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 import difflib
  39 import os
  40 import re
  41 import sys
  42
  43 # We don't bother checking files in these directories, because they're (a) auxiliary or (b)
  44 # imported code that doesn't follow our coding style.
  45 ignored_js_src_dirs = [
  46     "js/src/config/",  # auxiliary stuff
  47     "js/src/ctypes/libffi/",  # imported code
  48     "js/src/devtools/",  # auxiliary stuff
  49     "js/src/editline/",  # imported code
  50     "js/src/gdb/",  # auxiliary stuff
  51     "js/src/vtune/",  # imported code
  52     "js/src/zydis/",  # imported code
  53 ]
  54
  55 # We ignore #includes of these files, because they don't follow the usual rules.
  56 included_inclnames_to_ignore = set(
  57     [
  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         "frontend/smoosh_generated.h",  # generated in $OBJDIR
  64         "gc/StatsPhasesGenerated.h",  # generated in $OBJDIR
  65         "gc/StatsPhasesGenerated.inc",  # generated in $OBJDIR
  66         "jit/ABIFunctionTypeGenerated.h",  # generated in $OBJDIR"
  67         "jit/AtomicOperationsGenerated.h",  # generated in $OBJDIR
  68         "jit/CacheIROpsGenerated.h",  # generated in $OBJDIR
  69         "jit/LIROpsGenerated.h",  # generated in $OBJDIR
  70         "jit/MIROpsGenerated.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/locid.h",  # ICU
  92         "unicode/uchar.h",  # ICU
  93         "unicode/uniset.h",  # ICU
  94         "unicode/unistr.h",  # ICU
  95         "unicode/utypes.h",  # ICU
  96         "vtune/VTuneWrapper.h",  # VTune
  97         "wasm/WasmBuiltinModuleGenerated.h",  # generated in $OBJDIR"
  98         "zydis/ZydisAPI.h",  # Zydis
  99     ]
 100 )
 101
 102 deprecated_inclnames = {
 103     "mozilla/Unused.h": "Use [[nodiscard]] and (void)expr casts instead.",
 104 }
 105
 106 # JSAPI functions should be included through headers from js/public instead of
 107 # using the old, catch-all jsapi.h file.
 108 deprecated_inclnames_in_header = {
 109     "jsapi.h": "Prefer including headers from js/public.",
 110 }
 111
 112 # Temporary exclusions for files which still need to include jsapi.h.
 113 deprecated_inclnames_in_header_excludes = {
 114     "js/src/vm/Compartment-inl.h",  # for JS::InformalValueTypeName
 115     "js/src/jsapi-tests/tests.h",  # for JS_ValueToSource
 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     [
 122         "ctypes/typedefs.h",  # Included multiple times in the body of ctypes/CTypes.h
 123         # Included in the body of frontend/TokenStream.h
 124         "frontend/ReservedWordsGenerated.h",
 125         "gc/StatsPhasesGenerated.h",  # Included in the body of gc/Statistics.h
 126         "gc/StatsPhasesGenerated.inc",  # Included in the body of gc/Statistics.cpp
 127         "psapi.h",  # Must be included after "util/WindowsWrapper.h" on Windows
 128         "machine/endian.h",  # Must be included after <sys/types.h> on BSD
 129         "process.h",  # Windows-specific
 130         "winbase.h",  # Must precede other system headers(?)
 131         "windef.h",  # Must precede other system headers(?)
 132         "windows.h",  # Must precede other system headers(?)
 133     ]
 134 )
 135
 136 # The files in tests/style/ contain code that fails this checking in various
 137 # ways.  Here is the output we expect.  If the actual output differs from
 138 # this, one of the following must have happened.
 139 # - New SpiderMonkey code violates one of the checked rules.
 140 # - The tests/style/ files have changed without expected_output being changed
 141 #   accordingly.
 142 # - This script has been broken somehow.
 143 #
 144 expected_output = """\
 145 js/src/tests/style/BadIncludes.h:3: error:
 146     the file includes itself
 147
 148 js/src/tests/style/BadIncludes.h:6: error:
 149     "BadIncludes2.h" is included using the wrong path;
 150     did you forget a prefix, or is the file not yet committed?
