Add "log hashing" to see new failures

A common issue is that we see that a test goes from N failures to N+1,
but it's not clear at all which of the failures is new.

This adds a simple hashing mechanism to try to help us in these cases:
we hash each line in the log, and construct a small bloom filter of
which lines were seen.  We can then put this into the test file, and
when the test fails we can run the received output against the expected
bloom filter.

I tried testing it and it seems ok, but we'll have to see how well it
works in practice.

test/extra/M2Crypto_test.py

@@ -43,7 +43,12 @@ def install_and_test_lxml():
     subprocess.check_call([PYTHON_EXE, "setup.py", "install"], cwd=M2CRYPTO_DIR, env=env)
 
     expected = [{'ran': 235, 'errors': 5, 'skipped': 2}]
-    run_test([PYTHON_EXE, "setup.py", "test"], cwd=M2CRYPTO_DIR, expected=expected)
+    expected_log_hash = '''
+    rAAABwigI04NBogROC1ATTYBiAUIAMhCMKBLQAC1SMALgBCRQIIAgKOpBBGgIaMKAAUAkVgCUJAA
+    ABCMAIwJAQAAwONQATYSBmEYSACDAEUoRABkJKEAhBBEMgYMwQoFABBwCETByQggaAkAqAgUgAAG
+    QBWgAamiIaSEIIIiGjE=
+    '''
+    run_test([PYTHON_EXE, "setup.py", "test"], cwd=M2CRYPTO_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, None, force_create = True)
 install_and_test_lxml()

test/extra/avro_test.py

@@ -20,4 +20,9 @@ env["PYTHONPATH"] = os.path.abspath(os.path.join(ENV_NAME, "lib/python2.7/site-p
 
 # cpython has the same number of failures
 expected = [{'failed': 2, 'passed': 47}]
-run_test([PYTEST_EXE], env=env, cwd=AVRO_DIR, expected=expected)
+expected_log_hash = '''
+gBAAAACAAAABBAgAAAACAAgAIAAABAQAAAAAAACAAAgDAIAAABAAIAMBAQgACBAAFBAQAACBAAAA
+EAEAAAAQAABAAAAAAIAAAAAAAAQAAAgICCgAEBAAAAAQAAAAAACAAAAAAAEAgAAAAAIAAAAAgBAA
+AZQAAAAAAAAAIAAEAAA=
+'''
+run_test([PYTEST_EXE], env=env, cwd=AVRO_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/babel_test.py

@@ -20,4 +20,9 @@ subprocess.check_call([PYTHON_EXE, "setup.py", "build"], cwd=BABEL_DIR)
 subprocess.check_call([PYTHON_EXE, "setup.py", "install"], cwd=BABEL_DIR)
 
 expected = [{"ran": 227, "failures": 3, "errors": 3}]
-run_test([NOSETESTS_EXE], cwd=BABEL_DIR, expected=expected)
+expected_log_hash = '''
+gAIAAAAACQAAAABAAAAABAAAAIAAEAAAAAAAAAAAAEAEBAAAAAAAkAAAAAAAAAAAQAAEgAAAAAAA
+AAAAAAAAAQAACAgAAAAAAAAAIAAJAAAAAAAAAAAAAAAAAAAAAEAAAAAAAhAAAAAAAAAAEACAAAAA
+EIgAAAAQAAAAAIAAAAA=
+'''
+run_test([NOSETESTS_EXE], cwd=BABEL_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/cffi_1.7_test.py

