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     [
  60         "ffi.h",  # generated in ctypes/libffi/
  61         "devtools/Instruments.h",  # we ignore devtools/ in general
  62         "double-conversion/double-conversion.h",  # strange MFBT case
  63         "javascript-trace.h",  # generated in $OBJDIR if HAVE_DTRACE is defined
  64         "frontend/ReservedWordsGenerated.h",  # generated in $OBJDIR
  65         "frontend/smoosh_generated.h",  # generated in $OBJDIR
  66         "gc/StatsPhasesGenerated.h",  # generated in $OBJDIR
  67         "gc/StatsPhasesGenerated.inc",  # generated in $OBJDIR
  68         "jit/CacheIROpsGenerated.h",  # generated in $OBJDIR
  69         "jit/LOpcodesGenerated.h",  # generated in $OBJDIR
  70         "jit/MOpcodesGenerated.h",  # generated in $OBJDIR
  71         "js/ProfilingCategoryList.h",  # comes from mozglue/baseprofiler
  72         "jscustomallocator.h",  # provided by embedders;  allowed to be missing
  73         "js-config.h",  # generated in $OBJDIR
  74         "fdlibm.h",  # fdlibm
  75         "FuzzerDefs.h",  # included without a path
  76         "FuzzingInterface.h",  # included without a path
  77         "mozmemory.h",  # included without a path
  78         "pratom.h",  # NSPR
  79         "prcvar.h",  # NSPR
  80         "prerror.h",  # NSPR
  81         "prinit.h",  # NSPR
  82         "prio.h",  # NSPR
  83         "private/pprio.h",  # NSPR
  84         "prlink.h",  # NSPR
  85         "prlock.h",  # NSPR
  86         "prprf.h",  # NSPR
  87         "prthread.h",  # NSPR
  88         "prtypes.h",  # NSPR
  89         "selfhosted.out.h",  # generated in $OBJDIR
  90         "shellmoduleloader.out.h",  # generated in $OBJDIR
  91         "unicode/basictz.h",  # ICU
  92         "unicode/locid.h",  # ICU
  93         "unicode/plurrule.h",  # ICU
  94         "unicode/putil.h",  # ICU
  95         "unicode/timezone.h",  # ICU
  96         "unicode/ucal.h",  # ICU
  97         "unicode/uchar.h",  # ICU
  98         "unicode/uclean.h",  # ICU
  99         "unicode/ucol.h",  # ICU
 100         "unicode/ucurr.h",  # ICU
 101         "unicode/udat.h",  # ICU
 102         "unicode/udata.h",  # ICU
 103         "unicode/udateintervalformat.h",  # ICU
 104         "unicode/udatpg.h",  # ICU
 105         "unicode/udisplaycontext.h",  # ICU
 106         "unicode/uenum.h",  # ICU
 107         "unicode/ufieldpositer.h",  # ICU
 108         "unicode/uformattedvalue.h",  # ICU
 109         "unicode/ulistformatter.h",  # ICU
 110         "unicode/uldnames.h",  # ICU
 111         "unicode/uloc.h",  # ICU
 112         "unicode/umachine.h",  # ICU
 113         "unicode/uniset.h",  # ICU
 114         "unicode/unistr.h",  # ICU
 115         "unicode/unorm2.h",  # ICU
 116         "unicode/unum.h",  # ICU
 117         "unicode/unumberformatter.h",  # ICU
 118         "unicode/unumsys.h",  # ICU
 119         "unicode/upluralrules.h",  # ICU
 120         "unicode/ureldatefmt.h",  # ICU
 121         "unicode/ures.h",  # ICU
 122         "unicode/ustring.h",  # ICU
 123         "unicode/utypes.h",  # ICU
 124         "unicode/uversion.h",  # ICU
 125         "vtune/VTuneWrapper.h",  # VTune
 126         "zydis/ZydisAPI.h",  # Zydis
 127     ]
 128 )
 129
 130 # These files have additional constraints on where they are #included, so we
 131 # ignore #includes of them when checking #include ordering.
