Update cpplint.py to #150:...

cpplint/cpplint.py

 #!/usr/bin/python2.4
 #
 # Copyright (c) 2009 Google Inc. All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions are
 # met:
 #
 #    * Redistributions of source code must retain the above copyright
 # notice, this list of conditions and the following disclaimer.
@@ skipping 84 matching lines @@
                    [--counting=total|toplevel|detailed]
         <file> [file] ...
 
   The style guidelines this tries to follow are those in
     http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml
 
   Every problem is given a confidence score from 1-5, with 5 meaning we are
   certain of the problem, and 1 meaning it could be a legitimate construct.
   This will miss some errors, and is not a substitute for a code review.
 
-  To prevent specific lines from being linted, add a '// NOLINT' comment to the
-  end of the line.
+  To suppress false-positive errors of a certain category, add a
+  'NOLINT(category)' comment to the line.  NOLINT or NOLINT(*)
+  suppresses errors of all categories on that line.
 
   The files passed in will be linted; at least one file must be provided.
   Linted extensions are .cc, .cpp, and .h.  Other file types will be ignored.
 
   Flags:
 
     output=vs7
       By default, the output is formatted to ease emacs parsing.  Visual Studio
       compatible output (vs7) may also be used.  Other formats are unsupported.
 
@@ skipping 21 matching lines @@
       the top-level categories like 'build' and 'whitespace' will
       also be printed. If 'detailed' is provided, then a count
       is provided for each category like 'build/class'.
 """
 
 # We categorize each error message we print.  Here are the categories.
 # We want an explicit list so we can list them all in cpplint --filter=.
 # If you add a new error message with a new category, add it to the list
 # here!  cpplint_unittest.py should tell you if you forget to do this.
 # \ used for clearer layout -- pylint: disable-msg=C6013
-_ERROR_CATEGORIES = '''\
-  build/class
-  build/deprecated
-  build/endif_comment
-  build/forward_decl
-  build/header_guard
-  build/include
-  build/include_alpha
-  build/include_order
-  build/include_what_you_use
-  build/namespaces
-  build/printf_format
-  build/storage_class
-  legal/copyright
-  readability/braces
-  readability/casting
-  readability/check
-  readability/constructors
-  readability/fn_size
-  readability/function
-  readability/multiline_comment
-  readability/multiline_string
-  readability/streams
-  readability/todo
-  readability/utf8
-  runtime/arrays
-  runtime/casting
-  runtime/explicit
-  runtime/int
-  runtime/init
-  runtime/invalid_increment
-  runtime/member_string_references
-  runtime/memset
-  runtime/operator
-  runtime/printf
-  runtime/printf_format
-  runtime/references
-  runtime/rtti
-  runtime/sizeof
-  runtime/string
-  runtime/threadsafe_fn
-  runtime/virtual
-  whitespace/blank_line
-  whitespace/braces
-  whitespace/comma
-  whitespace/comments
-  whitespace/end_of_line
-  whitespace/ending_newline
-  whitespace/indent
-  whitespace/labels
-  whitespace/line_length
-  whitespace/newline
-  whitespace/operators
-  whitespace/parens
-  whitespace/semicolon
-  whitespace/tab
-  whitespace/todo
-'''
+_ERROR_CATEGORIES = [
+  'build/class',
+  'build/deprecated',
+  'build/endif_comment',
+  'build/forward_decl',
+  'build/header_guard',
+  'build/include',
+  'build/include_alpha',
+  'build/include_order',
+  'build/include_what_you_use',
+  'build/namespaces',
+  'build/printf_format',
+  'build/storage_class',
+  'legal/copyright',
+  'readability/braces',
+  'readability/casting',
+  'readability/check',
+  'readability/constructors',
+  'readability/fn_size',
+  'readability/function',
+  'readability/multiline_comment',
+  'readability/multiline_string',
+  'readability/nolint',
+  'readability/streams',
+  'readability/todo',
+  'readability/utf8',
+  'runtime/arrays',
+  'runtime/casting',
+  'runtime/explicit',
+  'runtime/int',
+  'runtime/init',
+  'runtime/invalid_increment',
+  'runtime/member_string_references',
+  'runtime/memset',
+  'runtime/operator',
+  'runtime/printf',
+  'runtime/printf_format',
+  'runtime/references',
+  'runtime/rtti',
+  'runtime/sizeof',
+  'runtime/string',
+  'runtime/threadsafe_fn',
+  'runtime/virtual',
+  'whitespace/blank_line',
+  'whitespace/braces',
+  'whitespace/comma',
+  'whitespace/comments',
+  'whitespace/end_of_line',
+  'whitespace/ending_newline',
+  'whitespace/indent',
+  'whitespace/labels',
+  'whitespace/line_length',
+  'whitespace/newline',
+  'whitespace/operators',
+  'whitespace/parens',
+  'whitespace/semicolon',
+  'whitespace/tab',
+  'whitespace/todo'
+  ]
 