@@ -30,7 +30,12 @@ def install_and_test_cffi():
     # dir_to_test = "."
     # I just picked a subdirectory; I don't really know what it's testing.
     dir_to_test = os.path.join(CFFI_DIR,  "testing", "cffi1")
-    run_test([PYTEST_EXE, dir_to_test], cwd=CFFI_DIR, expected=expected)
+    expected_log_hash = '''
+    gBEACAAQEAIAwICAAAAAAABAAAAAAAACAAAAEAAAEAAAAEQEQAAAAKAAARFEEACKAAABAACAAAAA
+    QgAAAEAQBAACgAEAAABAAAAAAAFAAAoAAAAAAACAAAACAAIAAUiAIAAAAAODgAAgEIQBABAACgAC
+    GBACAAAAICEAABAAgQA=
+    '''
+    run_test([PYTEST_EXE, dir_to_test], cwd=CFFI_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, ["pytest==2.8.7", "py==1.4.31", "pycparser==2.14"], force_create = True)
 install_and_test_cffi()

test/extra/cffi_test.py

@@ -26,9 +26,19 @@ def install_and_test_cffi():
     # looks like clang 3.5 causes more errors like: 214 != -42 doing casts
     if os.environ.has_key("CC") and "clang" in os.environ["CC"]:
         expected = [{ "failed": 20, "passed": 1659, "skipped": 73, "xfailed": 4}]
+        expected_log_hash = '''
+        oRkAgDIgEgAAwoKiAIQAIABAQAAAAAIKBOAIUABAEAAAIMQFgQCKhKEgERFEMAgAAAIBAAiCCBAC
+        CAIASESQBAQDpAEAAAogAAMBAoVQqkCKABBAAIDgAKECABJAAQiEIAAgAgOigAIwcoQBIAAACoAG
+        2FIHAAQAJIELIVABgwA=
+        '''
     else:
         expected = [{ "failed": 11, "passed": 1668, "skipped": 73, "xfailed": 4}]
-    run_test([PYTEST_EXE], cwd=CFFI_DIR, expected=expected)
+        expected_log_hash = '''
+        oRkAwBAg0gAEwoCiQIQgIQBAQAABQEKKBGAZVAhKcAAAAMQFAQAogKggFRFGEIgAAAKABgiGCBCC
+        CAIASEAQHAQSpAEADEugCJEBAoFgIECDBBBEAACgACECAAJKgQicIAAgAAOChBIyUoQBIAAACoAG
+        2FInAAQQpIEHARAJowE=
+        '''
+    run_test([PYTEST_EXE], cwd=CFFI_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, ["pytest==2.8.7", "py==1.4.31", "pycparser==2.14"], force_create = True)
 install_and_test_cffi()

test/extra/cheetah_test.py

@@ -10,4 +10,9 @@ cheetah_exe = os.path.join(ENV_NAME, "bin", "cheetah")
 env = os.environ
 env["PATH"] = os.path.join(ENV_NAME, "bin")
 expected = [{'ran': 2138, 'errors': 4}, {'ran': 2138, 'errors': 232, 'failures': 2}]
-run_test([cheetah_exe, "test"], cwd=ENV_NAME, expected=expected, env=env)
+expected_log_hash = '''
+jcoDAKUIQTpEDIDiMwAuQFEAKABjEbNAAAACgqABAAGgGsGQaQQLg/l0gIQXbEA4IKQisBIAAlOQ
+IG4lA5AAASAqqGdMCPAAALKbAEQAYAcCEgRHAQCAAhAVJIghShwAUpAAKaEwgk0GaEUkgQIIADgb
+pKTQYrIACAshhJ6Bwh0=
+'''
+run_test([cheetah_exe, "test"], cwd=ENV_NAME, expected=expected, env=env, expected_log_hash=expected_log_hash)

test/extra/formencode_test.py

@@ -16,4 +16,9 @@ create_virtenv(ENV_NAME, packages, force_create = True)
 subprocess.check_call(["patch", "-p1"], stdin=open(os.path.join(os.path.dirname(__file__), "formencode.patch")), cwd=SRC_DIR)
 
 expected = [{'ran': 201}]
-run_test([NOSETESTS_EXE], cwd=FORMENCODE_DIR, expected=expected)
+expected_log_hash = '''
+gAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAQAAAAAAAAAAgAAAAAAA
+AAAAAAAAAAAAAAAAAAAAAAAAAAgAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+AAAAAAAAAAAAAAAAAAA=
+'''
+run_test([NOSETESTS_EXE], cwd=FORMENCODE_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/geoip_test.py