 132 oddly_ordered_inclnames = set(
 133     [
 134         "ctypes/typedefs.h",  # Included multiple times in the body of ctypes/CTypes.h
 135         # Included in the body of frontend/TokenStream.h
 136         "frontend/ReservedWordsGenerated.h",
 137         "gc/StatsPhasesGenerated.h",  # Included in the body of gc/Statistics.h
 138         "gc/StatsPhasesGenerated.inc",  # Included in the body of gc/Statistics.cpp
 139         "psapi.h",  # Must be included after "util/Windows.h" on Windows
 140         "machine/endian.h",  # Must be included after <sys/types.h> on BSD
 141         "winbase.h",  # Must precede other system headers(?)
 142         "windef.h",  # Must precede other system headers(?)
 143     ]
 144 )
 145
 146 # The files in tests/style/ contain code that fails this checking in various
 147 # ways.  Here is the output we expect.  If the actual output differs from
 148 # this, one of the following must have happened.
 149 # - New SpiderMonkey code violates one of the checked rules.
 150 # - The tests/style/ files have changed without expected_output being changed
 151 #   accordingly.
 152 # - This script has been broken somehow.
 153 #
 154 expected_output = """\
 155 js/src/tests/style/BadIncludes.h:3: error:
 156     the file includes itself
 157
 158 js/src/tests/style/BadIncludes.h:6: error:
 159     "BadIncludes2.h" is included using the wrong path;
 160     did you forget a prefix, or is the file not yet committed?
 161
 162 js/src/tests/style/BadIncludes.h:8: error:
 163     <tests/style/BadIncludes2.h> should be included using
 164     the #include "..." form
 165
 166 js/src/tests/style/BadIncludes.h:10: error:
 167     "stdio.h" is included using the wrong path;
 168     did you forget a prefix, or is the file not yet committed?
 169
 170 js/src/tests/style/BadIncludes2.h:1: error:
 171     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 172
 173 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 174     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 175
 176 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 177     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 178
 179 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 180     "js/Value.h" should be included after "ds/LifoAlloc.h"
 181
 182 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 183     "ds/LifoAlloc.h" should be included after "jsapi.h"
 184
 185 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 186     "jsapi.h" should be included after <stdio.h>
 187
 188 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 189     <stdio.h> should be included after "mozilla/HashFunctions.h"
 190
 191 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 192     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 193
 194 (multiple files): error:
 195     header files form one or more cycles
 196
 197    tests/style/HeaderCycleA1.h
 198    -> tests/style/HeaderCycleA2.h
 199       -> tests/style/HeaderCycleA3.h
 200          -> tests/style/HeaderCycleA1.h
 201
 202    tests/style/HeaderCycleB1-inl.h
 203    -> tests/style/HeaderCycleB2-inl.h
 204       -> tests/style/HeaderCycleB3-inl.h
 205          -> tests/style/HeaderCycleB4-inl.h
 206             -> tests/style/HeaderCycleB1-inl.h
 207             -> tests/style/jsheadercycleB5inlines.h
 208                -> tests/style/HeaderCycleB1-inl.h
 209       -> tests/style/HeaderCycleB4-inl.h
 210
 211 """.splitlines(
 212     True
 213 )
 214
 215 actual_output = []
 216
 217
 218 def out(*lines):
 219     for line in lines:
 220         actual_output.append(line + "\n")
 221
 222
 223 def error(filename, linenum, *lines):
 224     location = filename
 225     if linenum is not None:
 226         location += ":" + str(linenum)
 227     out(location + ": error:")
 228     for line in lines:
 229         out("    " + line)
 230     out("")
 231
 232
 233 class FileKind(object):
 234     C = 1
 235     CPP = 2
 236     INL_H = 3
 237     H = 4
 238     TBL = 5
 239     MSG = 6
 240
 241     @staticmethod
 242     def get(filename):
 243         if filename.endswith(".c"):
 244             return FileKind.C
 245
 246         if filename.endswith(".cpp"):
 247             return FileKind.CPP
 248
 249         if filename.endswith(("inlines.h", "-inl.h")):
 250             return FileKind.INL_H
 251
 252         if filename.endswith(".h"):
 253             return FileKind.H
 254
 255         if filename.endswith(".tbl"):
 256             return FileKind.TBL
 257
 258         if filename.endswith(".msg"):
 259             return FileKind.MSG
 260
 261         error(filename, None, "unknown file kind")
 262
 263
 264 def check_style(enable_fixup):
 265     # We deal with two kinds of name.
