#!/usr/bin/env python
#
# 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.
# * Redistributions in binary form must reproduce the above
# copyright notice, this list of conditions and the following disclaimer
# in the documentation and/or other materials provided with the
# distribution.
# * Neither the name of Google Inc. nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
"""Does google-lint on c++ files.
The goal of this script is to identify places in the code that *may*
be in non-compliance with google style. It does not attempt to fix
up these problems -- the point is to educate. It does also not
attempt to find all problems, or to ensure that everything it does
find is legitimately a problem.
In particular, we can get very confused by /* and // inside strings!
We do a small hack, which is to ignore //'s with "'s after them on the
same line, but it is far from perfect (in either direction).
"""
import codecs
import copy
import getopt
import logging
import math # for log
import os
import re
import sre_compile
import string
import sys
import unicodedata
logger = logging.getLogger('testrunner')
_USAGE = """
Syntax: cpplint.py [--verbose=#] [--output=vs7] [--filter=-x,+y,...]
[--counting=total|toplevel|detailed] [--root=subdir]
[--linelength=digits] [--logfile=filename]
[file] ...
The style guidelines this tries to follow are those in
https://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 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.
Default linted extensions are .cc, .cpp, .cu, .cuh and .h. Change the
extensions with the --extensions flag.
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.
verbose=#
Specify a number 0-5 to restrict errors to certain verbosity levels.
filter=-x,+y,...
Specify a comma-separated list of category-filters to apply: only
error messages whose category names pass the filters will be printed.
(Category names are printed with the message and look like
"[whitespace/indent]".) Filters are evaluated left to right.
"-FOO" and "FOO" means "do not print categories that start with FOO".
"+FOO" means "do print categories that start with FOO".
Examples: --filter=-whitespace,+whitespace/braces
--filter=whitespace,runtime/printf,+runtime/printf_format
--filter=-,+build/include_what_you_use
To see a list of all the categories used in cpplint, pass no arg:
--filter=
counting=total|toplevel|detailed
The total number of errors found is always printed. If
'toplevel' is provided, then the count of errors in each of
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'.
root=subdir
The root directory used for deriving header guard CPP variable.
By default, the header guard CPP variable is calculated as the relative
path to the directory that contains .git, .hg, or .svn. When this flag
is specified, the relative path is calculated from the specified
directory. If the specified directory does not exist, this flag is
ignored.
Examples:
Assuming that src/.git exists, the header guard CPP variables for
src/chrome/browser/ui/browser.h are:
No flag => CHROME_BROWSER_UI_BROWSER_H_
--root=chrome => BROWSER_UI_BROWSER_H_
--root=chrome/browser => UI_BROWSER_H_
linelength=digits
This is the allowed line length for the project. The default value is
80 characters.
Examples:
--linelength=120
extensions=extension,extension,...
The allowed file extensions that cpplint will check
Examples:
--extensions=hpp,cpp
logfile=filename
Write TAP output to a logfile.
cpplint.py supports per-directory configurations specified in CPPLINT.cfg
files. CPPLINT.cfg file can contain a number of key=value pairs.
Currently the following options are supported:
set noparent
filter=+filter1,-filter2,...
exclude_files=regex
linelength=80
"set noparent" option prevents cpplint from traversing directory tree
upwards looking for more .cfg files in parent directories. This option
is usually placed in the top-level project directory.
The "filter" option is similar in function to --filter flag. It specifies
message filters in addition to the |_DEFAULT_FILTERS| and those specified
through --filter command-line flag.
"exclude_files" allows to specify a regular expression to be matched against
a file name. If the expression matches, the file is skipped and not run
through liner.
"linelength" allows to specify the allowed line length for the project.
CPPLINT.cfg has an effect on files in the same directory and all
sub-directories, unless overridden by a nested configuration file.
Example file:
filter=-build/include_order,+build/include_alpha
exclude_files=.*\.cc
The above example disables build/include_order warning and enables
build/include_alpha as well as excludes all .cc from being
processed by linter, in the current directory (where the .cfg
file is located) and all sub-directories.
"""
# 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.
_ERROR_CATEGORIES = [
'build/class',
'build/c++11',
'build/c++14',
'build/c++tr1',
'build/deprecated',
'build/endif_comment',
'build/explicit_make_pair',
'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/alt_tokens',
'readability/braces',
'readability/casting',
'readability/check',
'readability/constructors',
'readability/fn_size',
'readability/inheritance',
'readability/multiline_comment',
'readability/multiline_string',
'readability/namespace',
'readability/nolint',
'readability/nul',
'readability/strings',
'readability/todo',
'readability/utf8',
'runtime/arrays',
'runtime/casting',
'runtime/explicit',
'runtime/int',
'runtime/init',
'runtime/invalid_increment',
'runtime/member_string_references',
'runtime/memset',
'runtime/indentation_namespace',
'runtime/operator',
'runtime/printf',
'runtime/printf_format',
'runtime/references',
'runtime/string',
'runtime/threadsafe_fn',
'runtime/vlog',
'whitespace/blank_line',
'whitespace/braces',
'whitespace/comma',
'whitespace/comments',
'whitespace/empty_conditional_body',
'whitespace/empty_if_body',
'whitespace/empty_loop_body',
'whitespace/end_of_line',
'whitespace/ending_newline',
'whitespace/forcolon',
'whitespace/indent',
'whitespace/line_length',
'whitespace/newline',
'whitespace/operators',
'whitespace/parens',
'whitespace/semicolon',
'whitespace/tab',
'whitespace/todo',
]