@@ -11,4 +11,12 @@ pkg = ["nose==1.3.7", "-e", "git+http://github.com/maxmind/geoip-api-python.git@
 create_virtenv(ENV_NAME, pkg, force_create = True)
 GEOIP_DIR = os.path.abspath(os.path.join(SRC_DIR, "geoip"))
 expected = [{'ran': 10}]
-run_test([PYTHON_EXE, "setup.py", "test"], cwd=GEOIP_DIR, expected=expected)
+
+expected_log_hash = '''
+ggAAAAAAQAQAAAAACAAAAAAAAAAAAAIABIAAAAAAgAACAAAAAAAIAAAAAAAAAAIIBAAAgABABAgA
+AAAAAAAAAAAAAAAAAAAAAAQAAIgAAAAAAAAAAQBAABAAEAEAAAAAAAAAAAgAAAAIAIAAAAEAAAAA
+AIAAAAgAAAAAAAAAAAA=
+'''
+
+
+run_test([PYTHON_EXE, "setup.py", "test"], cwd=GEOIP_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/lxml_test.py

@@ -34,7 +34,12 @@ def install_and_test_lxml():
     subprocess.check_call([PYTHON_EXE, "setup.py", "build_ext", "-i", "--with-cython"], cwd=LXML_DIR)
 
     expected = [{'ran': 1381}]
-    run_test([PYTHON_EXE, "test.py"], cwd=LXML_DIR, expected=expected)
+    expected_log_hash = '''
+    gAAAAAAAAQAAAAAIAAAAAAAAAAAAgAAAAAAAAABAAACCAEgAAAAAgAIAAAAAAACAAAAAoAAAAAAA
+    ABAAAAAAAAAAAAAigAAAAAAAAAAQAAAwAgAAAAAAAAAAAAAAAAIAAAAEAAAACAAAAAAAAABBAAAA
+    AAAAAAAAAAAAAAAAAAA=
+    '''
+    run_test([PYTHON_EXE, "test.py"], cwd=LXML_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, None, force_create = True)
 install_and_test_lxml()

test/extra/mysqldb_test.py

@@ -31,7 +31,12 @@ def install_and_test_mysqldb():
     env = os.environ
     env["TESTDB"] = "travis.cnf"
     expected = [{"ran": 69}]
-    run_test([nosetests_exe], cwd=MYSQLDB_DIR, expected=expected, env=env)
+    expected_log_hash = '''
+    gAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAA
+    AAAAAAAEAAAAAAAAAAAAAAAAAAgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+    AAAAAAAAAAAQAAAAAAA=
+    '''
+    run_test([nosetests_exe], cwd=MYSQLDB_DIR, expected=expected, env=env, expected_log_hash=expected_log_hash)
 
 packages = ["nose==1.3.7"]
 create_virtenv(ENV_NAME, packages, force_create = True)

test/extra/numpy_fulltest.py

@@ -77,8 +77,13 @@ except:
 
     raise
 
+expected_log_hash = '''
+gAAAAAAAAAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+AAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAQAAAAAABAAACAAAAAAAAAAAAA
+AAAgAAAAAAAAAAAAAAA=
+'''
 test_helper.run_test(['sh', '-c', '. %s/bin/activate && python %s/numpy/tools/test-installed-numpy.py' % (ENV_DIR, ENV_DIR)],
-        ENV_NAME, [dict(ran=6139, failures=1)])
+        ENV_NAME, [dict(ran=6139, failures=1)], expected_log_hash=expected_log_hash)
 
 print
 print "PASSED"

