/beats/metricbeat/mb/module.PublishChannels.func1(0xc420318c00)
	/Users/tsg/src/github.com/elastic/beats/metricbeat/mb/module/publish.go:25 +0xd4
created by github.com/elastic/beats/metricbeat/mb/module.PublishChannels
	/Users/tsg/src/github.com/elastic/beats/metricbeat/mb/module/publish.go:32 +0xec

But works with 5.4.1

The issue seems related to the filter, which results in a nil event.

[andrewkroh]

We used to have nil checks at

beats/metricbeat/mb/module/wrapper.go

Lines 284 to 286 in 91970fd

	if event != nil {
	rtnEvents = append(rtnEvents, event)
	}

. But I forgot to put them back in while refactoring.