 # The default state of the category filter. This is overrided by the --filter=
 # flag. By default all errors are on, so only add here categories that should be
 # off by default (i.e., categories that must be enabled by the --filter= flags).
 # All entries here should start with a '-' or '+', as in the --filter= flag.
 _DEFAULT_FILTERS = [ '-build/include_alpha' ]
 
 # We used to check for high-bit characters, but after much discussion we
 # decided those were OK, as long as they were in UTF-8 and didn't represent
 # hard-coded international strings, which belong in a seperate i18n file.
 
 # Headers that we consider STL headers.
 _STL_HEADERS = frozenset([
     'algobase.h', 'algorithm', 'alloc.h', 'bitset', 'deque', 'exception',
     'function.h', 'functional', 'hash_map', 'hash_map.h', 'hash_set',
-    'hash_set.h', 'iterator', 'list', 'list.h', 'map', 'memory', 'pair.h',
-    'pthread_alloc', 'queue', 'set', 'set.h', 'sstream', 'stack',
+    'hash_set.h', 'iterator', 'list', 'list.h', 'map', 'memory', 'new',
+    'pair.h', 'pthread_alloc', 'queue', 'set', 'set.h', 'sstream', 'stack',
     'stl_alloc.h', 'stl_relops.h', 'type_traits.h',
     'utility', 'vector', 'vector.h',
     ])
 
 
 # Non-STL C++ system headers.
 _CPP_HEADERS = frozenset([
     'algo.h', 'builtinbuf.h', 'bvector.h', 'cassert', 'cctype',
     'cerrno', 'cfloat', 'ciso646', 'climits', 'clocale', 'cmath',
     'complex', 'complex.h', 'csetjmp', 'csignal', 'cstdarg', 'cstddef',
@@ skipping 47 matching lines @@
 # _IncludeState.CheckNextIncludeOrder().
 _C_SYS_HEADER = 1
 _CPP_SYS_HEADER = 2
 _LIKELY_MY_HEADER = 3
 _POSSIBLE_MY_HEADER = 4
 _OTHER_HEADER = 5
 
 
 _regexp_compile_cache = {}
 
+# Finds occurrences of NOLINT or NOLINT(...).
+_RE_SUPPRESSION = re.compile(r'\bNOLINT\b(\([^)]*\))?')
+
+# {str, set(int)}: a map from error categories to sets of linenumbers
+# on which those errors are expected and should be suppressed.
+_error_suppressions = {}
+
+def ParseNolintSuppressions(filename, raw_line, linenum, error):
+  """Updates the global list of error-suppressions.
+
+  Parses any NOLINT comments on the current line, updating the global
+  error_suppressions store.  Reports an error if the NOLINT comment
+  was malformed.
+
+  Args:
+    filename: str, the name of the input file.
+    raw_line: str, the line of input text, with comments.
+    linenum: int, the number of the current line.
+    error: function, an error handler.
+  """
+  # FIXME(adonovan): "NOLINT(" is misparsed as NOLINT(*).
+  m = _RE_SUPPRESSION.search(raw_line)
+  if m:
+    category = m.group(1)
+    if category in (None, '(*)'):  # => "suppress all"
+      _error_suppressions.setdefault(None, set()).add(linenum)
+    else:
+      if category.startswith('(') and category.endswith(')'):
+        category = category[1:-1]
+        if category in _ERROR_CATEGORIES:
+          _error_suppressions.setdefault(category, set()).add(linenum)
+        else:
+          error(filename, linenum, 'readability/nolint', 5,
+            'Unknown NOLINT error category: %s' % category)
+
+
+def ResetNolintSuppressions():
+  "Resets the set of NOLINT suppressions to empty."
+  _error_suppressions.clear()
+
+
+def IsErrorSuppressedByNolint(category, linenum):
+  """Returns true if the specified error category is suppressed on this line.
+
+  Consults the global error_suppressions map populated by
+  ParseNolintSuppressions/ResetNolintSuppressions.
+
+  Args:
+    category: str, the category of the error.
+    linenum: int, the current line number.
+  Returns:
+    bool, True iff the error should be suppressed due to a NOLINT comment.
+  """
+  return (linenum in _error_suppressions.get(category, set()) or
+          linenum in _error_suppressions.get(None, set()))
 
