run-tests.py works properly with python3 (#1285)

Fixes #1180 and #1181.

Author: binji

.gitattributes

@@ -14,9 +14,10 @@ demo/libwabt.js binary
 demo/third_party/codemirror/codemirror.css binary
 demo/third_party/codemirror/codemirror.js binary
 
-# Mark this test as binary so git doesn't change the line endings, since this
-# test is testing \r\n.
+# Mark these tests as binary so git doesn't change the line endings:
 test/parse/bad-crlf.txt binary
+test/parse/bad-string-eof.txt binary
+test/regress/regress-31.txt binary
 
 # Highlight tests like .wast files when displayed on GitHub.
 test/**/*.txt linguist-language=WebAssembly

.github/workflows/build.yml

@@ -9,10 +9,16 @@ jobs:
     strategy:
       matrix:
         os: [ubuntu-16.04, macos-latest, windows-latest]
+        python: [2.7, '3.x']
+        exclude:
+          - os: macos-latest
+            python: '3.x'
+          - os: windows-latest
+            python: '3.x'
     steps:
     - uses: actions/setup-python@v1
       with:
-        python-version: '2.7'
+        python-version: ${{ matrix.python }}
     - uses: actions/checkout@v1
       with:
         submodules: true

test/run-spec-wasm2c.py

@@ -95,13 +95,14 @@ def MangleTypes(types):
 
 
 def MangleName(s):
-    result = 'Z_'
-    for c in s.encode('utf-8'):
-        # NOTE(binji): Z is not allowed.
-        if c in '_abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXY0123456789':
-            result += c
-        else:
-            result += 'Z%02X' % ord(c)
+    def Mangle(match):
+        s = match.group(0)
+        c = ord(s) if isinstance(s, str) else s[0]  # Python2 vs Python3
+        return b'Z%02X' % c
+
+    # NOTE(binji): Z is not allowed.
+    pattern = b'([^_a-zA-Y0-9])'
+    result = 'Z_' + re.sub(pattern, Mangle, s.encode('utf-8')).decode('utf-8')
     return result
 
 

test/run-tests.py

@@ -162,8 +162,11 @@ def Indent(s, spaces):
 
 
 def DiffLines(expected, actual):
-    expected_lines = [line for line in expected.splitlines() if line]
-    actual_lines = [line for line in actual.splitlines() if line]
+    def Decode(s):
+        return s.decode('utf-8', 'replace')
+
+    expected_lines = [Decode(line) for line in expected.splitlines() if line]
+    actual_lines = [Decode(line) for line in actual.splitlines() if line]
     return list(
         difflib.unified_diff(expected_lines, actual_lines, fromfile='expected',
                              tofile='actual', lineterm=''))
@@ -268,7 +271,7 @@ def KillProcess(timeout=True):
             process = subprocess.Popen(self.args, cwd=cwd, env=env,
                                        stdout=None if console_out else subprocess.PIPE,
                                        stderr=None if console_out else subprocess.PIPE,
-                                       universal_newlines=True, **kwargs)
+                                       **kwargs)
             timer = threading.Timer(timeout, KillProcess)
             try:
                 timer.start()
@@ -315,8 +318,8 @@ class TestResult(object):
 
     def __init__(self):
         self.results = []
-        self.stdout = ''
-        self.stderr = ''
+        self.stdout = b''
+        self.stderr = b''
         self.duration = 0
 
     def GetLastCommand(self):
@@ -479,7 +482,8 @@ def Parse(self, filename):
         self.filename = filename
 
         test_path = os.path.join(REPO_ROOT_DIR, filename)
-        with open(test_path) as f:
+        # Read/write as binary because spec tests may have invalid UTF-8 codes.
+        with open(test_path, 'rb') as f:
             state = 'header'
             empty = True
             header_lines = []
@@ -488,19 +492,29 @@ def Parse(self, filename):
             stderr_lines = []
             for line in f.readlines():
                 empty = False
-                m = re.match(r'\s*\(;; (STDOUT|STDERR) ;;;$', line)
+                m = re.match(b'\\s*\\(;; (STDOUT|STDERR) ;;;$', line.strip())
                 if m:
-                    directive = m.group(1)
+                    directive = m.group(1).decode('utf-8')
                     if directive == 'STDERR':
                         state = 'stderr'
                         continue
                     elif directive == 'STDOUT':
                         state = 'stdout'
                         continue
                 else:
-                    m = re.match(r'\s*;;;(.*)$', line)
+                    m = re.match(b'\\s*;;;(.*)$', line)
                     if m:
-                        directive = m.group(1).strip()
+                        # The matched string has type bytes, but in python2
+                        # that is the same as str. In python3 that needs to be
+                        # decoded first. If we decode the string in python2 the
+                        # result is a unicode string, which doesn't work
+                        # everywhere (as used in a subprocess environment, for
+                        # example).
+                        if sys.version_info.major == 3:
+                            directive = m.group(1).decode('utf-8').strip()
+                        else:
+                            directive = m.group(1).strip()
+
                         if state == 'header':
                             key, value = directive.split(':', 1)
                             key = key.strip()
@@ -517,7 +531,7 @@ def Parse(self, filename):
                         state = 'input'
 
                 if state == 'header':
-                    header_lines.append(line)
+                    header_lines.append(line.decode('utf-8'))
                 if state == 'input':
                     if self.input_filename:
                         raise Error('Can\'t have STDIN_FILE and input')
@@ -530,9 +544,9 @@ def Parse(self, filename):
             raise Error('empty test file')
 
         self.header = ''.join(header_lines)
-        self.input_ = ''.join(input_lines)
-        self.expected_stdout = ''.join(stdout_lines)
-        self.expected_stderr = ''.join(stderr_lines)
+        self.input_ = b''.join(input_lines)
+        self.expected_stdout = b''.join(stdout_lines)
+        self.expected_stderr = b''.join(stderr_lines)
 
         if not self.cmds:
             raise Error('test has no commands')
@@ -555,23 +569,23 @@ def CreateInputFile(self):
             else:
                 # add an empty line for each header line so the line numbers match
                 gen_input_file.write(('\n' * self.header.count('\n')).encode('ascii'))
-                gen_input_file.write(self.input_.encode('ascii'))
+                gen_input_file.write(self.input_)
             gen_input_file.flush()
             return gen_input_file.name
 
     def Rebase(self, stdout, stderr):
         test_path = os.path.join(REPO_ROOT_DIR, self.filename)
         with open(test_path, 'wb') as f:
-            f.write(self.header)
+            f.write(self.header.encode('ascii'))
             f.write(self.input_)
             if stderr:
-                f.write('(;; STDERR ;;;\n')
+                f.write(b'(;; STDERR ;;;\n')
                 f.write(stderr)
-                f.write(';;; STDERR ;;)\n')
+                f.write(b';;; STDERR ;;)\n')
             if stdout:
-                f.write('(;; STDOUT ;;;\n')
+                f.write(b'(;; STDOUT ;;;\n')
                 f.write(stdout)
-                f.write(';;; STDOUT ;;)\n')
+                f.write(b';;; STDOUT ;;)\n')
 
     def Diff(self, stdout, stderr):
         msg = ''