Reduce the number of "false negative" test failures for Interactive/configure.py due to unreliable ordering of statements in the actual stdout output. Replace the expected literal text with a series of regular expressions in a custom match function. Update CHANGES.txt as well.

1 files changed, 68 insertions, 8 deletions

diff --git a/test/Interactive/configure.py b/test/Interactive/configure.py
index 44a53eb..dfe1c7f 100644
--- a/test/Interactive/configure.py
+++ b/test/Interactive/configure.py
@@ -32,6 +32,73 @@ Also tests that "b" can be used as a synonym for "build".
 __revision__ = "__FILE__ __REVISION__ __DATE__ __DEVELOPER__"
 
 import TestSCons
+import TestCmd
+import re
+
+# The order of the statements in the expected stdout is unreliable.
+
+# case 1
+# scons>>> .*foo\.cpp.*
+# scons>>> scons: `foo.obj' is up to date.
+# scons>>> .*foo\.cpp.*
+# scons>>> scons: `foo.obj' is up to date.
+# scons>>>\s*
+
+# case 2
+# scons>>> .*foo\.cpp.*
+# scons>>> .*foo\.cpp.*
+# scons>>> scons: `foo.obj' is up to date.
+# scons>>> scons: `foo.obj' is up to date.
+# scons>>>\s*
+
+# The order of this list is related to the order of the counts below
+expected_patterns = [
+    re.compile("^scons>>> .*foo\.cpp.*$"),
+    re.compile("^scons>>> scons: `foo.obj' is up to date\.$"),
+    re.compile("^scons>>>\s*$"),
+]
+
+# The order of this list is related to the order of the regular expressions above
+expected_counts = [
+    [2,2,1], # case 1 and 2
+]
+
+# This is used for the diff output when the test fails and to distinguish between
+# stdout and stderr in the match function below.
+expect_stdout = r"""scons>>> .*foo\.cpp.*
+scons>>> .*foo\.cpp.*
+scons>>> scons: `foo.obj' is up to date.
+scons>>> scons: `foo.obj' is up to date.
+scons>>>\s*
+"""
+
+def match_custom(lines, expect):
+
+    if expect != expect_stdout:
+        # stderr
+        if lines == expect:
+            return 1
+        return None
+
+    # taken directly from TestCmd
+    if not TestCmd.is_List(lines):
+        # CRs mess up matching (Windows) so split carefully
+        lines = re.split('\r?\n', lines)
+
+    # record number of matches for each regex
+    n_actual = [0] * len(expected_patterns)
+    for line in lines:
+        for i, regex in enumerate(expected_patterns):
+            if regex.search(line):
+                n_actual[i] += 1
+                break
+
+    # compare actual counts to expected counts
+    for n_expect in expected_counts:
+        if n_actual == n_expect:
+            return 1
+
+    return None
 
 _python_ = TestSCons._python_
 
@@ -108,14 +175,7 @@ scons.send("b foo.obj\n")
 
 scons.send("build foo.obj\n")
 
-expect_stdout = r"""scons>>> .*foo\.cpp.*
-scons>>> .*foo\.cpp.*
-scons>>> scons: `foo.obj' is up to date.
-scons>>> scons: `foo.obj' is up to date.
-scons>>>\s*
-"""
-
-test.finish(scons, stdout = expect_stdout, match=TestSCons.match_re)
+test.finish(scons, stdout = expect_stdout, stderr = '', match = match_custom)
 
 test.pass_test()