 266     # - A "filename" is a full path to a file from the repository root.
 267     # - An "inclname" is how a file is referred to in a #include statement.
 268     #
 269     # Examples (filename -> inclname)
 270     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 271     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 272     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 273     # - "js/public/Vector.h"        -> "js/Vector.h"
 274     # - "js/src/vm/String.h"        -> "vm/String.h"
 275
 276     non_js_dirnames = ("mfbt/", "memory/mozalloc/", "mozglue/")  # type: tuple(str)
 277     non_js_inclnames = set()  # type: set(inclname)
 278     js_names = dict()  # type: dict(filename, inclname)
 279
 280     # Process files in js/src.
 281     js_src_root = os.path.join("js", "src")
 282     for dirpath, dirnames, filenames in os.walk(js_src_root):
 283         if dirpath == js_src_root:
 284             # Skip any subdirectories that contain a config.status file
 285             # (likely objdirs).
 286             builddirs = []
 287             for dirname in dirnames:
 288                 path = os.path.join(dirpath, dirname, "config.status")
 289                 if os.path.isfile(path):
 290                     builddirs.append(dirname)
 291             for dirname in builddirs:
 292                 dirnames.remove(dirname)
 293         for filename in filenames:
 294             filepath = os.path.join(dirpath, filename).replace("\\", "/")
 295             if not filepath.startswith(
 296                 tuple(ignored_js_src_dirs)
 297             ) and filepath.endswith((".c", ".cpp", ".h", ".tbl", ".msg")):
 298                 inclname = filepath[len("js/src/") :]
 299                 js_names[filepath] = inclname
 300
 301     # Look for header files in directories in non_js_dirnames.
 302     for non_js_dir in non_js_dirnames:
 303         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 304             for filename in filenames:
 305                 if filename.endswith(".h"):
 306                     inclname = "mozilla/" + filename
 307                     non_js_inclnames.add(inclname)
 308
 309     # Look for header files in js/public.
 310     js_public_root = os.path.join("js", "public")
 311     for dirpath, dirnames, filenames in os.walk(js_public_root):
 312         for filename in filenames:
 313             if filename.endswith((".h", ".msg")):
 314                 filepath = os.path.join(dirpath, filename).replace("\\", "/")
 315                 inclname = "js/" + filepath[len("js/public/") :]
 316                 js_names[filepath] = inclname
 317
 318     all_inclnames = non_js_inclnames | set(js_names.values())
 319
 320     edges = dict()  # type: dict(inclname, set(inclname))
 321
 322     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 323     # are #included from JS files we have to add them to the inclusion graph.
 324     for inclname in non_js_inclnames:
 325         edges[inclname] = set()
 326
 327     # Process all the JS files.
 328     for filename in sorted(js_names.keys()):
 329         inclname = js_names[filename]
 330         file_kind = FileKind.get(filename)
 331         if (
 332             file_kind == FileKind.C
 333             or file_kind == FileKind.CPP
 334             or file_kind == FileKind.H
 335             or file_kind == FileKind.INL_H
 336         ):
 337             included_h_inclnames = set()  # type: set(inclname)
 338
 339             with open(filename, encoding="utf-8") as f:
 340                 code = read_file(f)
 341
 342             if enable_fixup:
 343                 code = code.sorted(inclname)
 344                 with open(filename, "w") as f:
 345                     f.write(code.to_source())
 346
 347             check_file(
 348                 filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 349             )
 350
 351         edges[inclname] = included_h_inclnames
 352
 353     find_cycles(all_inclnames, edges)
 354
 355     # Compare expected and actual output.