# These error categories are no longer enforced by cpplint, but for backwards-
# compatibility they may still appear in NOLINT comments.
_LEGACY_ERROR_CATEGORIES = [
'readability/streams',
'readability/function',
]
# The default state of the category filter. This is overridden 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',
'-build/include_alpha',
'-build/include_order',
'-legal/copyright',
]
# The default list of categories suppressed for C (not C++) files.
_DEFAULT_C_SUPPRESSED_CATEGORIES = [
'readability/casting',
]
# The default list of categories suppressed for Linux Kernel files.
_DEFAULT_KERNEL_SUPPRESSED_CATEGORIES = [
'whitespace/tab',
]
# 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 separate i18n file.
# C++ headers
_CPP_HEADERS = frozenset([
# Legacy
'algobase.h',
'algo.h',
'alloc.h',
'builtinbuf.h',
'bvector.h',
'complex.h',
'defalloc.h',
'deque.h',
'editbuf.h',
'fstream.h',
'function.h',
'hash_map',
'hash_map.h',
'hash_set',
'hash_set.h',
'hashtable.h',
'heap.h',
'indstream.h',
'iomanip.h',
'iostream.h',
'istream.h',
'iterator.h',
'list.h',
'map.h',
'multimap.h',
'multiset.h',
'ostream.h',
'pair.h',
'parsestream.h',
'pfstream.h',
'procbuf.h',
'pthread_alloc',
'pthread_alloc.h',
'rope',
'rope.h',
'ropeimpl.h',
'set.h',
'slist',
'slist.h',
'stack.h',
'stdiostream.h',
'stl_alloc.h',
'stl_relops.h',
'streambuf.h',
'stream.h',
'strfile.h',
'strstream.h',
'tempbuf.h',
'tree.h',
'type_traits.h',
'vector.h',
# 17.6.1.2 C++ library headers
'algorithm',
'array',
'atomic',
'bitset',
'chrono',
'codecvt',
'complex',
'condition_variable',
'deque',
'exception',
'forward_list',
'fstream',
'functional',
'future',
'initializer_list',
'iomanip',
'ios',
'iosfwd',
'iostream',
'istream',
'iterator',
'limits',
'list',
'locale',
'map',
'memory',
'mutex',
'new',
'numeric',
'ostream',
'queue',
'random',
'ratio',
'regex',
'scoped_allocator',
'set',
'sstream',
'stack',
'stdexcept',
'streambuf',
'string',
'strstream',
'system_error',
'thread',
'tuple',
'typeindex',
'typeinfo',
'type_traits',
'unordered_map',
'unordered_set',
'utility',
'valarray',
'vector',
# 17.6.1.2 C++ headers for C library facilities
'cassert',
'ccomplex',
'cctype',
'cerrno',
'cfenv',
'cfloat',
'cinttypes',
'ciso646',
'climits',
'clocale',
'cmath',
'csetjmp',
'csignal',
'cstdalign',
'cstdarg',
'cstdbool',
'cstddef',
'cstdint',
'cstdio',
'cstdlib',
'cstring',
'ctgmath',
'ctime',
'cuchar',
'cwchar',
'cwctype',
])
# Type names
_TYPES = re.compile(
r'^(?:'
# [dcl.type.simple]
r'(char(16_t|32_t)?)|wchar_t|'
r'bool|short|int|long|signed|unsigned|float|double|'
# [support.types]
r'(ptrdiff_t|size_t|max_align_t|nullptr_t)|'
# [cstdint.syn]
r'(u?int(_fast|_least)?(8|16|32|64)_t)|'
r'(u?int(max|ptr)_t)|'
r')$')
# These headers are excluded from [build/include] and [build/include_order]
# checks:
# - Anything not following google file name conventions (containing an
# uppercase character, such as Python.h or nsStringAPI.h, for example).
# - Lua headers.
_THIRD_PARTY_HEADERS_PATTERN = re.compile(
r'^(?:[^/]*[A-Z][^/]*\.h|lua\.h|lauxlib\.h|lualib\.h)$')
# Pattern for matching FileInfo.BaseName() against test file name
_TEST_FILE_SUFFIX = r'(_test|_unittest|_regtest)$'
# Pattern that matches only complete whitespace, possibly across multiple lines.
_EMPTY_CONDITIONAL_BODY_PATTERN = re.compile(r'^\s*$', re.DOTALL)
# Assertion macros. These are defined in base/logging.h and
# testing/base/public/gunit.h.
_CHECK_MACROS = [
'DCHECK', 'CHECK',
'EXPECT_TRUE', 'ASSERT_TRUE',
'EXPECT_FALSE', 'ASSERT_FALSE',
]
# Replacement macros for CHECK/DCHECK/EXPECT_TRUE/EXPECT_FALSE
_CHECK_REPLACEMENT = dict([(m, {}) for m in _CHECK_MACROS])
for op, replacement in [('==', 'EQ'), ('!=', 'NE'),
('>=', 'GE'), ('>', 'GT'),
('', 'LE'),
('= verbose, category passes filter and is not 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.