test/extra/paste_test.py

@@ -37,7 +37,10 @@ print ">> "
 # - no shiftjis encoding
 # - slightly different error messages
 expected = [{"failed" : 22, "passed" : 112}]
-run_test([PYTEST_EXE], cwd=PASTE_TEST_DIR, expected=expected)
-
-
+expected_log_hash = '''
+ghA0YC2ZyyvAxlQASAgCMAglIjZ2pwSCB8wuCIgiAYGKAITIQgIQLRgRYwA0e1BchxcHGsShlUij
+OBXOA0E4AQIkzEKIOCikJYUgRF5hh8YJVAioAI6FDJB8waBqoGC7hEIDCdEZkACMA4IDaLABMFSi
+n7AogIFoJ4iCeUEEggM=
+'''
+run_test([PYTEST_EXE], cwd=PASTE_TEST_DIR, expected=expected, expected_log_hash=expected_log_hash)
 

test/extra/protobuf_test.py

@@ -28,7 +28,12 @@ def install_and_test_protobuf():
     subprocess.check_call([PYTHON_EXE, "setup.py", "build"], cwd=PROTOBUF_PY_DIR, env=env)
 
     expected = [{"ran": 216}]
-    run_test([PYTHON_EXE, "setup.py", "test"], cwd=PROTOBUF_PY_DIR, expected=expected, env=env)
+    expected_log_hash = '''
+    gAAQSBxQEAxCwBwkAAREFCAUCQAAiAsIBggpNIQAAIBBBAAEAAQQAAADDEgABFI9QpcAlQAAgwEi
+    HEAJAESKkAKBGAAlpAAIAMggcAgAQQsQMwCkEgAisDKIAhEhABCMEE4CBAAEQQQAgIAIiIAEJBIy
+    gUBSkjAAIAUAQA8EIAI=
+    '''
+    run_test([PYTHON_EXE, "setup.py", "test"], cwd=PROTOBUF_PY_DIR, expected=expected, env=env, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, None, force_create = True)
 install_and_test_protobuf()

test/extra/pyicu_test.py

@@ -39,7 +39,12 @@ def install_and_test_pyicu():
     subprocess.check_call([PYTHON_EXE, "setup.py", "install"], cwd=PYICU_DIR, env=env)
 
     expected = [{'ran': 17}]
-    run_test([PYTHON_EXE, "setup.py", "test"], cwd=PYICU_DIR, expected=expected)
+    expected_log_hash = '''
+    gAAAAQAAABQAACBAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAARAAAAAAAAAAAAAAAIAIAAgAAAAAAA
+    AAAAAAgAAAACAAAAAAAAAAIAAAiAAAgAAQQAAAAAABAAAIEBAAAAAAAAACAAAAAAAAAAIIAIAAAA
+    AAAAAAAAAAAAAAACgAA=
+    '''
+    run_test([PYTHON_EXE, "setup.py", "test"], cwd=PYICU_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, None, force_create = True)
 install_and_test_pyicu()

test/extra/pylons_test.py

@@ -22,7 +22,12 @@ def install_and_test_pylons():
     # most of the errors are because of our coerceUnicodeToStr which raises a TypeError instead of a UnicodeError
     # but as we don't support the unicode string correctly every where I don't want to change this currently.
     expected = [{ "ran": 50, "errors": 7}]
-    run_test([NOSE_EXE], cwd=PYLONS_DIR, expected=expected)
+    expected_log_hash = '''
+    wLKBAAEAEQAABEAgAAUAYBABtBACiIFIAoAIIAiAYAIUBADgCOIAggAIBACQCAgIgAGBgCAsAIAB
+    FCIAQAAQAQQAmQoAAACEMQAiAaIAFIgAEEAAAUgAAGAIQAEAAEBQQABQAEAAAAAAAiEiIEAAAEIC
+    ECBAiigwIAAABAQIAQE=
+    '''
+    run_test([NOSE_EXE], cwd=PYLONS_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 pkg = [ "Mako==1.0.3",
         "decorator==4.0.9",