 356     difflines = difflib.unified_diff(
 357         expected_output,
 358         actual_output,
 359         fromfile="check_spidermonkey_style.py expected output",
 360         tofile="check_spidermonkey_style.py actual output",
 361     )
 362     ok = True
 363     for diffline in difflines:
 364         ok = False
 365         print(diffline, end="")
 366
 367     return ok
 368
 369
 370 def module_name(name):
 371     """Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename."""
 372
 373     return (
 374         name.replace("inlines.h", "")
 375         .replace("-inl.h", "")
 376         .replace(".h", "")
 377         .replace(".cpp", "")
 378     )  # NOQA: E501
 379
 380
 381 def is_module_header(enclosing_inclname, header_inclname):
 382     """Determine if an included name is the "module header", i.e. should be
 383     first in the file."""
 384
 385     module = module_name(enclosing_inclname)
 386
 387     # Normal case, for example:
 388     #   module == "vm/Runtime", header_inclname == "vm/Runtime.h".
 389     if module == module_name(header_inclname):
 390         return True
 391
 392     # A public header, for example:
 393     #
 394     #   module == "vm/CharacterEncoding",
 395     #   header_inclname == "js/CharacterEncoding.h"
 396     #
 397     # or (for implementation files for js/public/*/*.h headers)
 398     #
 399     #   module == "vm/SourceHook",
 400     #   header_inclname == "js/experimental/SourceHook.h"
 401     m = re.match(r"js\/.*?([^\/]+)\.h", header_inclname)
 402     if m is not None and module.endswith("/" + m.group(1)):
 403         return True
 404
 405     return False
 406
 407
 408 class Include(object):
 409     """Important information for a single #include statement."""
 410
 411     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 412         self.include_prefix = include_prefix
 413         self.line_suffix = line_suffix
 414         self.inclname = inclname
 415         self.linenum = linenum
 416         self.is_system = is_system
 417
 418     def is_style_relevant(self):
 419         # Includes are style-relevant; that is, they're checked by the pairwise
 420         # style-checking algorithm in check_file.
 421         return True
 422
 423     def section(self, enclosing_inclname):
 424         """Identify which section inclname belongs to.
 425
 426         The section numbers are as follows.
 427           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 428           1. mozilla/Foo.h
 429           2. <foo.h> or <foo>
 430           3. jsfoo.h, prmjtime.h, etc
 431           4. foo/Bar.h
 432           5. jsfooinlines.h
 433           6. foo/Bar-inl.h
 434           7. non-.h, e.g. *.tbl, *.msg (these can be scattered throughout files)
 435         """
 436
 437         if self.is_system:
 438             return 2
 439
 440         if not self.inclname.endswith(".h"):
 441             return 7
 442
 443         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 444         # special handling.
 445         if is_module_header(enclosing_inclname, self.inclname):
 446             return 0
 447
 448         if "/" in self.inclname:
 449             if self.inclname.startswith("mozilla/"):
 450                 return 1
 451
 452             if self.inclname.endswith("-inl.h"):
 453                 return 6
 454
 455             return 4
 456
 457         if self.inclname.endswith("inlines.h"):
 458             return 5
 459
 460         return 3
 461
 462     def quote(self):
 463         if self.is_system:
 464             return "<" + self.inclname + ">"
 465         else:
 466             return '"' + self.inclname + '"'
 467
 468     def sort_key(self, enclosing_inclname):
 469         return (self.section(enclosing_inclname), self.inclname.lower())
 470
 471     def to_source(self):
 472         return self.include_prefix + self.quote() + self.line_suffix + "\n"
 473
 474
 475 class CppBlock(object):
 476     """C preprocessor block: a whole file or a single #if/#elif/#else block.
 477
 478     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 479     The top-level block, which is not within a #if/#endif pair, is also
 480     considered a block.
 481
 482     Each kid is either an Include (representing a #include), OrdinaryCode, or
 483     a nested CppBlock."""
 484
 485     def __init__(self, start_line=""):
 486         self.start = start_line
 487         self.end = ""
 488         self.kids = []
 489
 490     def is_style_relevant(self):
 491         return True
 492
 493     def append_ordinary_line(self, line):
 494         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 495             self.kids.append(OrdinaryCode())
 496         self.kids[-1].lines.append(line)
 497
 498     def style_relevant_kids(self):
 499         """ Return a list of kids in this block that are style-relevant. """
 500         return [kid for kid in self.kids if kid.is_style_relevant()]
 501
 502     def sorted(self, enclosing_inclname):
 503         """Return a hopefully-sorted copy of this block. Implements --fixup.
