config/check_spidermonkey_style.py

   1 # vim: set ts=8 sts=4 et sw=4 tw=99:
   2 # This Source Code Form is subject to the terms of the Mozilla Public
   3 # License, v. 2.0. If a copy of the MPL was not distributed with this
   4 # file, You can obtain one at http://mozilla.org/MPL/2.0/.
   5
   6 # ----------------------------------------------------------------------------
   7 # This script checks various aspects of SpiderMonkey code style.  The current checks are as
   8 # follows.
   9 #
  10 # We check the following things in headers.
  11 #
  12 # - No cyclic dependencies.
  13 #
  14 # - No normal header should #include a inlines.h/-inl.h file.
  15 #
  16 # - #ifndef wrappers should have the right form. (XXX: not yet implemented)
  17 #   - Every header file should have one.
  18 #   - The guard name used should be appropriate for the filename.
  19 #
  20 # We check the following things in all files.
  21 #
  22 # - #includes should have full paths, e.g. "jit/Ion.h", not "Ion.h".
  23 #
  24 # - #includes should use the appropriate form for system headers (<...>) and
  25 #   local headers ("...").
  26 #
  27 # - #includes should be ordered correctly.
  28 #   - Each one should be in the correct section.
  29 #   - Alphabetical order should be used within sections.
  30 #   - Sections should be in the right order.
  31 #   Note that the presence of #if/#endif blocks complicates things, to the
  32 #   point that it's not always clear where a conditionally-compiled #include
  33 #   statement should go, even to a human.  Therefore, we check the #include
  34 #   statements within each #if/#endif block (including nested ones) in
  35 #   isolation, but don't try to do any order checking between such blocks.
  36 # ----------------------------------------------------------------------------
  37
  38 from __future__ import absolute_import, print_function
  39
  40 import difflib
  41 import os
  42 import re
  43 import sys
  44
  45 # We don't bother checking files in these directories, because they're (a) auxiliary or (b)
  46 # imported code that doesn't follow our coding style.
  47 ignored_js_src_dirs = [
  48     'js/src/config/',            # auxiliary stuff
  49     'js/src/ctypes/libffi/',     # imported code
  50     'js/src/devtools/',          # auxiliary stuff
  51     'js/src/editline/',          # imported code
  52     'js/src/gdb/',               # auxiliary stuff
  53     'js/src/vtune/',             # imported code
  54     'js/src/zydis/',             # imported code
  55 ]
  56
  57 # We ignore #includes of these files, because they don't follow the usual rules.
  58 included_inclnames_to_ignore = set([
  59     'ffi.h',                    # generated in ctypes/libffi/
  60     'devtools/Instruments.h',   # we ignore devtools/ in general
  61     'double-conversion/double-conversion.h',  # strange MFBT case
  62     'javascript-trace.h',       # generated in $OBJDIR if HAVE_DTRACE is defined
  63     'frontend/ReservedWordsGenerated.h',  # generated in $OBJDIR
  64     'gc/StatsPhasesGenerated.h',         # generated in $OBJDIR
  65     'gc/StatsPhasesGenerated.inc',       # generated in $OBJDIR
  66     'jit/LOpcodes.h',           # generated in $OBJDIR
  67     'jit/MOpcodes.h',           # generated in $OBJDIR
  68     'jscustomallocator.h',      # provided by embedders;  allowed to be missing
  69     'js-config.h',              # generated in $OBJDIR
  70     'fdlibm.h',                 # fdlibm
  71     'FuzzerDefs.h',             # included without a path
  72     'FuzzingInterface.h',       # included without a path
  73     'mozmemory.h',              # included without a path
  74     'pratom.h',                 # NSPR
  75     'prcvar.h',                 # NSPR