test/extra/pyopenssl_test.py

@@ -25,4 +25,9 @@ subprocess.check_call(["sed", "-i", 's/\\(def test_export_text.*\\)/\\1\\n
 print os.path.join(PYOPENSSL_DIR, "test", "test_crypto.py")
 
 expected = [{'ran': 438}]
-run_test([NOSETESTS_EXE], cwd=PYOPENSSL_DIR, expected=expected)
+expected_log_hash = '''
+gAAAAAAAAAAAAAAgAgAAAAAAAAAAAAAAAAAAAAAAQAAAAAAAAAAAAAAAAAAAAIAAAAAAgAAAAAAA
+AAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAAAAAEAAAAAAAA
+AAAAACAAAAAAgAAAAAA=
+'''
+run_test([NOSETESTS_EXE], cwd=PYOPENSSL_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/routes_test.py

@@ -13,4 +13,9 @@ create_virtenv(ENV_NAME, pkg, force_create = True)
 
 ROUTES_DIR = os.path.abspath(os.path.join(SRC_DIR, "routes"))
 expected = [{ "ran" : 141 }]
-run_test([PYTHON_EXE, "setup.py", "test"], cwd=ROUTES_DIR, expected=expected)
+expected_log_hash = '''
+ggSACQMCAQZABAAUIIMCEABCAAAgQAACDAUABAKgAAgEAAECIAAgIAgBABQICCDoIkNQgQAAIQQE
+xACSGAIAoIAAAgAQAQEIAAAQAFhWACgBBHAEYAAgIBQAUGAAAAIABCAEQEgAAAFAACAAAo5EgBAA
+AAAQCSBIURAAwDCAkQA=
+'''
+run_test([PYTHON_EXE, "setup.py", "test"], cwd=ROUTES_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/sass_test.py

@@ -15,5 +15,9 @@ packages += ["-e", "git+https://github.com/dahlia/libsass-python@0.8.3#egg=libsa
 create_virtenv(ENV_NAME, packages, force_create = True)
 
 expected = [{'ran': 75}]
-run_test([PYTHON_EXE, "setup.py", "test"], cwd=SASS_DIR, expected=expected)
-
+expected_log_hash = '''
+wEAIQAAAAAQEhBAAAAgQIF0QAAAQQAAAQigEAAwwABACAARUEBEAACTAAAQAEAAIAABJgQAEACAC
+AEAAUQQACBAAAAEBBABAARAAAQgCAHAWBQQAAABkaDABQAAYYAAEJtgCAIAgcoASgAwgUAAIRAQR
+QAQgIAQACCAgFEDEiFA=
+'''
+run_test([PYTHON_EXE, "setup.py", "test"], cwd=SASS_DIR, expected=expected, expected_log_hash=expected_log_hash)

test/extra/simplejson_test.py

@@ -20,7 +20,13 @@ def install_and_test_simplejson():
     subprocess.check_call([PYTHON_EXE, "setup.py", "install"], cwd=SIMPLEJSON_DIR)
 
     expected = [{'ran': 170}]
-    run_test([PYTHON_EXE, "setup.py", "test"], cwd=SIMPLEJSON_DIR, expected=expected)
+    expected_log_hash = '''
+    gAAAAAAAAAAAAABAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAA
+    AAAAAAAAAAAAAAAAAEAAAAQAAAgAAAAAAAAAAAAAABAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
+    AAIAAAAAAAAAAAAAAAA=
+    '''
+
+    run_test([PYTHON_EXE, "setup.py", "test"], cwd=SIMPLEJSON_DIR, expected=expected, expected_log_hash=expected_log_hash)
 
 create_virtenv(ENV_NAME, None, force_create = True)
 install_and_test_simplejson()

test/extra/unidecode_test.py

@@ -20,7 +20,12 @@ def install_and_test_unidecode():
     subprocess.check_call([PYTHON_EXE, "setup.py", "install"], cwd=UNIDECODE_DIR)
     
