Clean up after unit tests

[abetomo]

Which issue(s) this PR fixes:

Fixes #4048

What this PR does / why we need it:

After unit testing, clean up in /tmp.

(If you cannot terminate with SIGTERM and terminate with SIGKILL, some files cannot be cleaned up. #4050 (comment) #4050 (comment))

Docs Changes:

None

Release Note:

None

[abetomo]

With the committed changes and the following changes, all files were cleaned.

diff --git a/test/command/test_fluentd.rb b/test/command/test_fluentd.rb
index 461e19ef..9d62cd40 100644
--- a/test/command/test_fluentd.rb
+++ b/test/command/test_fluentd.rb
@@ -85,12 +85,12 @@ class TestFluentdCommand < ::Test::Unit::TestCase
         yield pid, io
         # p(here: "execute command", pid: pid, worker_pids: @worker_pids)
       ensure
-        Process.kill(:KILL, pid) rescue nil
+        Process.kill(:TERM, pid) rescue nil
         if @supervisor_pid
-          Process.kill(:KILL, @supervisor_pid) rescue nil
+          Process.kill(:TERM, @supervisor_pid) rescue nil
         end
         @worker_pids.each do |cpid|
-          Process.kill(:KILL, cpid) rescue nil
+          Process.kill(:TERM, cpid) rescue nil
         end
         # p(here: "execute command", pid: pid, exist: process_exist?(pid), worker_pids: @worker_pids, exists: @worker_pids.map{|i| process_exist?(i) })
         Timeout.timeout(10){ sleep 0.1 while process_exist?(pid) }

When you exit with SIGTERM, the shutdown process runs and the files are deleted.
However, the test was not stable because the process could not terminate in some cases.

[ashie]

How about trying SIGTERM first then SIGKILL after timeout?
Although we should make it stable but not needed to do it in this PR.

[abetomo]

@ashie Thanks!
I’ll give it a try.

[ashie]

LGTM

[ashie]

Retrying some unstable tests not concerned with this PR...

[daipom]

LGTM. Thanks for the fix!

[ashie]

Thanks!

[abetomo]

Thanks for your review!