  76     'prerror.h',                # NSPR
  77     'prinit.h',                 # NSPR
  78     'prio.h',                   # NSPR
  79     'private/pprio.h',          # NSPR
  80     'prlink.h',                 # NSPR
  81     'prlock.h',                 # NSPR
  82     'prprf.h',                  # NSPR
  83     'prthread.h',               # NSPR
  84     'prtypes.h',                # NSPR
  85     'selfhosted.out.h',         # generated in $OBJDIR
  86     'shellmoduleloader.out.h',  # generated in $OBJDIR
  87     'unicode/basictz.h',        # ICU
  88     'unicode/locid.h',          # ICU
  89     'unicode/plurrule.h',       # ICU
  90     'unicode/putil.h',          # ICU
  91     'unicode/timezone.h',       # ICU
  92     'unicode/ucal.h',           # ICU
  93     'unicode/uchar.h',          # ICU
  94     'unicode/uclean.h',         # ICU
  95     'unicode/ucol.h',           # ICU
  96     'unicode/udat.h',           # ICU
  97     'unicode/udata.h',          # ICU
  98     'unicode/udatpg.h',         # ICU
  99     'unicode/udisplaycontext.h',  # ICU
 100     'unicode/uenum.h',          # ICU
 101     'unicode/uformattedvalue.h',  # ICU
 102     'unicode/uloc.h',           # ICU
 103     'unicode/unistr.h',         # ICU
 104     'unicode/unorm2.h',         # ICU
 105     'unicode/unum.h',           # ICU
 106     'unicode/unumberformatter.h',  # ICU
 107     'unicode/unumsys.h',        # ICU
 108     'unicode/upluralrules.h',   # ICU
 109     'unicode/ureldatefmt.h',    # ICU
 110     'unicode/ures.h',           # ICU
 111     'unicode/ustring.h',        # ICU
 112     'unicode/utypes.h',         # ICU
 113     'unicode/uversion.h',       # ICU
 114     'vtune/VTuneWrapper.h',     # VTune
 115     'zydis/ZydisAPI.h',         # Zydis
 116 ])
 117
 118 # These files have additional constraints on where they are #included, so we
 119 # ignore #includes of them when checking #include ordering.
 120 oddly_ordered_inclnames = set([
 121     'ctypes/typedefs.h',        # Included multiple times in the body of ctypes/CTypes.h
 122     # Included in the body of frontend/TokenStream.h
 123     'frontend/ReservedWordsGenerated.h',
 124     'gc/StatsPhasesGenerated.h',         # Included in the body of gc/Statistics.h
 125     'gc/StatsPhasesGenerated.inc',       # Included in the body of gc/Statistics.cpp
 126     'psapi.h',                  # Must be included after "util/Windows.h" on Windows
 127     'machine/endian.h',         # Must be included after <sys/types.h> on BSD
 128     'winbase.h',                # Must precede other system headers(?)
 129     'windef.h'                  # Must precede other system headers(?)
 130 ])
 131
 132 # The files in tests/style/ contain code that fails this checking in various
 133 # ways.  Here is the output we expect.  If the actual output differs from
 134 # this, one of the following must have happened.
 135 # - New SpiderMonkey code violates one of the checked rules.
 136 # - The tests/style/ files have changed without expected_output being changed
 137 #   accordingly.
 138 # - This script has been broken somehow.
 139 #
 140 expected_output = '''\
 141 js/src/tests/style/BadIncludes.h:3: error:
 142     the file includes itself
 143
 144 js/src/tests/style/BadIncludes.h:6: error:
 145     "BadIncludes2.h" is included using the wrong path;
 146     did you forget a prefix, or is the file not yet committed?
 147
 148 js/src/tests/style/BadIncludes.h:8: error:
 149     <tests/style/BadIncludes2.h> should be included using
 150     the #include "..." form
 151
 152 js/src/tests/style/BadIncludes.h:10: error:
 153     "stdio.h" is included using the wrong path;
 154     did you forget a prefix, or is the file not yet committed?