 def Match(pattern, s):
   """Matches the string with the pattern, caching the compiled regexp."""
   # The regexp compilation caching is inlined in both Match and Search for
   # performance reasons; factoring it out into a separate function turns out
   # to be noticeably expensive.
   if not pattern in _regexp_compile_cache:
     _regexp_compile_cache[pattern] = sre_compile.compile(pattern)
   return _regexp_compile_cache[pattern].match(s)
 
@@ skipping 394 matching lines @@
 
   def NoExtension(self):
     """File has no source file extension."""
     return '/'.join(self.Split()[0:2])
 
   def IsSource(self):
     """File has a source file extension."""
     return self.Extension()[1:] in ('c', 'cc', 'cpp', 'cxx')
 
 
-def _ShouldPrintError(category, confidence):
-  """Returns true iff confidence >= verbose, and category passes filter."""
-  # There are two ways we might decide not to print an error message:
+def _ShouldPrintError(category, confidence, linenum):
+  """Returns true iff confidence >= verbose, category passes
+  filter and is not NOLINT-suppressed."""
+
+  # There are three ways we might decide not to print an error message:
+  # a "NOLINT(category)" comment appears in the source,
   # the verbosity level isn't high enough, or the filters filter it out.
+  if IsErrorSuppressedByNolint(category, linenum):
+    return False
   if confidence < _cpplint_state.verbose_level:
     return False
 
   is_filtered = False
   for one_filter in _Filters():
     if one_filter.startswith('-'):
       if category.startswith(one_filter[1:]):
         is_filtered = True
     elif one_filter.startswith('+'):
       if category.startswith(one_filter[1:]):
         is_filtered = False
     else:
       assert False  # should have been checked for in SetFilter.
   if is_filtered:
     return False
 
   return True
 
 
 def Error(filename, linenum, category, confidence, message):
   """Logs the fact we've found a lint error.
 
   We log where the error was found, and also our confidence in the error,
   that is, how certain we are this is a legitimate style regression, and
   not a misidentification or a use that's sometimes justified.
 
+  False positives can be suppressed by the use of
+  "cpplint(category)"  comments on the offending line.  These are
+  parsed into _error_suppressions.
+
   Args:
     filename: The name of the file containing the error.
     linenum: The number of the line containing the error.
     category: A string used to describe the "category" this bug
       falls under: "whitespace", say, or "runtime".  Categories
       may have a hierarchy separated by slashes: "whitespace/indent".
     confidence: A number from 1-5 representing a confidence score for
       the error, with 5 meaning that we are certain of the problem,
       and 1 meaning that it could be a legitimate construct.
     message: The error message.
   """
-  # There are two ways we might decide not to print an error message:
-  # the verbosity level isn't high enough, or the filters filter it out.
-  if _ShouldPrintError(category, confidence):
+  if _ShouldPrintError(category, confidence, linenum):
     _cpplint_state.IncrementErrorCount(category)
     if _cpplint_state.output_format == 'vs7':
       sys.stderr.write('%s(%s):  %s  [%s] [%d]\n' % (
           filename, linenum, message, category, confidence))
     else:
       sys.stderr.write('%s:%s:  %s  [%s] [%d]\n' % (
           filename, linenum, message, category, confidence))
 
 
 # Matches standard C++ escape esequences per 2.13.2.3 of the C++ standard.
@@ skipping 197 matching lines @@
 
   Args:
     filename: The name of a C++ header file.
 