     expected = [{'ran': 8}]
-    run_test([PYTHON_EXE, "setup.py", "test"], cwd=UNIDECODE_DIR, expected=expected)
+    expected_log_hash = '''
+    gECAAAAAAAAAAABAAAAAAAAAAAAAAAEAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAA
+    AAAAGAAAAAAAAAAAAAAAAAQAAggAAAAAAAAAAAAAABAAAAIAAAAAAAAAAAAAAAggAAAAAAAAAAAA
+    AAAAAAAAAAAAAAAAAAA=
+    '''
 
+    run_test([PYTHON_EXE, "setup.py", "test"], cwd=UNIDECODE_DIR, expected=expected, expected_log_hash=expected_log_hash)
 create_virtenv(ENV_NAME, None, force_create = True)
 install_and_test_unidecode()

test/integration/pycrypto_test.py

@@ -59,7 +59,13 @@ assert enc_data != test_string
 assert key.decrypt(enc_data) == test_string
 
 expected = [{'ran': 1891}]
-test_helper.run_test([sys.executable, "setup.py", "test"], pycrypto_dir, expected)
+expected_log_hash = '''
+gAAAAAAAAAAAAABAAAAAAAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAAAAAIAAIAgAAAAAAA
+BAABAABAAAAAAAAAAAAAAAQAAAgAAAAAAAAAAAAAABAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAA
+AAAAAAAAAAAAAAAAAAA=
+'''
+
+test_helper.run_test([sys.executable, "setup.py", "test"], pycrypto_dir, expected, expected_log_hash=expected_log_hash)
 
 print "-- Tests finished"
 

test/lib/test_helper.py

@@ -57,7 +57,8 @@ def parse_output(output):
                     result[-1][res_type] = int(m.group(1))
     return result
 
-def run_test(cmd, cwd, expected, env = None):
+def run_test(cmd, cwd, expected, expected_log_hash="", env=None):
+    assert isinstance(expected_log_hash, str)
     print "Running", cmd
     process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, cwd=cwd, env=env)
     output, unused_err = process.communicate()
@@ -68,12 +69,114 @@ def run_test(cmd, cwd, expected, env = None):
     print "Return code:", errcode
     assert errcode in (0, 1), "\n\n%s\nTest process crashed" % output
 
+    expected_log_hash = expected_log_hash.strip()
+    this_log_hash = log_hash(output)
+
+    if expected_log_hash == "":
+        raise Exception("please set the expected log hash: \nexpected_log_hash = '''\n%s\n'''" % (this_log_hash,))
+
     if expected == result:
         print "Received expected output"
+        different = check_hash(output, expected_log_hash)
+
+        # These checks are useful for making sure that we have the right expected
+        # hashes in our test files, but I don't think it's worth failing the build for them:
+        # assert not different, "expected_log_hash = '''\n%s\n'''" % (this_log_hash,)
+        # assert this_log_hash == expected_log_hash, "expected_log_hash = '''\n%s\n'''" % (this_log_hash,)
     else:
         print >> sys.stderr, '\n'.join(output.split('\n')[-500:])
+        print >> sys.stderr, '\n'
+        different = check_hash(output, expected_log_hash)
+        print >> sys.stderr, '\n'
         print >> sys.stderr, "WRONG output"
         print >> sys.stderr, "is:", result
         print >> sys.stderr, "expected:", expected
+
+        if not different:
+            print >> sys.stderr, "(log hash can't detect missing lines)"
+        if this_log_hash != expected_log_hash:
+            print >> sys.stderr, "expected_log_hash = '''\n%s\n'''" % (this_log_hash,)
         assert result == expected
 