"""
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))
elif _cpplint_state.output_format == 'eclipse':
sys.stderr.write('%s:%s: warning: %s [%s] [%d]\n' % (
filename, linenum, message, category, confidence))
elif _cpplint_state.output_format == 'tap':
template = ('not ok %(filename)s\n'
' ---\n'
' message: %(message)s\n'
' data:\n'
' line: %(linenum)d\n'
' ruleId: %(category)s\n'
' ...')
logger.info(template % locals())
else:
sys.stderr.write('%s:%s: %s [%s] [%d]\n' % (
filename, linenum, message, category, confidence))
# Matches standard C++ escape sequences per 2.13.2.3 of the C++ standard.
_RE_PATTERN_CLEANSE_LINE_ESCAPES = re.compile(
r'\\([abfnrtv?"\\\']|\d+|x[0-9a-fA-F]+)')
# Match a single C style comment on the same line.
_RE_PATTERN_C_COMMENTS = r'/\*(?:[^*]|\*(?!/))*\*/'
# Matches multi-line C style comments.
# This RE is a little bit more complicated than one might expect, because we
# have to take care of space removals tools so we can handle comments inside
# statements better.
# The current rule is: We only clear spaces from both sides when we're at the
# end of the line. Otherwise, we try to remove spaces from the right side,
# if this doesn't work we try on left side but only if there's a non-character
# on the right.
_RE_PATTERN_CLEANSE_LINE_C_COMMENTS = re.compile(
r'(\s*' + _RE_PATTERN_C_COMMENTS + r'\s*$|' +
_RE_PATTERN_C_COMMENTS + r'\s+|' +
r'\s+' + _RE_PATTERN_C_COMMENTS + r'(?=\W)|' +
_RE_PATTERN_C_COMMENTS + r')')
def IsCppString(line):
"""Does line terminate so, that the next symbol is in string constant.
This function does not consider single-line nor multi-line comments.
Args:
line: is a partial line of code starting from the 0..n.
Returns:
True, if next character appended to 'line' is inside a
string constant.
"""
line = line.replace(r'\\', 'XX') # after this, \\" does not match to \"
return ((line.count('"') - line.count(r'\"') - line.count("'\"'")) & 1) == 1
def CleanseRawStrings(raw_lines):
"""Removes C++11 raw strings from lines.
Before:
static const char kData[] = R"(
multi-line string
)";
After:
static const char kData[] = ""
(replaced by blank line)
"";
Args:
raw_lines: list of raw lines.
Returns:
list of lines with C++11 raw strings replaced by empty strings.
"""
delimiter = None
lines_without_raw_strings = []
for line in raw_lines:
if delimiter:
# Inside a raw string, look for the end
end = line.find(delimiter)
if end >= 0:
# Found the end of the string, match leading space for this
# line and resume copying the original lines, and also insert
# a "" on the last line.
leading_space = Match(r'^(\s*)\S', line)
line = leading_space.group(1) + '""' + line[end + len(delimiter):]
delimiter = None
else:
# Haven't found the end yet, append a blank line.
line = '""'
# Look for beginning of a raw string, and replace them with
# empty strings. This is done in a loop to handle multiple raw
# strings on the same line.
while delimiter is None:
# Look for beginning of a raw string.
# See 2.14.15 [lex.string] for syntax.
#
# Once we have matched a raw string, we check the prefix of the
# line to make sure that the line is not part of a single line
# comment. It's done this way because we remove raw strings
# before removing comments as opposed to removing comments
# before removing raw strings. This is because there are some
# cpplint checks that requires the comments to be preserved, but
# we don't want to check comments that are inside raw strings.
matched = Match(r'^(.*?)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line)
if (matched and
not Match(r'^([^\'"]|\'(\\.|[^\'])*\'|"(\\.|[^"])*")*//',
matched.group(1))):
delimiter = ')' + matched.group(2) + '"'
end = matched.group(3).find(delimiter)
if end >= 0:
# Raw string ended on same line
line = (matched.group(1) + '""' +
matched.group(3)[end + len(delimiter):])
delimiter = None
else:
# Start of a multi-line raw string
line = matched.group(1) + '""'
else:
break
lines_without_raw_strings.append(line)
# TODO(unknown): if delimiter is not None here, we might want to
# emit a warning for unterminated string.
return lines_without_raw_strings
def FindNextMultiLineCommentStart(lines, lineix):
"""Find the beginning marker for a multiline comment."""
while lineix < len(lines):
if lines[lineix].strip().startswith('/*'):
# Only return this marker if the comment goes beyond this line
if lines[lineix].strip().find('*/', 2) < 0:
return lineix
lineix += 1
return len(lines)
def FindNextMultiLineCommentEnd(lines, lineix):
"""We are inside a comment, find the end marker."""
while lineix < len(lines):
if lines[lineix].strip().endswith('*/'):
return lineix
lineix += 1
return len(lines)
def RemoveMultiLineCommentsFromRange(lines, begin, end):
"""Clears a range of lines for multi-line comments."""
# Having // dummy comments makes the lines non-empty, so we will not get
# unnecessary blank line warnings later in the code.
for i in range(begin, end):
lines[i] = '/**/'
def RemoveMultiLineComments(filename, lines, error):
"""Removes multiline (c-style) comments from lines."""
lineix = 0
while lineix < len(lines):
lineix_begin = FindNextMultiLineCommentStart(lines, lineix)
if lineix_begin >= len(lines):
return
lineix_end = FindNextMultiLineCommentEnd(lines, lineix_begin)
if lineix_end >= len(lines):
error(filename, lineix_begin + 1, 'readability/multiline_comment', 5,
'Could not find end of multi-line comment')
return
RemoveMultiLineCommentsFromRange(lines, lineix_begin, lineix_end + 1)
lineix = lineix_end + 1
def CleanseComments(line):
"""Removes //-comments and single-line C-style /* */ comments.
