Fix log_call test to not give random outcomes.

We shouldn't play around with the carriage return char,
seems like using that together with readline gives undefined results (wrt ordering).

tests/integration/subprocess.sh

@@ -6,9 +6,9 @@
 #
 # (cat <<EOF
 # 2 1 one\\\\n
-# 1 2 two\\\\n
+# 1 1 two\\\\n
-# 1 5 four
+# 1 3 four
-# 1 6 \\\\rNo, three..\\\\n
+# 1 1 \\\\rNo, three..\\\\n
 # EOF
 # ) | ./subprocess.sh
 #
@@ -24,6 +24,7 @@ while read line; do
 	rest=${line#* *}
 	delay=${rest%% *}
 	message=${rest#* }
-	eval "sleep $delay; printf \"$message\" >&$stream" &
+	sleep $delay
+	printf "$message" >&$stream
 done
 wait

tests/integration/tools_tests.py

@@ -21,19 +21,20 @@ def setup_logger():
 def test_log_call_output_order():
 	logged = setup_logger()
 	fixture = """
-2 0.1 one\\\\n
+2 0.00 one\\\\n
-1 0.2 two\\\\n
+1 0.02 two\\\\n
-1 0.5 four
+1 0.02 four\\\\n
-2 0.6 \\\\rNo, three..
+2 0.02 No, three..\\\\n
-1 0.8 \\\\rthree
+1 0.02 three\\\\n
 """
 	status, stdout, stderr = log_call([subprocess_path], stdin=fixture)
 	eq_(status, 0)
 	eq_(stderr, ['one', 'No, three..'])
-	eq_(stdout, ['two', 'four\rthree'])
+	eq_(stdout, ['two', 'four', 'three'])
 	expected_order = ['one',
 	                  'two',
-	                  'four\rthree',
+	                  'four',
 	                  'No, three..',
+	                  'three',
 	                  ]
-	eq_(logged.getvalue().split("\n")[8:-1], expected_order)
+	eq_(expected_order, logged.getvalue().split("\n")[8:-1])