 504
 505         When in doubt, this leaves the code unchanged.
 506         """
 507
 508         def pretty_sorted_includes(includes):
 509             """Return a new list containing the given includes, in order,
 510             with blank lines separating sections."""
 511             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 512             if sorted(keys) == keys:
 513                 return includes  # if nothing is out of order, don't touch anything
 514
 515             output = []
 516             current_section = None
 517             for (section, _), inc in sorted(zip(keys, includes)):
 518                 if current_section is not None and section != current_section:
 519                     output.append(OrdinaryCode(["\n"]))  # blank line
 520                 output.append(inc)
 521                 current_section = section
 522             return output
 523
 524         def should_try_to_sort(includes):
 525             if "tests/style/BadIncludes" in enclosing_inclname:
 526                 return False  # don't straighten the counterexample
 527             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 528                 return False  # don't sort batches containing odd includes
 529             if includes == sorted(
 530                 includes, key=lambda inc: inc.sort_key(enclosing_inclname)
 531             ):
 532                 return False  # it's already sorted, avoid whitespace-only fixups
 533             return True
 534
 535         # The content of the eventual output of this method.
 536         output = []
 537
 538         # The current batch of includes to sort. This list only ever contains Include objects
 539         # and whitespace OrdinaryCode objects.
 540         batch = []
 541
 542         def flush_batch():
 543             """Sort the contents of `batch` and move it to `output`."""
 544
 545             assert all(
 546                 isinstance(item, Include)
 547                 or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 548                 for item in batch
 549             )
 550
 551             # Here we throw away the blank lines.
 552             # `pretty_sorted_includes` puts them back.
 553             includes = []
 554             last_include_index = -1
 555             for i, item in enumerate(batch):
 556                 if isinstance(item, Include):
 557                     includes.append(item)
 558                     last_include_index = i
 559             cutoff = last_include_index + 1
 560
 561             if should_try_to_sort(includes):
 562                 output.extend(pretty_sorted_includes(includes) + batch[cutoff:])
 563             else:
 564                 output.extend(batch)
 565             del batch[:]
 566
 567         for kid in self.kids:
 568             if isinstance(kid, CppBlock):
 569                 flush_batch()
 570                 output.append(kid.sorted(enclosing_inclname))
 571             elif isinstance(kid, Include):
 572                 batch.append(kid)
 573             else:
 574                 assert isinstance(kid, OrdinaryCode)
 575                 if kid.to_source().isspace():
 576                     batch.append(kid)
 577                 else:
 578                     flush_batch()
 579                     output.append(kid)
 580         flush_batch()
 581
 582         result = CppBlock()
 583         result.start = self.start
 584         result.end = self.end
 585         result.kids = output
 586         return result
 587
 588     def to_source(self):
 589         return self.start + "".join(kid.to_source() for kid in self.kids) + self.end
 590
 591
 592 class OrdinaryCode(object):
 593     """ A list of lines of code that aren't #include/#if/#else/#endif lines. """
 594
 595     def __init__(self, lines=None):
 596         self.lines = lines if lines is not None else []
 597
 598     def is_style_relevant(self):
 599         return False
 600
 601     def to_source(self):
 602         return "".join(self.lines)
 603
 604
 605 # A "snippet" is one of:
 606 #
 607 # *   Include - representing an #include line
 608 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 609 # *   OrdinaryCode - representing lines of non-#include-relevant code
 610
 611
 612 def read_file(f):
 613     block_stack = [CppBlock()]
 614
 615     # Extract the #include statements as a tree of snippets.
 616     for linenum, line in enumerate(f, start=1):
 617         if line.lstrip().startswith("#"):
 618             # Look for a |#include "..."| line.