Args:
line: A line of C++ source.
Returns:
The line with single-line comments removed.
"""
commentpos = line.find('//')
if commentpos != -1 and not IsCppString(line[:commentpos]):
line = line[:commentpos].rstrip()
# get rid of /* ... */
return _RE_PATTERN_CLEANSE_LINE_C_COMMENTS.sub('', line)
class CleansedLines(object):
"""Holds 4 copies of all lines with different preprocessing applied to them.
1) elided member contains lines without strings and comments.
2) lines member contains lines without comments.
3) raw_lines member contains all the lines without processing.
4) lines_without_raw_strings member is same as raw_lines, but with C++11 raw
strings removed.
All these members are of , and of the same length.
"""
def __init__(self, lines):
self.elided = []
self.lines = []
self.raw_lines = lines
self.num_lines = len(lines)
self.lines_without_raw_strings = CleanseRawStrings(lines)
for linenum in range(len(self.lines_without_raw_strings)):
self.lines.append(CleanseComments(
self.lines_without_raw_strings[linenum]))
elided = self._CollapseStrings(self.lines_without_raw_strings[linenum])
self.elided.append(CleanseComments(elided))
def NumLines(self):
"""Returns the number of lines represented."""
return self.num_lines
@staticmethod
def _CollapseStrings(elided):
"""Collapses strings and chars on a line to simple "" or '' blocks.
We nix strings first so we're not fooled by text like '"http://"'
Args:
elided: The line being processed.
Returns:
The line with collapsed strings.
"""
if _RE_PATTERN_INCLUDE.match(elided):
return elided
# Remove escaped characters first to make quote/single quote collapsing
# basic. Things that look like escaped characters shouldn't occur
# outside of strings and chars.
elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub('', elided)
# Replace quoted strings and digit separators. Both single quotes
# and double quotes are processed in the same loop, otherwise
# nested quotes wouldn't work.
collapsed = ''
while True:
# Find the first quote character
match = Match(r'^([^\'"]*)([\'"])(.*)$', elided)
if not match:
collapsed += elided
break
head, quote, tail = match.groups()
if quote == '"':
# Collapse double quoted strings
second_quote = tail.find('"')
if second_quote >= 0:
collapsed += head + '""'
elided = tail[second_quote + 1:]
else:
# Unmatched double quote, don't bother processing the rest
# of the line since this is probably a multiline string.
collapsed += elided
break
else:
# Found single quote, check nearby text to eliminate digit separators.
#
# There is no special handling for floating point here, because
# the integer/fractional/exponent parts would all be parsed
# correctly as long as there are digits on both sides of the
# separator. So we are fine as long as we don't see something
# like "0.'3" (gcc 4.9.0 will not allow this literal).
if Search(r'\b(?:0[bBxX]?|[1-9])[0-9a-fA-F]*$', head):
match_literal = Match(r'^((?:\'?[0-9a-zA-Z_])*)(.*)$', "'" + tail)
collapsed += head + match_literal.group(1).replace("'", '')
elided = match_literal.group(2)
else:
second_quote = tail.find('\'')
if second_quote >= 0:
collapsed += head + "''"
elided = tail[second_quote + 1:]
else:
# Unmatched single quote
collapsed += elided
break
return collapsed
def FindEndOfExpressionInLine(line, startpos, stack):
"""Find the position just after the end of current parenthesized expression.
Args:
line: a CleansedLines line.
startpos: start searching at this position.
stack: nesting stack at startpos.
Returns:
On finding matching end: (index just after matching end, None)
On finding an unclosed expression: (-1, None)
Otherwise: (-1, new stack at end of this line)
"""
for i in xrange(startpos, len(line)):
char = line[i]
if char in '([{':
# Found start of parenthesized expression, push to expression stack
stack.append(char)
elif char == ' 0 and
(line[i - 1] == '-' or Search(r'\boperator\s*$', line[0:i - 1]))):
continue
# Pop the stack if there is a matching '' since it must be an operator.
if stack:
if stack[-1] == '', the matching ' 0 and
(line[i - 1] == '-' or
Match(r'\s>=\s', line[i - 1:]) or
Search(r'\boperator\s*$', line[0:i]))):
i -= 1
else:
stack.append('>')
elif char == '' on the stack, they must be
# operators. Remove those.
while stack and stack[-1] == '>':
stack.pop()
if not stack:
return (-1, None)
if ((char == '(' and stack[-1] == ')') or
(char == '[' and stack[-1] == ']') or
(char == '{' and stack[-1] == '}')):
stack.pop()
if not stack:
return (i, None)
else:
# Mismatched parentheses
return (-1, None)
elif char == ';':
# Found something that look like end of statements. If we are currently
# expecting a '' must have been an operator, since
# template argument list should not contain statements.
while stack and stack[-1] == '>':
stack.pop()
if not stack:
return (-1, None)
i -= 1
return (-1, stack)
def ReverseCloseExpression(clean_lines, linenum, pos):
"""If input points to ) or } or ] or >, finds the position that opens it.
If lines[linenum][pos] points to a ')' or '}' or ']' or '>', finds the
linenum/pos that correspond to the opening of the expression.
Args:
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
pos: A position on the line.
Returns:
A tuple (line, linenum, pos) pointer *at* the opening brace, or
(line, 0, -1) if we never find the matching opening brace. Note
we ignore strings and comments when matching; and the line we
return is the 'cleansed' line at linenum.
