Since metricbeat light modules support processors (elastic#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 25b8bf1)

Since metricbeat light modules support processors (elastic#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 25b8bf1)

Since metricbeat light modules support processors (elastic#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 25b8bf1)

…-stage-level

* upstream/master: (30 commits)
  Add a GRPC listener service for Agent (elastic#18827)
  Disable host.* fields by default for iptables module (elastic#18756)
  [WIP] Clarify capabilities of the Filebeat auditd module (elastic#17068)
  fix: rename file and remove extra separator (elastic#18881)
  ci: enable JJBB (elastic#18812)
  Disable host.* fields by default for Checkpoint module (elastic#18754)
  Disable host.* fields by default for Cisco module (elastic#18753)
  Update latest.yml testing env to 7.7.0 (elastic#18535)
  Upgrade k8s.io/client-go and k8s keystore tests (elastic#18817)
  Add missing Jenkins stages for Auditbeat (elastic#18835)
  [Elastic Log Driver] Create a config shim between libbeat and the user (elastic#18605)
  Use indexers and matchers in config when defaults are enabled (elastic#18818)
  Fix panic on `metricbeat test modules` (elastic#18797)
  [CI] Fix permissions in MacOSX agents (elastic#18847)
  [Ingest Manager] When not port are specified and the https is used fallback to 443 (elastic#18844)
  [Ingest Manager] Fix install service script for windows (elastic#18814)
  [Metricbeat] Fix getting compute instance metadata with partial zone/region config (elastic#18757)
  Improve error messages in s3 input (elastic#18824)
  Add memory metrics into compute googlecloud (elastic#18802)
  include bucket name when logging error (elastic#18679)
  ...

Since metricbeat light modules support processors (#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 25b8bf1)

Since metricbeat light modules support processors (#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 25b8bf1)

…8853)

Since metricbeat light modules support processors (#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 25b8bf1)

Add nil pipeline from #16715, required for the fix.

(cherry picked from commit 7a1b524)

Co-authored-by: Steffen Siering <steffen.siering@elastic.co>

…es` (elastic#18853)

Since metricbeat light modules support processors (elastic#15923), module
initialization requires a publisher in the beat so modules can attach
their processors. `metricbeat test modules` is not initializing as
normal metricbeat commands, and it is not initializing any output or
publisher pipeline, so metricbeat panics when trying to initialize
modules with the new method.

This change adds a dummy publisher for this case, and fixes also a
condition that was adding a `nil` module option, causing additional
panics. A test that reproduced the issues is also added.

(cherry picked from commit 316793d)

Add nil pipeline from elastic#16715, required for the fix.

(cherry picked from commit 257fdfc)

Co-authored-by: Steffen Siering <steffen.siering@elastic.co>