 151
 152 js/src/tests/style/BadIncludes.h:8: error:
 153     <tests/style/BadIncludes2.h> should be included using
 154     the #include "..." form
 155
 156 js/src/tests/style/BadIncludes.h:10: error:
 157     "stdio.h" is included using the wrong path;
 158     did you forget a prefix, or is the file not yet committed?
 159
 160 js/src/tests/style/BadIncludes.h:12: error:
 161     "mozilla/Unused.h" is deprecated: Use [[nodiscard]] and (void)expr casts instead.
 162
 163 js/src/tests/style/BadIncludes2.h:1: error:
 164     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 165
 166 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 167     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 168
 169 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 170     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 171
 172 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 173     "js/Value.h" should be included after "ds/LifoAlloc.h"
 174
 175 js/src/tests/style/BadIncludesOrder-inl.h:9: error:
 176     "jsapi.h" is deprecated: Prefer including headers from js/public.
 177
 178 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 179     "ds/LifoAlloc.h" should be included after "jsapi.h"
 180
 181 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 182     "jsapi.h" should be included after <stdio.h>
 183
 184 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 185     <stdio.h> should be included after "mozilla/HashFunctions.h"
 186
 187 js/src/tests/style/BadIncludesOrder-inl.h:20: error:
 188     "jsapi.h" is deprecated: Prefer including headers from js/public.
 189
 190 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 191     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 192
 193 (multiple files): error:
 194     header files form one or more cycles
 195
 196    tests/style/HeaderCycleA1.h
 197    -> tests/style/HeaderCycleA2.h
 198       -> tests/style/HeaderCycleA3.h
 199          -> tests/style/HeaderCycleA1.h
 200
 201    tests/style/HeaderCycleB1-inl.h
 202    -> tests/style/HeaderCycleB2-inl.h
 203       -> tests/style/HeaderCycleB3-inl.h
 204          -> tests/style/HeaderCycleB4-inl.h
 205             -> tests/style/HeaderCycleB1-inl.h
 206             -> tests/style/jsheadercycleB5inlines.h
 207                -> tests/style/HeaderCycleB1-inl.h
 208       -> tests/style/HeaderCycleB4-inl.h
 209
 210 """.splitlines(
 211     True
 212 )
 213
 214 actual_output = []
 215
 216
 217 def out(*lines):
 218     for line in lines:
 219         actual_output.append(line + "\n")
 220
 221
 222 def error(filename, linenum, *lines):
 223     location = filename
 224     if linenum is not None:
 225         location += ":" + str(linenum)
 226     out(location + ": error:")
 227     for line in lines:
 228         out("    " + line)
 229     out("")
 230
 231
 232 class FileKind(object):
 233     C = 1
 234     CPP = 2
 235     INL_H = 3
 236     H = 4
 237     TBL = 5
 238     MSG = 6
 239
 240     @staticmethod
 241     def get(filename):
 242         if filename.endswith(".c"):
 243             return FileKind.C
 244
 245         if filename.endswith(".cpp"):
 246             return FileKind.CPP
 247
 248         if filename.endswith(("inlines.h", "-inl.h")):
 249             return FileKind.INL_H
 250
 251         if filename.endswith(".h"):
 252             return FileKind.H
 253
 254         if filename.endswith(".tbl"):
 255             return FileKind.TBL
 256
 257         if filename.endswith(".msg"):
 258             return FileKind.MSG
 259
 260         error(filename, None, "unknown file kind")
 261
 262
 263 def check_style(enable_fixup):
 264     # We deal with two kinds of name.
 265     # - A "filename" is a full path to a file from the repository root.
 266     # - An "inclname" is how a file is referred to in a #include statement.
 267     #
 268     # Examples (filename -> inclname)
 269     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 270     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 271     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 272     # - "js/public/Vector.h"        -> "js/Vector.h"
 273     # - "js/src/vm/String.h"        -> "vm/String.h"
 274
 275     non_js_dirnames = (
 276         "mfbt/",
 277         "memory/mozalloc/",
 278         "mozglue/",
 279         "intl/components/",
 280     )  # type: tuple(str)
 281     non_js_inclnames = set()  # type: set(inclname)
 282     js_names = dict()  # type: dict(filename, inclname)
 283
 284     # Process files in js/src.