 619             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 620             if m is not None:
 621                 prefix, inclname, suffix = m.groups()
 622                 block_stack[-1].kids.append(
 623                     Include(prefix, inclname, suffix, linenum, is_system=False)
 624                 )
 625                 continue
 626
 627             # Look for a |#include <...>| line.
 628             m = re.match(r"(\s*#\s*include\s+)<([^>]*)>(.*)", line)
 629             if m is not None:
 630                 prefix, inclname, suffix = m.groups()
 631                 block_stack[-1].kids.append(
 632                     Include(prefix, inclname, suffix, linenum, is_system=True)
 633                 )
 634                 continue
 635
 636             # Look for a |#{if,ifdef,ifndef}| line.
 637             m = re.match(r"\s*#\s*(if|ifdef|ifndef)\b", line)
 638             if m is not None:
 639                 # Open a new block.
 640                 new_block = CppBlock(line)
 641                 block_stack[-1].kids.append(new_block)
 642                 block_stack.append(new_block)
 643                 continue
 644
 645             # Look for a |#{elif,else}| line.
 646             m = re.match(r"\s*#\s*(elif|else)\b", line)
 647             if m is not None:
 648                 # Close the current block, and open an adjacent one.
 649                 block_stack.pop()
 650                 new_block = CppBlock(line)
 651                 block_stack[-1].kids.append(new_block)
 652                 block_stack.append(new_block)
 653                 continue
 654
 655             # Look for a |#endif| line.
 656             m = re.match(r"\s*#\s*endif\b", line)
 657             if m is not None:
 658                 # Close the current block.
 659                 block_stack.pop().end = line
 660                 if len(block_stack) == 0:
 661                     raise ValueError("#endif without #if at line " + str(linenum))
 662                 continue
 663
 664         # Otherwise, we have an ordinary line.
 665         block_stack[-1].append_ordinary_line(line)
 666
 667     if len(block_stack) > 1:
 668         raise ValueError("unmatched #if")
 669     return block_stack[-1]
 670
 671
 672 def check_file(
 673     filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 674 ):
 675     def check_include_statement(include):
 676         """Check the style of a single #include statement."""
 677
 678         if include.is_system:
 679             # Check it is not a known local file (in which case it's probably a system header).
 680             if (
 681                 include.inclname in included_inclnames_to_ignore
 682                 or include.inclname in all_inclnames
 683             ):
 684                 error(
 685                     filename,
 686                     include.linenum,
 687                     include.quote() + " should be included using",
 688                     'the #include "..." form',
 689                 )
 690
 691         else:
 692             if include.inclname not in included_inclnames_to_ignore:
 693                 included_kind = FileKind.get(include.inclname)
 694
 695                 # Check the #include path has the correct form.
 696                 if include.inclname not in all_inclnames:
 697                     error(
 698                         filename,
 699                         include.linenum,
 700                         include.quote() + " is included using the wrong path;",
 701                         "did you forget a prefix, or is the file not yet committed?",
 702                     )
 703
 704                 # Record inclusions of .h files for cycle detection later.
 705                 # (Exclude .tbl and .msg files.)
 706                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 707                     included_h_inclnames.add(include.inclname)
 708
 709                 # Check a H file doesn't #include an INL_H file.
 710                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 711                     error(
 712                         filename,
 713                         include.linenum,
 714                         "vanilla header includes an inline-header file "
 715                         + include.quote(),
 716                     )
 717
 718                 # Check a file doesn't #include itself.  (We do this here because the cycle
 719                 # detection below doesn't detect this case.)
 720                 if inclname == include.inclname:
 721                     error(filename, include.linenum, "the file includes itself")
 722
 723     def check_includes_order(include1, include2):
 724         """Check the ordering of two #include statements."""
 725
 726         if (
 727             include1.inclname in oddly_ordered_inclnames
 728             or include2.inclname in oddly_ordered_inclnames
 729         ):
 730             return
 731
 732         section1 = include1.section(inclname)
 733         section2 = include2.section(inclname)
 734         if (section1 > section2) or (
 735             (section1 == section2)
 736             and (include1.inclname.lower() > include2.inclname.lower())
 737         ):
 738             error(
 739                 filename,
 740                 str(include1.linenum) + ":" + str(include2.linenum),
 741                 include1.quote() + " should be included after " + include2.quote(),
 742             )
 743
 744     # Check the extracted #include statements, both individually, and the ordering of
 745     # adjacent pairs that live in the same block.
