| """test script for a few new invalid token catches""" | |
| import sys | |
| from codecs import BOM_UTF8 | |
| from test.support import force_not_colorized | |
| from test.support import os_helper | |
| from test.support import script_helper | |
| from test.support import warnings_helper | |
| import unittest | |
| class EOFTestCase(unittest.TestCase): | |
| def test_EOF_single_quote(self): | |
| expect = "unterminated string literal (detected at line 1) (<string>, line 1)" | |
| for quote in ("'", "\""): | |
| with self.assertRaises(SyntaxError) as cm: | |
| eval(f"""{quote}this is a test\ | |
| """) | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.offset, 1) | |
| def test_EOFS(self): | |
| expect = ("unterminated triple-quoted string literal (detected at line 3) (<string>, line 1)") | |
| with self.assertRaises(SyntaxError) as cm: | |
| eval("""ä = '''thîs is \na \ntest""") | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.text, "ä = '''thîs is ") | |
| self.assertEqual(cm.exception.offset, 5) | |
| with self.assertRaises(SyntaxError) as cm: | |
| eval("""ä = '''thîs is \na \ntest""".encode()) | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.text, "ä = '''thîs is ") | |
| self.assertEqual(cm.exception.offset, 5) | |
| with self.assertRaises(SyntaxError) as cm: | |
| eval(BOM_UTF8 + """ä = '''thîs is \na \ntest""".encode()) | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.text, "ä = '''thîs is ") | |
| self.assertEqual(cm.exception.offset, 5) | |
| with self.assertRaises(SyntaxError) as cm: | |
| eval("""# coding: latin1\nä = '''thîs is \na \ntest""".encode('latin1')) | |
| self.assertEqual(str(cm.exception), "unterminated triple-quoted string literal (detected at line 4) (<string>, line 2)") | |
| self.assertEqual(cm.exception.text, "ä = '''thîs is ") | |
| self.assertEqual(cm.exception.offset, 5) | |
| @force_not_colorized | |
| def test_EOFS_with_file(self): | |
| expect = ("(<string>, line 1)") | |
| with os_helper.temp_dir() as temp_dir: | |
| file_name = script_helper.make_script(temp_dir, 'foo', | |
| """ä = '''thîs is \na \ntest""") | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| " ä = '''thîs is ", | |
| ' ^', | |
| 'SyntaxError: unterminated triple-quoted string literal (detected at line 3)']) | |
| file_name = script_helper.make_script(temp_dir, 'foo', | |
| """ä = '''thîs is \na \ntest""".encode()) | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| " ä = '''thîs is ", | |
| ' ^', | |
| 'SyntaxError: unterminated triple-quoted string literal (detected at line 3)']) | |
| file_name = script_helper.make_script(temp_dir, 'foo', | |
| BOM_UTF8 + """ä = '''thîs is \na \ntest""".encode()) | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| " ä = '''thîs is ", | |
| ' ^', | |
| 'SyntaxError: unterminated triple-quoted string literal (detected at line 3)']) | |
| file_name = script_helper.make_script(temp_dir, 'foo', | |
| """# coding: latin1\nä = '''thîs is \na \ntest""".encode('latin1')) | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| " ä = '''thîs is ", | |
| ' ^', | |
| 'SyntaxError: unterminated triple-quoted string literal (detected at line 4)']) | |
| @warnings_helper.ignore_warnings(category=SyntaxWarning) | |
| def test_eof_with_line_continuation(self): | |
| expect = "unexpected EOF while parsing (<string>, line 1)" | |
| with self.assertRaises(SyntaxError) as cm: | |
| compile('"\\Xhh" \\', '<string>', 'exec') | |
| self.assertEqual(str(cm.exception), expect) | |
| def test_line_continuation_EOF(self): | |
| """A continuation at the end of input must be an error; bpo2180.""" | |
| expect = 'unexpected EOF while parsing (<string>, line 1)' | |
| with self.assertRaises(SyntaxError) as cm: | |
| exec('ä = 5\\') | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.text, 'ä = 5\\\n') | |
| self.assertEqual(cm.exception.offset, 7) | |
| with self.assertRaises(SyntaxError) as cm: | |
| exec('ä = 5\\'.encode()) | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.text, 'ä = 5\\\n') | |
| self.assertEqual(cm.exception.offset, 7) | |
| with self.assertRaises(SyntaxError) as cm: | |
| exec('# coding:latin1\nä = 5\\'.encode('latin1')) | |
| self.assertEqual(str(cm.exception), | |
| 'unexpected EOF while parsing (<string>, line 2)') | |
| self.assertEqual(cm.exception.text, 'ä = 5\\\n') | |
| self.assertEqual(cm.exception.offset, 7) | |
| with self.assertRaises(SyntaxError) as cm: | |
| exec(BOM_UTF8 + 'ä = 5\\'.encode()) | |
| self.assertEqual(str(cm.exception), expect) | |
| self.assertEqual(cm.exception.text, 'ä = 5\\\n') | |
| self.assertEqual(cm.exception.offset, 7) | |
| with self.assertRaises(SyntaxError) as cm: | |
| exec('\\') | |
| self.assertEqual(str(cm.exception), expect) | |
| @unittest.skipIf(not sys.executable, "sys.executable required") | |
| @force_not_colorized | |
| def test_line_continuation_EOF_from_file_bpo2180(self): | |
| """Ensure tok_nextc() does not add too many ending newlines.""" | |
| with os_helper.temp_dir() as temp_dir: | |
| file_name = script_helper.make_script(temp_dir, 'foo', '\\') | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-2:], [ | |
| ' \\', | |
| 'SyntaxError: unexpected EOF while parsing']) | |
| self.assertEqual(err[-3][-8:], ', line 1', err) | |
| file_name = script_helper.make_script(temp_dir, 'foo', 'ä = 6\\') | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| ' ä = 6\\', | |
| ' ^', | |
| 'SyntaxError: unexpected EOF while parsing']) | |
| self.assertEqual(err[-4][-8:], ', line 1', err) | |
| file_name = script_helper.make_script(temp_dir, 'foo', | |
| '# coding:latin1\n' | |
| 'ä = 7\\'.encode('latin1')) | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| ' ä = 7\\', | |
| ' ^', | |
| 'SyntaxError: unexpected EOF while parsing']) | |
| self.assertEqual(err[-4][-8:], ', line 2', err) | |
| file_name = script_helper.make_script(temp_dir, 'foo', | |
| BOM_UTF8 + 'ä = 8\\'.encode()) | |
| rc, out, err = script_helper.assert_python_failure('-X', 'utf8', file_name) | |
| err = err.decode().splitlines() | |
| self.assertEqual(err[-3:], [ | |
| ' ä = 8\\', | |
| ' ^', | |
| 'SyntaxError: unexpected EOF while parsing']) | |
| self.assertEqual(err[-4][-8:], ', line 1', err) | |
| if __name__ == "__main__": | |
| unittest.main() |