 285     js_src_root = os.path.join("js", "src")
 286     for dirpath, dirnames, filenames in os.walk(js_src_root):
 287         if dirpath == js_src_root:
 288             # Skip any subdirectories that contain a config.status file
 289             # (likely objdirs).
 290             builddirs = []
 291             for dirname in dirnames:
 292                 path = os.path.join(dirpath, dirname, "config.status")
 293                 if os.path.isfile(path):
 294                     builddirs.append(dirname)
 295             for dirname in builddirs:
 296                 dirnames.remove(dirname)
 297         for filename in filenames:
 298             filepath = os.path.join(dirpath, filename).replace("\\", "/")
 299             if not filepath.startswith(
 300                 tuple(ignored_js_src_dirs)
 301             ) and filepath.endswith((".c", ".cpp", ".h", ".tbl", ".msg")):
 302                 inclname = filepath[len("js/src/") :]
 303                 js_names[filepath] = inclname
 304
 305     # Look for header files in directories in non_js_dirnames.
 306     for non_js_dir in non_js_dirnames:
 307         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 308             for filename in filenames:
 309                 if filename.endswith(".h"):
 310                     inclname = "mozilla/" + filename
 311                     if non_js_dir == "intl/components/":
 312                         inclname = "mozilla/intl/" + filename
 313                     non_js_inclnames.add(inclname)
 314
 315     # Look for header files in js/public.
 316     js_public_root = os.path.join("js", "public")
 317     for dirpath, dirnames, filenames in os.walk(js_public_root):
 318         for filename in filenames:
 319             if filename.endswith((".h", ".msg")):
 320                 filepath = os.path.join(dirpath, filename).replace("\\", "/")
 321                 inclname = "js/" + filepath[len("js/public/") :]
 322                 js_names[filepath] = inclname
 323
 324     all_inclnames = non_js_inclnames | set(js_names.values())
 325
 326     edges = dict()  # type: dict(inclname, set(inclname))
 327
 328     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 329     # are #included from JS files we have to add them to the inclusion graph.
 330     for inclname in non_js_inclnames:
 331         edges[inclname] = set()
 332
 333     # Process all the JS files.
 334     for filename in sorted(js_names.keys()):
 335         inclname = js_names[filename]
 336         file_kind = FileKind.get(filename)
 337         if (
 338             file_kind == FileKind.C
 339             or file_kind == FileKind.CPP
 340             or file_kind == FileKind.H
 341             or file_kind == FileKind.INL_H
 342         ):
 343             included_h_inclnames = set()  # type: set(inclname)
 344
 345             with open(filename, encoding="utf-8") as f:
 346                 code = read_file(f)
 347
 348             if enable_fixup:
 349                 code = code.sorted(inclname)
 350                 with open(filename, "w") as f:
 351                     f.write(code.to_source())
 352
 353             check_file(
 354                 filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 355             )
 356
 357         edges[inclname] = included_h_inclnames
 358
 359     find_cycles(all_inclnames, edges)
 360
 361     # Compare expected and actual output.
 362     difflines = difflib.unified_diff(
 363         expected_output,
 364         actual_output,
 365         fromfile="check_spidermonkey_style.py expected output",
 366         tofile="check_spidermonkey_style.py actual output",
 367     )
 368     ok = True
 369     for diffline in difflines:
 370         ok = False
 371         print(diffline, end="")
 372
 373     return ok
 374
 375
 376 def module_name(name):
 377     """Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename."""
 378
 379     return (
 380         name.replace("inlines.h", "")
 381         .replace("-inl.h", "")
 382         .replace(".h", "")
 383         .replace(".cpp", "")
 384     )  # NOQA: E501
 385
 386
 387 def is_module_header(enclosing_inclname, header_inclname):
 388     """Determine if an included name is the "module header", i.e. should be
 389     first in the file."""
