Print debug stacktraces from within a forked worker

Fixes CHEF-4357: http://tickets.opscode.com/browse/CHEF-4357

Rescue exceptions from within a forker worker and feed them to
`Chef::Application.debug_stacktrace()`, then exit non-zero. This fixes
an issue where Chef would spew ruby stacktraces with forking enabled
(stacktraces should only be printed to console with `-l debug`).
Additionally fixes an issue where a generic exception was passed to
`debug_stacktrace`, leaving a stacktrace.out file with no useful
information.

In addition to the above fixes, includes the following improvements:
* Update the forked worker's argv0 to clearly indicate it's the worker
  process, and include ppid and run start time.
* When reaping the worker process, include more useful information when
  the worker exited non-successfully, such as exit code or signal that
  killed the process. In particular, OOM kill (e.g., when running
  resource intensive code compile, etc. on small machine) should be a
  little more obvious.

1 files changed, 0 insertions, 2 deletions

diff --git a/lib/chef/application/client.rb b/lib/chef/application/client.rb
index 42a7744795..b08a795697 100644
--- a/lib/chef/application/client.rb
+++ b/lib/chef/application/client.rb
@@ -316,7 +316,6 @@ class Chef::Application::Client < Chef::Application
       rescue Exception => e
         if Chef::Config[:interval]
           Chef::Log.error("#{e.class}: #{e}")
-          Chef::Application.debug_stacktrace(e)
           Chef::Log.error("Sleeping for #{Chef::Config[:interval]} seconds before trying again")
           unless SELF_PIPE.empty?
             client_sleep Chef::Config[:interval]
@@ -326,7 +325,6 @@ class Chef::Application::Client < Chef::Application
           end
           retry
         else
-          Chef::Application.debug_stacktrace(e)
           Chef::Application.fatal!("#{e.class}: #{e.message}", 1)
         end
       end