"""
line = clean_lines.elided[linenum]
if line[pos] not in ')}]>':
return (line, 0, -1)
# Check last line
(start_pos, stack) = FindStartOfExpressionInLine(line, pos, [])
if start_pos > -1:
return (line, linenum, start_pos)
# Continue scanning backward
while stack and linenum > 0:
linenum -= 1
line = clean_lines.elided[linenum]
(start_pos, stack) = FindStartOfExpressionInLine(line, len(line) - 1, stack)
if start_pos > -1:
return (line, linenum, start_pos)
# Did not find start of expression before beginning of file, give up
return (line, 0, -1)
def CheckForCopyright(filename, lines, error):
"""Logs an error if no Copyright message appears at the top of the file."""
# We'll say it should occur by line 10. Don't forget there's a
# dummy line at the front.
for line in xrange(1, min(len(lines), 11)):
if re.search(r'Copyright', lines[line], re.I): break
else: # means no copyright line was found
error(filename, 0, 'legal/copyright', 5,
'No copyright message found. '
'You should have a line: "Copyright [year] "')
def GetIndentLevel(line):
"""Return the number of leading spaces in line.
Args:
line: A string to check.
Returns:
An integer count of leading spaces, possibly zero.
"""
indent = Match(r'^( *)\S', line)
if indent:
return len(indent.group(1))
else:
return 0
def GetHeaderGuardCPPVariable(filename):
"""Returns the CPP variable that should be used as a header guard.
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$', '.h', filename)
filename = re.sub(r'/\.flymake/([^/]*)$', r'/\1', filename)
# Replace 'c++' with 'cpp'.
filename = filename.replace('C++', 'cpp').replace('c++', 'cpp')
fileinfo = FileInfo(filename)
file_path_from_root = fileinfo.RepositoryName()
if _root:
file_path_from_root = re.sub('^' + _root + os.sep, '', file_path_from_root)
return re.sub(r'[^a-zA-Z0-9]', '_', file_path_from_root).upper() + '_'
def CheckForHeaderGuard(filename, clean_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.
Args:
filename: The name of the C++ header file.
clean_lines: A CleansedLines instance containing the file.
error: The function to call with any errors found.
"""
# Don't check for header guards if there are error suppression
# comments somewhere in this file.
#
# Because this is silencing a warning for a nonexistent line, we
# only support the very specific NOLINT(build/header_guard) syntax,
# and not the general NOLINT or NOLINT(*) syntax.
raw_lines = clean_lines.lines_without_raw_strings
for i in raw_lines:
if Search(r'//\s*NOLINT\(build/header_guard\)', i):
return
cppvar = GetHeaderGuardCPPVariable(filename)
ifndef = ''
ifndef_linenum = 0
define = ''
endif = ''
endif_linenum = 0
for linenum, line in enumerate(raw_lines):
linesplit = line.split()
if len(linesplit) >= 2:
# find the first occurrence of #ifndef and #define, save arg
if not ifndef and linesplit[0] == '#ifndef':
# set ifndef to the header guard presented on the #ifndef line.
ifndef = linesplit[1]
ifndef_linenum = linenum
if not define and linesplit[0] == '#define':
define = linesplit[1]
# find the last occurrence of #endif, save entire line
if line.startswith('#endif'):
endif = line
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:
error_level = 0
if ifndef != cppvar + '_':
error_level = 5
ParseNolintSuppressions(filename, raw_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)
# Check for "//" comments on endif line.
ParseNolintSuppressions(filename, raw_lines[endif_linenum], endif_linenum,
error)
match = Match(r'#endif\s*//\s*' + cppvar + r'(_)?\b', endif)
if match:
if match.group(1) == '_':
# Issue low severity warning for deprecated double trailing underscore
error(filename, endif_linenum, 'build/header_guard', 0,
'#endif line should be "#endif // %s"' % cppvar)
return
# Didn't find the corresponding "//" comment. If this file does not
# contain any "//" comments at all, it could be that the compiler
# only wants "/**/" comments, look for those instead.
no_single_line_comments = True
for i in xrange(1, len(raw_lines) - 1):
line = raw_lines[i]
if Match(r'^(?:(?:\'(?:\.|[^\'])*\')|(?:"(?:\.|[^"])*")|[^\'"])*//', line):
no_single_line_comments = False
break
if no_single_line_comments:
match = Match(r'#endif\s*/\*\s*' + cppvar + r'(_)?\s*\*/', endif)
if match:
if match.group(1) == '_':
# Low severity warning for double trailing underscore
error(filename, endif_linenum, 'build/header_guard', 0,
'#endif line should be "#endif /* %s */"' % cppvar)
return
# Didn't find anything
error(filename, endif_linenum, 'build/header_guard', 5,
'#endif line should be "#endif // %s"' % cppvar)
def CheckHeaderFileIncluded(filename, include_state, error):
"""Logs an error if a .cc file does not include its header."""
# Do not check test files
fileinfo = FileInfo(filename)
if Search(_TEST_FILE_SUFFIX, fileinfo.BaseName()):
return
headerfile = filename[0:len(filename) - len(fileinfo.Extension())] + '.h'
if not os.path.exists(headerfile):
return
headername = FileInfo(headerfile).RepositoryName()
first_include = 0
for section_list in include_state.include_list:
for f in section_list:
if headername in f[0] or f[0] in headername:
return
if not first_include:
first_include = f[1]
error(filename, first_include, 'build/include', 5,
'%s should include its header file %s' % (fileinfo.RepositoryName(),
headername))
def CheckForBadCharacters(filename, lines, error):
"""Logs an error for each line containing bad characters.