   Returns:
     The CPP variable that should be used as a header guard in the
     named file.
 
   """
 
+  # Restores original filename in case that cpplint is invoked from Emacs's
+  # flymake.
+  filename = re.sub(r'_flymake\.h$', '.h', filename)
+
   fileinfo = FileInfo(filename)
   return re.sub(r'[-./\s]', '_', fileinfo.RepositoryName()).upper() + '_'
 
 
 def CheckForHeaderGuard(filename, lines, error):
   """Checks that the file contains a header guard.
 
   Logs an error if no #ifndef header guard is present.  For other
   headers, checks that the full pathname is used.
 
@@ skipping 26 matching lines @@
       endif_linenum = linenum
 
   if not ifndef or not define or ifndef != define:
     error(filename, 0, 'build/header_guard', 5,
           'No #ifndef header guard found, suggested CPP variable is: %s' %
           cppvar)
     return
 
   # The guard should be PATH_FILE_H_, but we also allow PATH_FILE_H__
   # for backward compatibility.
-  if ifndef != cppvar and not Search(r'\bNOLINT\b', lines[ifndef_linenum]):
+  if ifndef != cppvar:
     error_level = 0
     if ifndef != cppvar + '_':
       error_level = 5
 
+    ParseNolintSuppressions(filename, lines[ifndef_linenum], ifndef_linenum,
+                            error)
     error(filename, ifndef_linenum, 'build/header_guard', error_level,
           '#ifndef header guard has wrong style, please use: %s' % cppvar)
 
-  if (endif != ('#endif  // %s' % cppvar) and
-      not Search(r'\bNOLINT\b', lines[endif_linenum])):
+  if endif != ('#endif  // %s' % cppvar):
     error_level = 0
     if endif != ('#endif  // %s' % (cppvar + '_')):
       error_level = 5
 
+    ParseNolintSuppressions(filename, lines[endif_linenum], endif_linenum,
+                            error)
     error(filename, endif_linenum, 'build/header_guard', error_level,
           '#endif line should be "#endif  // %s"' % cppvar)
 
 
 def CheckForUnicodeReplacementCharacters(filename, lines, error):
   """Logs an error for each line containing Unicode replacement characters.
 
   These indicate that either the file contained invalid UTF-8 (likely)
   or Unicode replacement characters (which it shouldn't).  Note that
   it's possible for this to throw off line numbering if the invalid
@@ skipping 477 matching lines @@
             function += parameter_regexp.group(1)
         else:
           function += '()'
         function_state.Begin(function)
         break
     if not body_found:
       # No body for the function (or evidence of a non-function) was found.
       error(filename, linenum, 'readability/fn_size', 5,
             'Lint failed to find start of function body.')
   elif Match(r'^\}\s*$', line):  # function end
-    if not Search(r'\bNOLINT\b', raw_line):
-      function_state.Check(error, filename, linenum)
+    function_state.Check(error, filename, linenum)
     function_state.End()
   elif not Match(r'^\s*$', line):
     function_state.Count()  # Count non-blank/non-comment lines.
 
 
 _RE_PATTERN_TODO = re.compile(r'^//(\s*)TODO(\(.+?\))?:?(\s|$)?')
 
 
 def CheckComment(comment, filename, linenum, error):
   """Checks for common mistakes in TODO comments.
@@ skipping 486 matching lines @@
   elif ((initial_spaces == 1 or initial_spaces == 3) and
         not Match(r'\s*\w+\s*:\s*$', cleansed_line)):
     error(filename, linenum, 'whitespace/indent', 3,
           'Weird number of spaces at line-start.  '
           'Are you using a 2-space indent?')
   # Labels should always be indented at least one space.
   elif not initial_spaces and line[:2] != '//' and Search(r'[^:]:\s*$',
                                                           line):
     error(filename, linenum, 'whitespace/labels', 4,
           'Labels should always be indented at least one space.  '
-          'If this is a member-initializer list in a constructor, '
-          'the colon should be on the line after the definition header.')
+          'If this is a member-initializer list in a constructor or '
+          'the base class list in a class definition, the colon should '
+          'be on the following line.')
+
 
