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