Two kinds of bad characters:
1. 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 UTF-8 occurred adjacent to a newline.
2. NUL bytes. These are problematic for some tools.
Args:
filename: The name of the current file.
lines: An array of strings, each representing a line of the file.
error: The function to call with any errors found.
"""
for linenum, line in enumerate(lines):
if u'\ufffd' in line:
error(filename, linenum, 'readability/utf8', 5,
'Line contains invalid UTF-8 (or Unicode replacement character).')
if '\0' in line:
error(filename, linenum, 'readability/nul', 5, 'Line contains NUL byte.')
def CheckForNewlineAtEOF(filename, lines, error):
"""Logs an error if there is no newline char at the end of the file.
Args:
filename: The name of the current file.
lines: An array of strings, each representing a line of the file.
error: The function to call with any errors found.
"""
# The array lines() was created by adding two newlines to the
# original file (go figure), then splitting on \n.
# To verify that the file ends in \n, we just have to make sure the
# last-but-two element of lines() exists and is empty.
if len(lines) < 3 or lines[-2]:
error(filename, len(lines) - 2, 'whitespace/ending_newline', 5,
'Could not find a newline character at the end of the file.')
def CheckForMultilineCommentsAndStrings(filename, clean_lines, linenum, error):
"""Logs an error if we see /* ... */ or "..." that extend past one line.
/* ... */ comments are legit inside macros, for one line.
Otherwise, we prefer // comments, so it's ok to warn about the
other. Likewise, it's ok for strings to extend across multiple
lines, as long as a line continuation character (backslash)
terminates each line. Although not currently prohibited by the C++
style guide, it's ugly and unnecessary. We don't do well with either
in this lint program, so we warn about both.
Args:
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
line = clean_lines.elided[linenum]
# Remove all \\ (escaped backslashes) from the line. They are OK, and the
# second (escaped) slash may trigger later \" detection erroneously.
line = line.replace('\\\\', '')
if line.count('/*') > line.count('*/'):
error(filename, linenum, 'readability/multiline_comment', 5,
'Complex multi-line /*...*/-style comment found. '
'Lint may give bogus warnings. '
'Consider replacing these with //-style comments, '
'with #if 0...#endif, '
'or with more clearly structured multi-line comments.')
if (line.count('"') - line.count('\\"')) % 2:
error(filename, linenum, 'readability/multiline_string', 5,
'Multi-line string ("...") found. This lint script doesn\'t '
'do well with such strings, and may give bogus warnings. '
'Use C++11 raw strings or concatenation instead.')
# (non-threadsafe name, thread-safe alternative, validation pattern)
#
# The validation pattern is used to eliminate false positives such as:
# _rand(); // false positive due to substring match.
# ->rand(); // some member function rand().
# ACMRandom rand(seed); // some variable named rand.
# ISAACRandom rand(); // another variable named rand.
#
# Basically we require the return value of these functions to be used
# in some expression context on the same line by matching on some
# operator before the function name. This eliminates constructors and
# member function calls.
_UNSAFE_FUNC_PREFIX = r'(?:[-+*/=%^&|(\s+)'
_THREADING_LIST = (
('asctime(', 'asctime_r(', _UNSAFE_FUNC_PREFIX + r'asctime\([^)]+\)'),
('ctime(', 'ctime_r(', _UNSAFE_FUNC_PREFIX + r'ctime\([^)]+\)'),
('getgrgid(', 'getgrgid_r(', _UNSAFE_FUNC_PREFIX + r'getgrgid\([^)]+\)'),
('getgrnam(', 'getgrnam_r(', _UNSAFE_FUNC_PREFIX + r'getgrnam\([^)]+\)'),
('getlogin(', 'getlogin_r(', _UNSAFE_FUNC_PREFIX + r'getlogin\(\)'),
('getpwnam(', 'getpwnam_r(', _UNSAFE_FUNC_PREFIX + r'getpwnam\([^)]+\)'),
('getpwuid(', 'getpwuid_r(', _UNSAFE_FUNC_PREFIX + r'getpwuid\([^)]+\)'),
('gmtime(', 'gmtime_r(', _UNSAFE_FUNC_PREFIX + r'gmtime\([^)]+\)'),
('localtime(', 'localtime_r(', _UNSAFE_FUNC_PREFIX + r'localtime\([^)]+\)'),
('rand(', 'rand_r(', _UNSAFE_FUNC_PREFIX + r'rand\(\)'),
('strtok(', 'strtok_r(',
_UNSAFE_FUNC_PREFIX + r'strtok\([^)]+\)'),
('ttyname(', 'ttyname_r(', _UNSAFE_FUNC_PREFIX + r'ttyname\([^)]+\)'),
)
def CheckPosixThreading(filename, clean_lines, linenum, error):
"""Checks for calls to thread-unsafe functions.
Much code has been originally written without consideration of
multi-threading. Also, engineers are relying on their old experience;
they have learned posix before threading extensions were added. These
tests guide the engineers to use thread-safe functions (when using
posix directly).