 155
 156 js/src/tests/style/BadIncludes2.h:1: error:
 157     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 158
 159 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 160     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 161
 162 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 163     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 164
 165 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 166     "js/Value.h" should be included after "ds/LifoAlloc.h"
 167
 168 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 169     "ds/LifoAlloc.h" should be included after "jsapi.h"
 170
 171 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 172     "jsapi.h" should be included after <stdio.h>
 173
 174 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 175     <stdio.h> should be included after "mozilla/HashFunctions.h"
 176
 177 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 178     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 179
 180 (multiple files): error:
 181     header files form one or more cycles
 182
 183    tests/style/HeaderCycleA1.h
 184    -> tests/style/HeaderCycleA2.h
 185       -> tests/style/HeaderCycleA3.h
 186          -> tests/style/HeaderCycleA1.h
 187
 188    tests/style/HeaderCycleB1-inl.h
 189    -> tests/style/HeaderCycleB2-inl.h
 190       -> tests/style/HeaderCycleB3-inl.h
 191          -> tests/style/HeaderCycleB4-inl.h
 192             -> tests/style/HeaderCycleB1-inl.h
 193             -> tests/style/jsheadercycleB5inlines.h
 194                -> tests/style/HeaderCycleB1-inl.h
 195       -> tests/style/HeaderCycleB4-inl.h
 196
 197 '''.splitlines(True)
 198
 199 actual_output = []
 200
 201
 202 def out(*lines):
 203     for line in lines:
 204         actual_output.append(line + '\n')
 205
 206
 207 def error(filename, linenum, *lines):
 208     location = filename
 209     if linenum is not None:
 210         location += ':' + str(linenum)
 211     out(location + ': error:')
 212     for line in (lines):
 213         out('    ' + line)
 214     out('')
 215
 216
 217 class FileKind(object):
 218     C = 1
 219     CPP = 2
 220     INL_H = 3
 221     H = 4
 222     TBL = 5
 223     MSG = 6
 224
 225     @staticmethod
 226     def get(filename):
 227         if filename.endswith('.c'):
 228             return FileKind.C
 229
 230         if filename.endswith('.cpp'):
 231             return FileKind.CPP
 232
 233         if filename.endswith(('inlines.h', '-inl.h')):
 234             return FileKind.INL_H
 235
 236         if filename.endswith('.h'):
 237             return FileKind.H
 238
 239         if filename.endswith('.tbl'):
 240             return FileKind.TBL
 241
 242         if filename.endswith('.msg'):
 243             return FileKind.MSG
 244
 245         error(filename, None, 'unknown file kind')
 246
 247
 248 def check_style(enable_fixup):
 249     # We deal with two kinds of name.
 250     # - A "filename" is a full path to a file from the repository root.
 251     # - An "inclname" is how a file is referred to in a #include statement.
 252     #
 253     # Examples (filename -> inclname)
 254     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 255     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 256     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 257     # - "js/public/Vector.h"        -> "js/Vector.h"
 258     # - "js/src/vm/String.h"        -> "vm/String.h"
 259
 260     non_js_dirnames = ('mfbt/',
 261                        'memory/mozalloc/',
 262                        'mozglue/')  # type: tuple(str)
 263     non_js_inclnames = set()        # type: set(inclname)
 264     js_names = dict()               # type: dict(filename, inclname)
 265
 266     # Process files in js/src.
 267     js_src_root = os.path.join('js', 'src')
 268     for dirpath, dirnames, filenames in os.walk(js_src_root):
 269         if dirpath == js_src_root:
 270             # Skip any subdirectories that contain a config.status file
 271             # (likely objdirs).
 272             builddirs = []
 273             for dirname in dirnames:
 274                 path = os.path.join(dirpath, dirname, 'config.status')
 275                 if os.path.isfile(path):
 276                     builddirs.append(dirname)
 277             for dirname in builddirs:
 278                 dirnames.remove(dirname)
 279         for filename in filenames:
 280             filepath = os.path.join(dirpath, filename).replace('\\', '/')
 281             if not filepath.startswith(tuple(ignored_js_src_dirs)) and \
 282                filepath.endswith(('.c', '.cpp', '.h', '.tbl', '.msg')):
 283                 inclname = filepath[len('js/src/'):]
 284                 js_names[filepath] = inclname
 285
 286     # Look for header files in directories in non_js_dirnames.
 287     for non_js_dir in non_js_dirnames:
 288         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 289             for filename in filenames:
 290                 if filename.endswith('.h'):
 291                     inclname = 'mozilla/' + filename
 292                     non_js_inclnames.add(inclname)
 293
 294     # Look for header files in js/public.
 295     js_public_root = os.path.join('js', 'public')
 296     for dirpath, dirnames, filenames in os.walk(js_public_root):
 297         for filename in filenames:
 298             if filename.endswith('.h'):
 299                 filepath = os.path.join(dirpath, filename).replace('\\', '/')
 300                 inclname = 'js/' + filepath[len('js/public/'):]
 301                 js_names[filepath] = inclname
 302
 303     all_inclnames = non_js_inclnames | set(js_names.values())
 304
 305     edges = dict()      # type: dict(inclname, set(inclname))
 306
 307     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 308     # are #included from JS files we have to add them to the inclusion graph.