   # Check if the line is a header guard.
   is_header_guard = False
   if file_extension == 'h':
     cppvar = GetHeaderGuardCPPVariable(filename)
     if (line.startswith('#ifndef %s' % cppvar) or
         line.startswith('#define %s' % cppvar) or
         line.startswith('#endif  // %s' % cppvar)):
       is_header_guard = True
   # #include lines and header guards can be long, since there's no clean way to
@@ skipping 352 matching lines @@
       error(filename, linenum, 'runtime/int', 4,
             'Use "unsigned short" for ports, not "short"')
   else:
     match = Search(r'\b(short|long(?! +double)|long long)\b', line)
     if match:
       error(filename, linenum, 'runtime/int', 4,
             'Use int16/int64/etc, rather than the C type %s' % match.group(1))
 
   # When snprintf is used, the second argument shouldn't be a literal.
   match = Search(r'snprintf\s*\(([^,]*),\s*([0-9]*)\s*,', line)
-  if match:
+  if match and match.group(2) != '0':
+    # If 2nd arg is zero, snprintf is used to calculate size.
     error(filename, linenum, 'runtime/printf', 3,
           'If you can, use sizeof(%s) instead of %s as the 2nd arg '
           'to snprintf.' % (match.group(1), match.group(2)))
 
   # Check if some verboten C functions are being used.
   if Search(r'\bsprintf\b', line):
     error(filename, linenum, 'runtime/printf', 5,
           'Never use sprintf.  Use snprintf instead.')
   match = Search(r'\b(strcpy|strcat)\b', line)
   if match:
@@ skipping 328 matching lines @@
   """
   required = {}  # A map of header name to linenumber and the template entity.
                  # Example of required: { '<functional>': (1219, 'less<>') }
 
   for linenum in xrange(clean_lines.NumLines()):
     line = clean_lines.elided[linenum]
     if not line or line[0] == '#':
       continue
 
     # String is special -- it is a non-templatized type in STL.
-    if _RE_PATTERN_STRING.search(line):
-      required['<string>'] = (linenum, 'string')
+    m = _RE_PATTERN_STRING.search(line)
+    if m:
+      # Don't warn about strings in non-STL namespaces:
+      # (We check only the first match per line; good enough.)
+      prefix = line[:m.start()]
+      if prefix.endswith('std::') or not prefix.endswith('::'):
+        required['<string>'] = (linenum, 'string')
 
     for pattern, template, header in _re_pattern_algorithm_header:
       if pattern.search(line):
         required[header] = (linenum, template)
 
     # The following function is just a speed up, no semantics are changed.
     if not '<' in line:  # Reduces the cpu time usage by skipping lines.
       continue
 
     for pattern, template, header in _re_pattern_templates:
@@ skipping 11 matching lines @@
   # Use the absolute path so that matching works properly.
   abs_filename = os.path.abspath(filename)
 
   # For Emacs's flymake.
   # If cpplint is invoked from Emacs's flymake, a temporary file is generated
   # by flymake and that file name might end with '_flymake.cc'. In that case,
   # restore original file name here so that the corresponding header file can be
   # found.
   # e.g. If the file name is 'foo_flymake.cc', we should search for 'foo.h'
   # instead of 'foo_flymake.h'
-  emacs_flymake_suffix = '_flymake.cc'
-  if abs_filename.endswith(emacs_flymake_suffix):
-    abs_filename = abs_filename[:-len(emacs_flymake_suffix)] + '.cc'
+  abs_filename = re.sub(r'_flymake\.cc$', '.cc', abs_filename)
 
