Fix race condition in test

The write wait guarantees that the input was written, but not that the output was read into the log. I modified the tests to partition the output so the order fo the assertions matters. This prevents the text `> hello` from earlier from triggering a false positive.

Author: schneems

test/integration/background_stdin_test.rb

@@ -67,15 +67,24 @@ def test_print_output_on_exit
           on_success_dir: dir.join(SUCCESS_DIRNAME)
         ).call
 
-        assert_include(actual: io.string, include_str: "Warning background task is still running, cleaning up: `background_ungraceful_exit`")
-        assert_include(actual: io.string, include_str: "Log contents for `/usr/bin/env bash -c")
-        assert_include(actual: io.string, include_str: "You said: hello")
+        logs = io.string
+
+        match_after = partition_match_after(actual: logs, include_str: "Warning background task is still running, cleaning up: `background_ungraceful_exit`")
+        match_after = partition_match_after(actual: match_after, include_str: "Log contents for `/usr/bin/env bash -c")
+        partition_match_after(actual: match_after, include_str: "> hello")
       end
     end
   end
 
-  def assert_include(actual:, include_str:)
-    assert actual.include?(include_str), "Expected to find `#{include_str}` in output, but did not. Output:\n#{actual}"
+  # Finds the include_str if it exists or raises an error
+  # Returns the contents of that string and everything after it
+  #
+  # Used to handle the case where output might be in the logs twice and we want to verify the order
+  def partition_match_after(actual:, include_str:)
+    _before, match, after = actual.partition(include_str)
+    found = match && !match.empty?
+    assert found, "Expected to find `#{include_str}` in output, but did not. Output:\n#{actual}"
+    [match, after].join
   end
 
   def loop_script