 309     for inclname in non_js_inclnames:
 310         edges[inclname] = set()
 311
 312     # Process all the JS files.
 313     for filename in sorted(js_names.keys()):
 314         inclname = js_names[filename]
 315         file_kind = FileKind.get(filename)
 316         if file_kind == FileKind.C or file_kind == FileKind.CPP or \
 317            file_kind == FileKind.H or file_kind == FileKind.INL_H:
 318             included_h_inclnames = set()    # type: set(inclname)
 319
 320             with open(filename) as f:
 321                 code = read_file(f)
 322
 323             if enable_fixup:
 324                 code = code.sorted(inclname)
 325                 with open(filename, 'w') as f:
 326                     f.write(code.to_source())
 327
 328             check_file(filename, inclname, file_kind, code,
 329                        all_inclnames, included_h_inclnames)
 330
 331         edges[inclname] = included_h_inclnames
 332
 333     find_cycles(all_inclnames, edges)
 334
 335     # Compare expected and actual output.
 336     difflines = difflib.unified_diff(expected_output, actual_output,
 337                                      fromfile='check_spidermonkey_style.py expected output',
 338                                      tofile='check_spidermonkey_style.py actual output')
 339     ok = True
 340     for diffline in difflines:
 341         ok = False
 342         print(diffline, end='')
 343
 344     return ok
 345
 346
 347 def module_name(name):
 348     '''Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename.'''
 349
 350     return name.replace('inlines.h', '').replace('-inl.h', '').replace('.h', '').replace('.cpp', '')  # NOQA: E501
 351
 352
 353 def is_module_header(enclosing_inclname, header_inclname):
 354     '''Determine if an included name is the "module header", i.e. should be
 355     first in the file.'''
 356
 357     module = module_name(enclosing_inclname)
 358
 359     # Normal case, e.g. module == "foo/Bar", header_inclname == "foo/Bar.h".
 360     if module == module_name(header_inclname):
 361         return True
 362
 363     # A public header, e.g. module == "foo/Bar", header_inclname == "js/Bar.h".
 364     m = re.match(r'js\/(.*)\.h', header_inclname)
 365     if m is not None and module.endswith('/' + m.group(1)):
 366         return True
 367
 368     return False
 369
 370
 371 class Include(object):
 372     '''Important information for a single #include statement.'''
 373
 374     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 375         self.include_prefix = include_prefix
 376         self.line_suffix = line_suffix
 377         self.inclname = inclname
 378         self.linenum = linenum
 379         self.is_system = is_system
 380
 381     def is_style_relevant(self):
 382         # Includes are style-relevant; that is, they're checked by the pairwise
 383         # style-checking algorithm in check_file.
 384         return True
 385
 386     def section(self, enclosing_inclname):
 387         '''Identify which section inclname belongs to.
 388
 389         The section numbers are as follows.
 390           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 391           1. mozilla/Foo.h
 392           2. <foo.h> or <foo>
 393           3. jsfoo.h, prmjtime.h, etc
 394           4. foo/Bar.h
 395           5. jsfooinlines.h
 396           6. foo/Bar-inl.h
 397           7. non-.h, e.g. *.tbl, *.msg
 398         '''
 399
 400         if self.is_system:
 401             return 2
 402
 403         if not self.inclname.endswith('.h'):
 404             return 7
 405
 406         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 407         # special handling.
 408         if is_module_header(enclosing_inclname, self.inclname):
 409             return 0
 410
 411         if '/' in self.inclname:
 412             if self.inclname.startswith('mozilla/'):
 413                 return 1
 414
 415             if self.inclname.endswith('-inl.h'):
 416                 return 6
 417
 418             return 4
 419
 420         if self.inclname.endswith('inlines.h'):
 421             return 5
 422
 423         return 3
 424
 425     def quote(self):
 426         if self.is_system:
 427             return '<' + self.inclname + '>'
 428         else:
 429             return '"' + self.inclname + '"'
 430
 431     def sort_key(self, enclosing_inclname):
 432         return (self.section(enclosing_inclname), self.inclname.lower())
 433
 434     def to_source(self):
 435         return self.include_prefix + self.quote() + self.line_suffix + '\n'
 436
 437
 438 class CppBlock(object):
 439     '''C preprocessor block: a whole file or a single #if/#elif/#else block.