 390
 391     module = module_name(enclosing_inclname)
 392
 393     # Normal case, for example:
 394     #   module == "vm/Runtime", header_inclname == "vm/Runtime.h".
 395     if module == module_name(header_inclname):
 396         return True
 397
 398     # A public header, for example:
 399     #
 400     #   module == "vm/CharacterEncoding",
 401     #   header_inclname == "js/CharacterEncoding.h"
 402     #
 403     # or (for implementation files for js/public/*/*.h headers)
 404     #
 405     #   module == "vm/SourceHook",
 406     #   header_inclname == "js/experimental/SourceHook.h"
 407     m = re.match(r"js\/.*?([^\/]+)\.h", header_inclname)
 408     if m is not None and module.endswith("/" + m.group(1)):
 409         return True
 410
 411     return False
 412
 413
 414 class Include(object):
 415     """Important information for a single #include statement."""
 416
 417     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 418         self.include_prefix = include_prefix
 419         self.line_suffix = line_suffix
 420         self.inclname = inclname
 421         self.linenum = linenum
 422         self.is_system = is_system
 423
 424     def is_style_relevant(self):
 425         # Includes are style-relevant; that is, they're checked by the pairwise
 426         # style-checking algorithm in check_file.
 427         return True
 428
 429     def section(self, enclosing_inclname):
 430         """Identify which section inclname belongs to.
 431
 432         The section numbers are as follows.
 433           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 434           1. mozilla/Foo.h
 435           2. <foo.h> or <foo>
 436           3. jsfoo.h, prmjtime.h, etc
 437           4. foo/Bar.h
 438           5. jsfooinlines.h
 439           6. foo/Bar-inl.h
 440           7. non-.h, e.g. *.tbl, *.msg (these can be scattered throughout files)
 441         """
 442
 443         if self.is_system:
 444             return 2
 445
 446         if not self.inclname.endswith(".h"):
 447             return 7
 448
 449         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 450         # special handling.
 451         if is_module_header(enclosing_inclname, self.inclname):
 452             return 0
 453
 454         if "/" in self.inclname:
 455             if self.inclname.startswith("mozilla/"):
 456                 return 1
 457
 458             if self.inclname.endswith("-inl.h"):
 459                 return 6
 460
 461             return 4
 462
 463         if self.inclname.endswith("inlines.h"):
 464             return 5
 465
 466         return 3
 467
 468     def quote(self):
 469         if self.is_system:
 470             return "<" + self.inclname + ">"
 471         else:
 472             return '"' + self.inclname + '"'
 473
 474     def sort_key(self, enclosing_inclname):
 475         return (self.section(enclosing_inclname), self.inclname.lower())
 476
 477     def to_source(self):
 478         return self.include_prefix + self.quote() + self.line_suffix + "\n"
 479
 480
 481 class CppBlock(object):
 482     """C preprocessor block: a whole file or a single #if/#elif/#else block.
 483
 484     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 485     The top-level block, which is not within a #if/#endif pair, is also
 486     considered a block.
 487
 488     Each kid is either an Include (representing a #include), OrdinaryCode, or
 489     a nested CppBlock."""
 490
 491     def __init__(self, start_line=""):
 492         self.start = start_line
 493         self.end = ""
 494         self.kids = []
 495
 496     def is_style_relevant(self):
 497         return True
 498
 499     def append_ordinary_line(self, line):
 500         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 501             self.kids.append(OrdinaryCode())
 502         self.kids[-1].lines.append(line)
 503
 504     def style_relevant_kids(self):
 505         """Return a list of kids in this block that are style-relevant."""
 506         return [kid for kid in self.kids if kid.is_style_relevant()]
 507
 508     def sorted(self, enclosing_inclname):
 509         """Return a hopefully-sorted copy of this block. Implements --fixup.
 510
 511         When in doubt, this leaves the code unchanged.
 512         """
 513
 514         def pretty_sorted_includes(includes):
 515             """Return a new list containing the given includes, in order,
 516             with blank lines separating sections."""