Args:
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
line = clean_lines.elided[linenum]
for single_thread_func, multithread_safe_func, pattern in _THREADING_LIST:
# Additional pattern matching check to confirm that this is the
# function we are looking for
if Search(pattern, line):
error(filename, linenum, 'runtime/threadsafe_fn', 2,
'Consider using ' + multithread_safe_func +
'...) instead of ' + single_thread_func +
'...) for improved thread safety.')
def CheckVlogArguments(filename, clean_lines, linenum, error):
"""Checks that VLOG() is only used for defining a logging level.
For example, VLOG(2) is correct. VLOG(INFO), VLOG(WARNING), VLOG(ERROR), and
VLOG(FATAL) are not.
Args:
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
line = clean_lines.elided[linenum]
if Search(r'\bVLOG\((INFO|ERROR|WARNING|DFATAL|FATAL)\)', line):
error(filename, linenum, 'runtime/vlog', 5,
'VLOG() should be used with numeric verbosity level. '
'Use LOG() if you want symbolic severity levels.')
# Matches invalid increment: *count++, which moves pointer instead of
# incrementing a value.
_RE_PATTERN_INVALID_INCREMENT = re.compile(
r'^\s*\*\w+(\+\+|--);')
def CheckInvalidIncrement(filename, clean_lines, linenum, error):
"""Checks for invalid increment *count++.
For example following function:
void increment_counter(int* count) {
*count++;
}
is invalid, because it effectively does count++, moving pointer, and should
be replaced with ++*count, (*count)++ or *count += 1.
Args:
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
line = clean_lines.elided[linenum]
if _RE_PATTERN_INVALID_INCREMENT.match(line):
error(filename, linenum, 'runtime/invalid_increment', 5,
'Changing pointer instead of value (or unused value of operator*).')
def IsMacroDefinition(clean_lines, linenum):
if Search(r'^#define', clean_lines[linenum]):
return True
if linenum > 0 and Search(r'\\$', clean_lines[linenum - 1]):
return True
return False
def IsForwardClassDeclaration(clean_lines, linenum):
return Match(r'^\s*(\btemplate\b)*.*class\s+\w+;\s*$', clean_lines[linenum])
class _BlockInfo(object):
"""Stores information about a generic block of code."""
def __init__(self, linenum, seen_open_brace):
self.starting_linenum = linenum
self.seen_open_brace = seen_open_brace
self.open_parentheses = 0
self.inline_asm = _NO_ASM
self.check_namespace_indentation = False
def CheckBegin(self, filename, clean_lines, linenum, error):
"""Run checks that applies to text up to the opening brace.
This is mostly for checking the text after the class identifier
and the "{", usually where the base class is specified. For other
blocks, there isn't much to check, so we always pass.
Args:
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
pass
def CheckEnd(self, filename, clean_lines, linenum, error):
"""Run checks that applies to text after the closing brace.
This is mostly used for checking end of namespace comments.
Args:
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
pass
def IsBlockInfo(self):
"""Returns true if this block is a _BlockInfo.
This is convenient for verifying that an object is an instance of
a _BlockInfo, but not an instance of any of the derived classes.
Returns:
True for this class, False for derived classes.
"""
return self.__class__ == _BlockInfo
class _ExternCInfo(_BlockInfo):
"""Stores information about an 'extern "C"' block."""
def __init__(self, linenum):
_BlockInfo.__init__(self, linenum, True)
class _ClassInfo(_BlockInfo):
"""Stores information about a class."""
def __init__(self, name, class_or_struct, clean_lines, linenum):
_BlockInfo.__init__(self, linenum, False)
self.name = name
self.is_derived = False
self.check_namespace_indentation = True
if class_or_struct == 'struct':
self.access = 'public'
self.is_struct = True
else:
self.access = 'private'
self.is_struct = False
# Remember initial indentation level for this class. Using raw_lines here
# instead of elided to account for leading comments.
self.class_indent = GetIndentLevel(clean_lines.raw_lines[linenum])
# Try to find the end of the class. This will be confused by things like:
# class A {
# } *x = { ...
#
# But it's still good enough for CheckSectionSpacing.
self.last_line = 0
depth = 0
for i in range(linenum, clean_lines.NumLines()):
line = clean_lines.elided[i]
depth += line.count('{') - line.count('}')
if not depth:
self.last_line = i
break
def CheckBegin(self, filename, clean_lines, linenum, error):
# Look for a bare ':'
if Search('(^|[^:]):($|[^:])', clean_lines.elided[linenum]):
self.is_derived = True
def CheckEnd(self, filename, clean_lines, linenum, error):
# If there is a DISALLOW macro, it should appear near the end of
# the class.
seen_last_thing_in_class = False
for i in xrange(linenum - 1, self.starting_linenum, -1):
match = Search(
r'\b(DISALLOW_COPY_AND_ASSIGN|DISALLOW_IMPLICIT_CONSTRUCTORS)\(' +
self.name + r'\)',
clean_lines.elided[i])
if match:
if seen_last_thing_in_class:
error(filename, i, 'readability/constructors', 3,
match.group(1) + ' should be the last thing in the class')
break
if not Match(r'^\s*$', clean_lines.elided[i]):
seen_last_thing_in_class = True
# Check that closing brace is aligned with beginning of the class.
# Only do this if the closing brace is indented by only whitespaces.
# This means we will not check single-line class definitions.
indent = Match(r'^( *)\}', clean_lines.elided[linenum])
if indent and len(indent.group(1)) != self.class_indent:
if self.is_struct:
parent = 'struct ' + self.name
else:
parent = 'class ' + self.name
error(filename, linenum, 'whitespace/indent', 3,
'Closing brace should be aligned with beginning of %s' % parent)
class _NamespaceInfo(_BlockInfo):
"""Stores information about a namespace."""