 440
 441     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 442     The top-level block, which is not within a #if/#endif pair, is also
 443     considered a block.
 444
 445     Each kid is either an Include (representing a #include), OrdinaryCode, or
 446     a nested CppBlock.'''
 447
 448     def __init__(self, start_line=""):
 449         self.start = start_line
 450         self.end = ''
 451         self.kids = []
 452
 453     def is_style_relevant(self):
 454         return True
 455
 456     def append_ordinary_line(self, line):
 457         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 458             self.kids.append(OrdinaryCode())
 459         self.kids[-1].lines.append(line)
 460
 461     def style_relevant_kids(self):
 462         """ Return a list of kids in this block that are style-relevant. """
 463         return [kid for kid in self.kids if kid.is_style_relevant()]
 464
 465     def sorted(self, enclosing_inclname):
 466         """Return a hopefully-sorted copy of this block. Implements --fixup.
 467
 468         When in doubt, this leaves the code unchanged.
 469         """
 470
 471         def pretty_sorted_includes(includes):
 472             """ Return a new list containing the given includes, in order,
 473             with blank lines separating sections. """
 474             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 475             if sorted(keys) == keys:
 476                 return includes  # if nothing is out of order, don't touch anything
 477
 478             output = []
 479             current_section = None
 480             for (section, _), inc in sorted(zip(keys, includes)):
 481                 if current_section is not None and section != current_section:
 482                     output.append(OrdinaryCode(["\n"]))  # blank line
 483                 output.append(inc)
 484                 current_section = section
 485             return output
 486
 487         def should_try_to_sort(includes):
 488             if 'tests/style/BadIncludes' in enclosing_inclname:
 489                 return False  # don't straighten the counterexample
 490             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 491                 return False  # don't sort batches containing odd includes
 492             if includes == sorted(includes, key=lambda inc: inc.sort_key(enclosing_inclname)):
 493                 return False  # it's already sorted, avoid whitespace-only fixups
 494             return True
 495
 496         # The content of the eventual output of this method.
 497         output = []
 498
 499         # The current batch of includes to sort. This list only ever contains Include objects
 500         # and whitespace OrdinaryCode objects.
 501         batch = []
 502
 503         def flush_batch():
 504             """Sort the contents of `batch` and move it to `output`."""
 505
 506             assert all(isinstance(item, Include)
 507                        or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 508                        for item in batch)
 509
 510             # Here we throw away the blank lines.
 511             # `pretty_sorted_includes` puts them back.
 512             includes = []
 513             last_include_index = -1
 514             for i, item in enumerate(batch):
 515                 if isinstance(item, Include):
 516                     includes.append(item)
 517                     last_include_index = i
 518             cutoff = last_include_index + 1
 519
 520             if should_try_to_sort(includes):
 521                 output.extend(pretty_sorted_includes(
 522                     includes) + batch[cutoff:])
 523             else:
 524                 output.extend(batch)
 525             del batch[:]
 526
 527         for kid in self.kids:
 528             if isinstance(kid, CppBlock):
 529                 flush_batch()
 530                 output.append(kid.sorted(enclosing_inclname))
 531             elif isinstance(kid, Include):
 532                 batch.append(kid)
 533             else:
 534                 assert isinstance(kid, OrdinaryCode)
 535                 if kid.to_source().isspace():
 536                     batch.append(kid)
 537                 else:
 538                     flush_batch()
 539                     output.append(kid)
 540         flush_batch()
 541
 542         result = CppBlock()
 543         result.start = self.start
 544         result.end = self.end
 545         result.kids = output
 546         return result
 547
 548     def to_source(self):
 549         return self.start + ''.join(kid.to_source() for kid in self.kids) + self.end
 550
 551
 552 class OrdinaryCode(object):
 553     ''' A list of lines of code that aren't #include/#if/#else/#endif lines. '''
 554
 555     def __init__(self, lines=None):
 556         self.lines = lines if lines is not None else []
 557
 558     def is_style_relevant(self):
 559         return False
 560
 561     def to_source(self):
 562         return ''.join(self.lines)
 563
 564
 565 # A "snippet" is one of:
 566 #
 567 # *   Include - representing an #include line
 568 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 569 # *   OrdinaryCode - representing lines of non-#include-relevant code
 570
 571 def read_file(f):
 572     block_stack = [CppBlock()]
 573
 574     # Extract the #include statements as a tree of snippets.