 517             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 518             if sorted(keys) == keys:
 519                 return includes  # if nothing is out of order, don't touch anything
 520
 521             output = []
 522             current_section = None
 523             for (section, _), inc in sorted(zip(keys, includes)):
 524                 if current_section is not None and section != current_section:
 525                     output.append(OrdinaryCode(["\n"]))  # blank line
 526                 output.append(inc)
 527                 current_section = section
 528             return output
 529
 530         def should_try_to_sort(includes):
 531             if "tests/style/BadIncludes" in enclosing_inclname:
 532                 return False  # don't straighten the counterexample
 533             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 534                 return False  # don't sort batches containing odd includes
 535             if includes == sorted(
 536                 includes, key=lambda inc: inc.sort_key(enclosing_inclname)
 537             ):
 538                 return False  # it's already sorted, avoid whitespace-only fixups
 539             return True
 540
 541         # The content of the eventual output of this method.
 542         output = []
 543
 544         # The current batch of includes to sort. This list only ever contains Include objects
 545         # and whitespace OrdinaryCode objects.
 546         batch = []
 547
 548         def flush_batch():
 549             """Sort the contents of `batch` and move it to `output`."""
 550
 551             assert all(
 552                 isinstance(item, Include)
 553                 or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 554                 for item in batch
 555             )
 556
 557             # Here we throw away the blank lines.
 558             # `pretty_sorted_includes` puts them back.
 559             includes = []
 560             last_include_index = -1
 561             for i, item in enumerate(batch):
 562                 if isinstance(item, Include):
 563                     includes.append(item)
 564                     last_include_index = i
 565             cutoff = last_include_index + 1
 566
 567             if should_try_to_sort(includes):
 568                 output.extend(pretty_sorted_includes(includes) + batch[cutoff:])
 569             else:
 570                 output.extend(batch)
 571             del batch[:]
 572
 573         for kid in self.kids:
 574             if isinstance(kid, CppBlock):
 575                 flush_batch()
 576                 output.append(kid.sorted(enclosing_inclname))
 577             elif isinstance(kid, Include):
 578                 batch.append(kid)
 579             else:
 580                 assert isinstance(kid, OrdinaryCode)
 581                 if kid.to_source().isspace():
 582                     batch.append(kid)
 583                 else:
 584                     flush_batch()
 585                     output.append(kid)
 586         flush_batch()
 587
 588         result = CppBlock()
 589         result.start = self.start
 590         result.end = self.end
 591         result.kids = output
 592         return result
 593
 594     def to_source(self):
 595         return self.start + "".join(kid.to_source() for kid in self.kids) + self.end
 596
 597
 598 class OrdinaryCode(object):
 599     """A list of lines of code that aren't #include/#if/#else/#endif lines."""
 600
 601     def __init__(self, lines=None):
 602         self.lines = lines if lines is not None else []
 603
 604     def is_style_relevant(self):
 605         return False
 606
 607     def to_source(self):
 608         return "".join(self.lines)
 609
 610
 611 # A "snippet" is one of:
 612 #
 613 # *   Include - representing an #include line
 614 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 615 # *   OrdinaryCode - representing lines of non-#include-relevant code
 616
 617
 618 def read_file(f):
 619     block_stack = [CppBlock()]
 620
 621     # Extract the #include statements as a tree of snippets.
 622     for linenum, line in enumerate(f, start=1):
 623         if line.lstrip().startswith("#"):
 624             # Look for a |#include "..."| line.
 625             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 626             if m is not None:
 627                 prefix, inclname, suffix = m.groups()
 628                 block_stack[-1].kids.append(
 629                     Include(prefix, inclname, suffix, linenum, is_system=False)
 630                 )
 631                 continue
 632
 633             # Look for a |#include <...>| line.
 634             m = re.match(r"(\s*#\s*include\s+)<([^>]*)>(.*)", line)
 635             if m is not None:
 636                 prefix, inclname, suffix = m.groups()
 637                 block_stack[-1].kids.append(
 638                     Include(prefix, inclname, suffix, linenum, is_system=True)
 639                 )
 640                 continue
 641
 642             # Look for a |#{if,ifdef,ifndef}| line.