def __init__(self, name, linenum):
_BlockInfo.__init__(self, linenum, False)
self.name = name or ''
self.check_namespace_indentation = True
def CheckEnd(self, filename, clean_lines, linenum, error):
"""Check end of namespace comments."""
line = clean_lines.raw_lines[linenum]
# Check how many lines is enclosed in this namespace. Don't issue
# warning for missing namespace comments if there aren't enough
# lines. However, do apply checks if there is already an end of
# namespace comment and it's incorrect.
#
# TODO(unknown): We always want to check end of namespace comments
# if a namespace is large, but sometimes we also want to apply the
# check if a short namespace contained nontrivial things (something
# other than forward declarations). There is currently no logic on
# deciding what these nontrivial things are, so this check is
# triggered by namespace size only, which works most of the time.
if (linenum - self.starting_linenum < 10
and not Match(r'^\s*};*\s*(//|/\*).*\bnamespace\b', line)):
return
# Look for matching comment at end of namespace.
#
# Note that we accept C style "/* */" comments for terminating
# namespaces, so that code that terminate namespaces inside
# preprocessor macros can be cpplint clean.
#
# We also accept stuff like "// end of namespace ." with the
# period at the end.
#
# Besides these, we don't accept anything else, otherwise we might
# get false negatives when existing comment is a substring of the
# expected namespace.
if self.name:
# Named namespace
if not Match((r'^\s*};*\s*(//|/\*).*\bnamespace\s+' +
re.escape(self.name) + r'[\*/\.\\\s]*$'),
line):
error(filename, linenum, 'readability/namespace', 5,
'Namespace should be terminated with "// namespace %s"' %
self.name)
else:
# Anonymous namespace
if not Match(r'^\s*};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$', line):
# If "// namespace anonymous" or "// anonymous namespace (more text)",
# mention "// anonymous namespace" as an acceptable form
if Match(r'^\s*}.*\b(namespace anonymous|anonymous namespace)\b', line):
error(filename, linenum, 'readability/namespace', 5,
'Anonymous namespace should be terminated with "// namespace"'
' or "// anonymous namespace"')
else:
error(filename, linenum, 'readability/namespace', 5,
'Anonymous namespace should be terminated with "// namespace"')
class _PreprocessorInfo(object):
"""Stores checkpoints of nesting stacks when #if/#else is seen."""
def __init__(self, stack_before_if):
# The entire nesting stack before #if
self.stack_before_if = stack_before_if
# The entire nesting stack up to #else
self.stack_before_else = []
# Whether we have already seen #else or #elif
self.seen_else = False
class NestingState(object):
"""Holds states related to parsing braces."""
def __init__(self):
# Stack for tracking all braces. An object is pushed whenever we
# see a "{", and popped when we see a "}". Only 3 types of
# objects are possible:
# - _ClassInfo: a class or struct.
# - _NamespaceInfo: a namespace.
# - _BlockInfo: some other type of block.
self.stack = []
# Top of the previous stack before each Update().
#
# Because the nesting_stack is updated at the end of each line, we
# had to do some convoluted checks to find out what is the current
# scope at the beginning of the line. This check is simplified by
# saving the previous top of nesting stack.
#
# We could save the full stack, but we only need the top. Copying
# the full nesting stack would slow down cpplint by ~10%.
self.previous_stack_top = []
# Stack of _PreprocessorInfo objects.
self.pp_stack = []
def SeenOpenBrace(self):
"""Check if we have seen the opening brace for the innermost block.
Returns:
True if we have seen the opening brace, False if the innermost
block is still expecting an opening brace.
"""
return (not self.stack) or self.stack[-1].seen_open_brace
def InNamespaceBody(self):
"""Check if we are currently one level inside a namespace body.
Returns:
True if top of the stack is a namespace block, False otherwise.
"""
return self.stack and isinstance(self.stack[-1], _NamespaceInfo)
def InExternC(self):
"""Check if we are currently one level inside an 'extern "C"' block.
Returns:
True if top of the stack is an extern block, False otherwise.
"""
return self.stack and isinstance(self.stack[-1], _ExternCInfo)
def InClassDeclaration(self):
"""Check if we are currently one level inside a class or struct declaration.
Returns:
True if top of the stack is a class/struct, False otherwise.
"""
return self.stack and isinstance(self.stack[-1], _ClassInfo)
def InAsmBlock(self):
"""Check if we are currently one level inside an inline ASM block.
Returns:
True if the top of the stack is a block containing inline ASM.
"""
return self.stack and self.stack[-1].inline_asm != _NO_ASM
def InTemplateArgumentList(self, clean_lines, linenum, pos):
"""Check if current position is inside template argument list.
Args:
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
pos: position just after the suspected template argument.
Returns:
True if (linenum, pos) is inside template arguments.
"""
while linenum < clean_lines.NumLines():
# Find the earliest character that might indicate a template argument
line = clean_lines.elided[linenum]
match = Match(r'^[^{};=\[\]\.]*(.)', line[pos:])
if not match:
linenum += 1
pos = 0
continue
token = match.group(1)
pos += len(match.group(0))
# These things do not look like template argument list:
# class Suspect {
# class Suspect x; }
if token in ('{', '}', ';'): return False
# These things look like template argument list:
# template
# template
# template
# template
if token in ('>', '=', '[', ']', '.'): return True
# Check if token is an unmatched '? and