+# Try to canonicalize the log to remove most spurious differences.
+# We won't be able to get 100% of them, since there will always be differences in the number of
+# python warnings or compiler messages.
+# But try to remove the most egregious things (filename differences, timing differences) so that the output is easier to parse.
+def process_log(log):
+    r = []
+    for l in log.split('\n'):
+        # Remove timing data:
+        l = re.sub("tests in ([\\d\\.])+s", "", l)
+        l = re.sub("in ([\\d\\.])+ seconds", "", l)
+
+        # Remove filenames:
+        # log = re.sub("/[^ ]*.py:\\d", "", log)
+        # log = re.sub("/[^ ]*.py.*line \\d", "", log)
+        if "http://" not in l:
+            l = re.sub("(^|[ \"\'/])/[^ :\"\']*($|[ \":\'])", "", l)
+
+        # Remove pointer ids:
+        l = re.sub('0x([0-9a-f]{8,})', "", l)
+
+        r.append(l)
+
+    return r
+
+def log_hash(log, nbits=1024):
+    log_lines = process_log(log)
+
+    bits = [0] * nbits
+
+    for l in log_lines:
+        bits[hash(l) % nbits] = 1
+
+    assert sum(bits) < nbits * .67, "hash is very full!"
+
+    l = []
+    for i in xrange(0, nbits, 8):
+        t = 0
+        for j in xrange(8):
+            if bits[i + j]:
+                t += 1 << (7 - j)
+        l.append(chr(t))
+    return ''.join(l).encode('base64').strip()
+
+def check_hash(log, expected_hash):
+    orig_log_lines = log.split('\n')
+    log_lines = process_log(log)
+
+    s = expected_hash.decode('base64')
+    nbits = len(s) * 8
+    bits = [0] * nbits
+
+    for i in xrange(len(s)):
+        c = ord(s[i])
+        for j in xrange(8):
+            bit = (c >> (7 - j)) & 1
+            if bit:
+                bits[i * 8 + j] = True
+
+    missing = [False] * len(log_lines)
+    for i, l in enumerate(log_lines):
+        if not bits[hash(l) % nbits]:
+            missing[i] = True
+
+    ncontext = 2
+    def ismissing(idx, within):
+        for i in xrange(max(0, idx-within), min(len(log_lines), idx+within+1)):
+            if missing[i]:
+                return True
+        return False
+    different = False
+    for i in xrange(len(log_lines)):
+        if ismissing(i, 0):
+            different = True
+            if orig_log_lines[i] != log_lines[i]:
+                print >>sys.stderr, "\033[30m+ % 4d: %s\033[0m" % (i + 1, orig_log_lines[i])
+                print >>sys.stderr, "+ % 4d: %s" % (i + 1, log_lines[i])
+            else:
+                print >>sys.stderr, "+ % 4d: %s" % (i + 1, orig_log_lines[i])
+        elif ismissing(i, ncontext):
+            print >>sys.stderr, "  % 4d: %s" % (i + 1, orig_log_lines[i])
+    assert different == any(missing)
+    return any(missing)

tools/tester.py

@@ -535,7 +535,8 @@ def main(orig_dir):
 
     TEST_DIR = os.path.join(orig_dir, opts.test_dir)
     EXTMODULE_DIR_PYSTON = os.path.abspath(os.path.dirname(os.path.realpath(IMAGE)) + "/test/test_extension/")
-    EXTMODULE_DIR = os.path.abspath(os.path.dirname(os.path.realpath(IMAGE)) + "/test/test_extension/build/lib.linux-x86_64-2.7/")
+    # EXTMODULE_DIR = os.path.abspath(os.path.dirname(os.path.realpath(IMAGE)) + "/test/test_extension/build/lib.linux-x86_64-2.7/")
+    EXTMODULE_DIR = os.path.abspath(orig_dir) + "/test/test_extension/build/lib.linux-x86_64-2.7/"
     patterns = opts.pattern
 
     IS_OPTIMIZED = int(subprocess.check_output([IMAGE, "-c", 'import sysconfig; print int("-O0" not in sysconfig.get_config_var(\"CFLAGS\"))']))