 575     for linenum, line in enumerate(f, start=1):
 576         if line.lstrip().startswith('#'):
 577             # Look for a |#include "..."| line.
 578             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 579             if m is not None:
 580                 prefix, inclname, suffix = m.groups()
 581                 block_stack[-1].kids.append(Include(prefix,
 582                                                     inclname, suffix, linenum, is_system=False))
 583                 continue
 584
 585             # Look for a |#include <...>| line.
 586             m = re.match(r'(\s*#\s*include\s+)<([^>]*)>(.*)', line)
 587             if m is not None:
 588                 prefix, inclname, suffix = m.groups()
 589                 block_stack[-1].kids.append(Include(prefix,
 590                                                     inclname, suffix, linenum, is_system=True))
 591                 continue
 592
 593             # Look for a |#{if,ifdef,ifndef}| line.
 594             m = re.match(r'\s*#\s*(if|ifdef|ifndef)\b', line)
 595             if m is not None:
 596                 # Open a new block.
 597                 new_block = CppBlock(line)
 598                 block_stack[-1].kids.append(new_block)
 599                 block_stack.append(new_block)
 600                 continue
 601
 602             # Look for a |#{elif,else}| line.
 603             m = re.match(r'\s*#\s*(elif|else)\b', line)
 604             if m is not None:
 605                 # Close the current block, and open an adjacent one.
 606                 block_stack.pop()
 607                 new_block = CppBlock(line)
 608                 block_stack[-1].kids.append(new_block)
 609                 block_stack.append(new_block)
 610                 continue
 611
 612             # Look for a |#endif| line.
 613             m = re.match(r'\s*#\s*endif\b', line)
 614             if m is not None:
 615                 # Close the current block.
 616                 block_stack.pop().end = line
 617                 if len(block_stack) == 0:
 618                     raise ValueError(
 619                         "#endif without #if at line " + str(linenum))
 620                 continue
 621
 622         # Otherwise, we have an ordinary line.
 623         block_stack[-1].append_ordinary_line(line)
 624
 625     if len(block_stack) > 1:
 626         raise ValueError("unmatched #if")
 627     return block_stack[-1]
 628
 629
 630 def check_file(filename, inclname, file_kind, code, all_inclnames, included_h_inclnames):
 631
 632     def check_include_statement(include):
 633         '''Check the style of a single #include statement.'''
 634
 635         if include.is_system:
 636             # Check it is not a known local file (in which case it's probably a system header).
 637             if include.inclname in included_inclnames_to_ignore or \
 638                include.inclname in all_inclnames:
 639                 error(filename, include.linenum,
 640                       include.quote() + ' should be included using',
 641                       'the #include "..." form')
 642
 643         else:
 644             if include.inclname not in included_inclnames_to_ignore:
 645                 included_kind = FileKind.get(include.inclname)
 646
 647                 # Check the #include path has the correct form.
 648                 if include.inclname not in all_inclnames:
 649                     error(filename, include.linenum,
 650                           include.quote() + ' is included using the wrong path;',
 651                           'did you forget a prefix, or is the file not yet committed?')
 652
 653                 # Record inclusions of .h files for cycle detection later.
 654                 # (Exclude .tbl and .msg files.)
 655                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 656                     included_h_inclnames.add(include.inclname)
 657
 658                 # Check a H file doesn't #include an INL_H file.
 659                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 660                     error(filename, include.linenum,
 661                           'vanilla header includes an inline-header file ' + include.quote())
 662
 663                 # Check a file doesn't #include itself.  (We do this here because the cycle
 664                 # detection below doesn't detect this case.)
 665                 if inclname == include.inclname:
 666                     error(filename, include.linenum,
 667                           'the file includes itself')
 668
 669     def check_includes_order(include1, include2):
 670         '''Check the ordering of two #include statements.'''