 643             m = re.match(r"\s*#\s*(if|ifdef|ifndef)\b", line)
 644             if m is not None:
 645                 # Open a new block.
 646                 new_block = CppBlock(line)
 647                 block_stack[-1].kids.append(new_block)
 648                 block_stack.append(new_block)
 649                 continue
 650
 651             # Look for a |#{elif,else}| line.
 652             m = re.match(r"\s*#\s*(elif|else)\b", line)
 653             if m is not None:
 654                 # Close the current block, and open an adjacent one.
 655                 block_stack.pop()
 656                 new_block = CppBlock(line)
 657                 block_stack[-1].kids.append(new_block)
 658                 block_stack.append(new_block)
 659                 continue
 660
 661             # Look for a |#endif| line.
 662             m = re.match(r"\s*#\s*endif\b", line)
 663             if m is not None:
 664                 # Close the current block.
 665                 block_stack.pop().end = line
 666                 if len(block_stack) == 0:
 667                     raise ValueError("#endif without #if at line " + str(linenum))
 668                 continue
 669
 670         # Otherwise, we have an ordinary line.
 671         block_stack[-1].append_ordinary_line(line)
 672
 673     if len(block_stack) > 1:
 674         raise ValueError("unmatched #if")
 675     return block_stack[-1]
 676
 677
 678 def check_file(
 679     filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 680 ):
 681     def check_include_statement(include):
 682         """Check the style of a single #include statement."""
 683
 684         if include.is_system:
 685             # Check it is not a known local file (in which case it's probably a system header).
 686             if (
 687                 include.inclname in included_inclnames_to_ignore
 688                 or include.inclname in all_inclnames
 689             ):
 690                 error(
 691                     filename,
 692                     include.linenum,
 693                     include.quote() + " should be included using",
 694                     'the #include "..." form',
 695                 )
 696
 697         else:
 698             msg = deprecated_inclnames.get(include.inclname)
 699             if msg:
 700                 error(
 701                     filename,
 702                     include.linenum,
 703                     include.quote() + " is deprecated: " + msg,
 704                 )
 705
 706             if file_kind == FileKind.H or file_kind == FileKind.INL_H:
 707                 msg = deprecated_inclnames_in_header.get(include.inclname)
 708                 if msg and filename not in deprecated_inclnames_in_header_excludes:
 709                     error(
 710                         filename,
 711                         include.linenum,
 712                         include.quote() + " is deprecated: " + msg,
 713                     )
 714
 715             if include.inclname not in included_inclnames_to_ignore:
 716                 included_kind = FileKind.get(include.inclname)
 717
 718                 # Check the #include path has the correct form.
 719                 if include.inclname not in all_inclnames:
 720                     error(
 721                         filename,
 722                         include.linenum,
 723                         include.quote() + " is included using the wrong path;",
 724                         "did you forget a prefix, or is the file not yet committed?",
 725                     )
 726
 727                 # Record inclusions of .h files for cycle detection later.
 728                 # (Exclude .tbl and .msg files.)
 729                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 730                     included_h_inclnames.add(include.inclname)
 731
 732                 # Check a H file doesn't #include an INL_H file.
 733                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 734                     error(
 735                         filename,
 736                         include.linenum,
 737                         "vanilla header includes an inline-header file "
 738                         + include.quote(),
 739                     )
 740
 741                 # Check a file doesn't #include itself.  (We do this here because the cycle
 742                 # detection below doesn't detect this case.)
 743                 if inclname == include.inclname:
 744                     error(filename, include.linenum, "the file includes itself")
 745
 746     def check_includes_order(include1, include2):
 747         """Check the ordering of two #include statements."""
 748
 749         if (
 750             include1.inclname in oddly_ordered_inclnames
 751             or include2.inclname in oddly_ordered_inclnames
 752         ):
 753             return
 754
 755         section1 = include1.section(inclname)
 756         section2 = include2.section(inclname)
 757         if (section1 > section2) or (
 758             (section1 == section2)
 759             and (include1.inclname.lower() > include2.inclname.lower())
 760         ):
 761             error(
 762                 filename,
 763                 str(include1.linenum) + ":" + str(include2.linenum),
 764                 include1.quote() + " should be included after " + include2.quote(),
 765             )
 766
 767     # Check the extracted #include statements, both individually, and the ordering of
 768     # adjacent pairs that live in the same block.