 746     def pair_traverse(prev, this):
 747         if isinstance(this, Include):
 748             check_include_statement(this)
 749             if isinstance(prev, Include):
 750                 check_includes_order(prev, this)
 751         else:
 752             kids = this.style_relevant_kids()
 753             for prev2, this2 in zip([None] + kids[0:-1], kids):
 754                 pair_traverse(prev2, this2)
 755
 756     pair_traverse(None, code)
 757
 758
 759 def find_cycles(all_inclnames, edges):
 760     """Find and draw any cycles."""
 761
 762     SCCs = tarjan(all_inclnames, edges)
 763
 764     # The various sorted() calls below ensure the output is deterministic.
 765
 766     def draw_SCC(c):
 767         cset = set(c)
 768         drawn = set()
 769
 770         def draw(v, indent):
 771             out("   " * indent + ("-> " if indent else "   ") + v)
 772             if v in drawn:
 773                 return
 774             drawn.add(v)
 775             for succ in sorted(edges[v]):
 776                 if succ in cset:
 777                     draw(succ, indent + 1)
 778
 779         draw(sorted(c)[0], 0)
 780         out("")
 781
 782     have_drawn_an_SCC = False
 783     for scc in sorted(SCCs):
 784         if len(scc) != 1:
 785             if not have_drawn_an_SCC:
 786                 error("(multiple files)", None, "header files form one or more cycles")
 787                 have_drawn_an_SCC = True
 788
 789             draw_SCC(scc)
 790
 791
 792 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 793 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 794 def tarjan(V, E):
 795     vertex_index = {}
 796     vertex_lowlink = {}
 797     index = 0
 798     S = []
 799     all_SCCs = []
 800
 801     def strongconnect(v, index):
 802         # Set the depth index for v to the smallest unused index
 803         vertex_index[v] = index
 804         vertex_lowlink[v] = index
 805         index += 1
 806         S.append(v)
 807
 808         # Consider successors of v
 809         for w in E[v]:
 810             if w not in vertex_index:
 811                 # Successor w has not yet been visited; recurse on it
 812                 index = strongconnect(w, index)
 813                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 814             elif w in S:
 815                 # Successor w is in stack S and hence in the current SCC
 816                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 817
 818         # If v is a root node, pop the stack and generate an SCC
 819         if vertex_lowlink[v] == vertex_index[v]:
 820             i = S.index(v)
 821             scc = S[i:]
 822             del S[i:]
 823             all_SCCs.append(scc)
 824
 825         return index
 826
 827     for v in V:
 828         if v not in vertex_index:
 829             index = strongconnect(v, index)
 830
 831     return all_SCCs
 832
 833
 834 def main():
 835     if sys.argv[1:] == ["--fixup"]:
 836         # Sort #include directives in-place.  Fixup mode doesn't solve
 837         # all possible silliness that the script checks for; it's just a
 838         # hack for the common case where renaming a header causes style
 839         # errors.
 840         fixup = True
 841     elif sys.argv[1:] == []:
 842         fixup = False
 843     else:
 844         print(
 845             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 846             "line options: " + repr(sys.argv[1:])
 847         )
 848         sys.exit(1)
 849
 850     ok = check_style(fixup)
 851
 852     if ok:
 853         print("TEST-PASS | check_spidermonkey_style.py | ok")
 854     else:
 855         print(
 856             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 857             + "actual output does not match expected output;  diff is above."
 858         )
 859         print(
 860             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 861             + "Hint: If the problem is that you renamed a header, and many #includes "
 862             + "are no longer in alphabetical order, commit your work and then try "
 863             + "`check_spidermonkey_style.py --fixup`. "
 864             + "You need to commit first because --fixup modifies your files in place."
 865         )
 866
 867     sys.exit(0 if ok else 1)
 868
 869
 870 if __name__ == "__main__":
 871     main()