 671
 672         if include1.inclname in oddly_ordered_inclnames or \
 673            include2.inclname in oddly_ordered_inclnames:
 674             return
 675
 676         section1 = include1.section(inclname)
 677         section2 = include2.section(inclname)
 678         if (section1 > section2) or \
 679            ((section1 == section2) and (include1.inclname.lower() > include2.inclname.lower())):
 680             error(filename, str(include1.linenum) + ':' + str(include2.linenum),
 681                   include1.quote() + ' should be included after ' + include2.quote())
 682
 683     # Check the extracted #include statements, both individually, and the ordering of
 684     # adjacent pairs that live in the same block.
 685     def pair_traverse(prev, this):
 686         if isinstance(this, Include):
 687             check_include_statement(this)
 688             if isinstance(prev, Include):
 689                 check_includes_order(prev, this)
 690         else:
 691             kids = this.style_relevant_kids()
 692             for prev2, this2 in zip([None] + kids[0:-1], kids):
 693                 pair_traverse(prev2, this2)
 694
 695     pair_traverse(None, code)
 696
 697
 698 def find_cycles(all_inclnames, edges):
 699     '''Find and draw any cycles.'''
 700
 701     SCCs = tarjan(all_inclnames, edges)
 702
 703     # The various sorted() calls below ensure the output is deterministic.
 704
 705     def draw_SCC(c):
 706         cset = set(c)
 707         drawn = set()
 708
 709         def draw(v, indent):
 710             out('   ' * indent + ('-> ' if indent else '   ') + v)
 711             if v in drawn:
 712                 return
 713             drawn.add(v)
 714             for succ in sorted(edges[v]):
 715                 if succ in cset:
 716                     draw(succ, indent + 1)
 717         draw(sorted(c)[0], 0)
 718         out('')
 719
 720     have_drawn_an_SCC = False
 721     for scc in sorted(SCCs):
 722         if len(scc) != 1:
 723             if not have_drawn_an_SCC:
 724                 error('(multiple files)', None,
 725                       'header files form one or more cycles')
 726                 have_drawn_an_SCC = True
 727
 728             draw_SCC(scc)
 729
 730
 731 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 732 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 733 def tarjan(V, E):
 734     vertex_index = {}
 735     vertex_lowlink = {}
 736     index = 0
 737     S = []
 738     all_SCCs = []
 739
 740     def strongconnect(v, index):
 741         # Set the depth index for v to the smallest unused index
 742         vertex_index[v] = index
 743         vertex_lowlink[v] = index
 744         index += 1
 745         S.append(v)
 746
 747         # Consider successors of v
 748         for w in E[v]:
 749             if w not in vertex_index:
 750                 # Successor w has not yet been visited; recurse on it
 751                 index = strongconnect(w, index)
 752                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 753             elif w in S:
 754                 # Successor w is in stack S and hence in the current SCC
 755                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 756
 757         # If v is a root node, pop the stack and generate an SCC
 758         if vertex_lowlink[v] == vertex_index[v]:
 759             i = S.index(v)
 760             scc = S[i:]
 761             del S[i:]
 762             all_SCCs.append(scc)
 763
 764         return index
 765
 766     for v in V:
 767         if v not in vertex_index:
 768             index = strongconnect(v, index)
 769
 770     return all_SCCs
 771
 772
 773 def main():
 774     if sys.argv[1:] == ["--fixup"]:
 775         # Sort #include directives in-place.  Fixup mode doesn't solve
 776         # all possible silliness that the script checks for; it's just a
 777         # hack for the common case where renaming a header causes style
 778         # errors.
 779         fixup = True
 780     elif sys.argv[1:] == []:
 781         fixup = False
 782     else:
 783         print("TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 784               "line options: " + repr(sys.argv[1:]))
 785         sys.exit(1)
 786
 787     ok = check_style(fixup)
 788
 789     if ok:
 790         print('TEST-PASS | check_spidermonkey_style.py | ok')
 791     else:
 792         print('TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | ' +
 793               'actual output does not match expected output;  diff is above.')
 794         print('TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | ' +
 795               'Hint: If the problem is that you renamed a header, and many #includes ' +
 796               'are no longer in alphabetical order, commit your work and then try ' +
 797               '`check_spidermonkey_style.py --fixup`. ' +
 798               'You need to commit first because --fixup modifies your files in place.')
 799
 800     sys.exit(0 if ok else 1)
 801
 802
 803 if __name__ == '__main__':
 804     main()