 769     def pair_traverse(prev, this):
 770         if isinstance(this, Include):
 771             check_include_statement(this)
 772             if isinstance(prev, Include):
 773                 check_includes_order(prev, this)
 774         else:
 775             kids = this.style_relevant_kids()
 776             for prev2, this2 in zip([None] + kids[0:-1], kids):
 777                 pair_traverse(prev2, this2)
 778
 779     pair_traverse(None, code)
 780
 781
 782 def find_cycles(all_inclnames, edges):
 783     """Find and draw any cycles."""
 784
 785     SCCs = tarjan(all_inclnames, edges)
 786
 787     # The various sorted() calls below ensure the output is deterministic.
 788
 789     def draw_SCC(c):
 790         cset = set(c)
 791         drawn = set()
 792
 793         def draw(v, indent):
 794             out("   " * indent + ("-> " if indent else "   ") + v)
 795             if v in drawn:
 796                 return
 797             drawn.add(v)
 798             for succ in sorted(edges[v]):
 799                 if succ in cset:
 800                     draw(succ, indent + 1)
 801
 802         draw(sorted(c)[0], 0)
 803         out("")
 804
 805     have_drawn_an_SCC = False
 806     for scc in sorted(SCCs):
 807         if len(scc) != 1:
 808             if not have_drawn_an_SCC:
 809                 error("(multiple files)", None, "header files form one or more cycles")
 810                 have_drawn_an_SCC = True
 811
 812             draw_SCC(scc)
 813
 814
 815 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 816 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 817 def tarjan(V, E):
 818     vertex_index = {}
 819     vertex_lowlink = {}
 820     index = 0
 821     S = []
 822     all_SCCs = []
 823
 824     def strongconnect(v, index):
 825         # Set the depth index for v to the smallest unused index
 826         vertex_index[v] = index
 827         vertex_lowlink[v] = index
 828         index += 1
 829         S.append(v)
 830
 831         # Consider successors of v
 832         for w in E[v]:
 833             if w not in vertex_index:
 834                 # Successor w has not yet been visited; recurse on it
 835                 index = strongconnect(w, index)
 836                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 837             elif w in S:
 838                 # Successor w is in stack S and hence in the current SCC
 839                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 840
 841         # If v is a root node, pop the stack and generate an SCC
 842         if vertex_lowlink[v] == vertex_index[v]:
 843             i = S.index(v)
 844             scc = S[i:]
 845             del S[i:]
 846             all_SCCs.append(scc)
 847
 848         return index
 849
 850     for v in V:
 851         if v not in vertex_index:
 852             index = strongconnect(v, index)
 853
 854     return all_SCCs
 855
 856
 857 def main():
 858     if sys.argv[1:] == ["--fixup"]:
 859         # Sort #include directives in-place.  Fixup mode doesn't solve
 860         # all possible silliness that the script checks for; it's just a
 861         # hack for the common case where renaming a header causes style
 862         # errors.
 863         fixup = True
 864     elif sys.argv[1:] == []:
 865         fixup = False
 866     else:
 867         print(
 868             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 869             "line options: " + repr(sys.argv[1:])
 870         )
 871         sys.exit(1)
 872
 873     ok = check_style(fixup)
 874
 875     if ok:
 876         print("TEST-PASS | check_spidermonkey_style.py | ok")
 877     else:
 878         print(
 879             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 880             + "actual output does not match expected output;  diff is above."
 881         )
 882         print(
 883             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 884             + "Hint: If the problem is that you renamed a header, and many #includes "
 885             + "are no longer in alphabetical order, commit your work and then try "
 886             + "`check_spidermonkey_style.py --fixup`. "
 887             + "You need to commit first because --fixup modifies your files in place."
 888         )
 889
 890     sys.exit(0 if ok else 1)
 891
 892
 893 if __name__ == "__main__":
 894     main()