   # include_state is modified during iteration, so we iterate over a copy of
   # the keys.
   for header in include_state.keys():  #NOLINT
     (same_module, common_path) = FilesBelongToSameModule(abs_filename, header)
     fullpath = common_path + header
     if same_module and UpdateIncludeState(fullpath, include_state, io):
       header_found = True
 
   # If we can't find the header file for a .cc, assume it's because we don't
@@ skipping 30 matching lines @@
     line: Number of line being processed.
     include_state: An _IncludeState instance in which the headers are inserted.
     function_state: A _FunctionState instance which counts function lines, etc.
     class_state: A _ClassState instance which maintains information about
                  the current stack of nested class declarations being parsed.
     error: A callable to which errors are reported, which takes 4 arguments:
            filename, line number, error level, and message
 
   """
   raw_lines = clean_lines.raw_lines
+  ParseNolintSuppressions(filename, raw_lines[line], line, error)
   CheckForFunctionLengths(filename, clean_lines, line, function_state, error)
-  if Search(r'\bNOLINT\b', raw_lines[line]):  # ignore nolint lines
-    return
   CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error)
   CheckStyle(filename, clean_lines, line, file_extension, error)
   CheckLanguage(filename, clean_lines, line, file_extension, include_state,
                 error)
   CheckForNonStandardConstructs(filename, clean_lines, line,
                                 class_state, error)
   CheckPosixThreading(filename, clean_lines, line, error)
   CheckInvalidIncrement(filename, clean_lines, line, error)
 
 
 def ProcessFileData(filename, file_extension, lines, error):
   """Performs lint checks and reports any errors to the given error function.
 
   Args:
     filename: Filename of the file that is being processed.
     file_extension: The extension (dot not included) of the file.
     lines: An array of strings, each representing a line of the file, with the
            last element being empty if the file is termined with a newline.
     error: A callable to which errors are reported, which takes 4 arguments:
   """
   lines = (['// marker so line numbers and indices both start at 1'] + lines +
            ['// marker so line numbers end in a known way'])
 
   include_state = _IncludeState()
   function_state = _FunctionState()
   class_state = _ClassState()
 
+  ResetNolintSuppressions()
+
   CheckForCopyright(filename, lines, error)
 
   if file_extension == 'h':
     CheckForHeaderGuard(filename, lines, error)
 
   RemoveMultiLineComments(filename, lines, error)
   clean_lines = CleansedLines(lines)
   for line in xrange(clean_lines.NumLines()):
     ProcessLine(filename, file_extension, clean_lines, line,
                 include_state, function_state, class_state, error)
   class_state.CheckFinished(filename, error)
 
   CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error)
 
   # We check here rather than inside ProcessLine so that we see raw
   # lines rather than "cleaned" lines.
   CheckForUnicodeReplacementCharacters(filename, lines, error)
 
   CheckForNewlineAtEOF(filename, lines, error)
 
-
 def ProcessFile(filename, vlevel):
   """Does google-lint on a single file.
 
   Args:
     filename: The name of the file to parse.
 
     vlevel: The level of errors to report.  Every error of confidence
     >= verbose_level will be reported.  0 is a good default.
   """
 
@@ skipping 61 matching lines @@
     sys.exit('\nFATAL ERROR: ' + message)
   else:
     sys.exit(1)
 
 
 def PrintCategories():
   """Prints a list of all the error-categories used by error messages.
 
   These are the categories used to filter messages via --filter.
   """
-  sys.stderr.write(_ERROR_CATEGORIES)
+  sys.stderr.write(''.join('  %s\n' % cat for cat in _ERROR_CATEGORIES))
   sys.exit(0)
 
 
 def ParseArguments(args):
   """Parses the command line arguments.
 
   This may set the output format and verbosity level as side-effects.
 
   Args:
     args: The command line arguments:
@@ skipping 55 matching lines @@
   _cpplint_state.ResetErrorCounts()
   for filename in filenames:
     ProcessFile(filename, _cpplint_state.verbose_level)
   _cpplint_state.PrintErrorCounts()
 
   sys.exit(_cpplint_state.error_count > 0)
 
 
 if __name__ == '__main__':
   main()