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         "mozilla/glue/Debug.h",  # comes from mozglue/misc, shadowed by <mozilla/Debug.h>
  73         "jscustomallocator.h",  # provided by embedders;  allowed to be missing
  74         "js-config.h",  # generated in $OBJDIR
  75         "fdlibm.h",  # fdlibm
  76         "FuzzerDefs.h",  # included without a path
  77         "FuzzingInterface.h",  # included without a path
  78         "ICU4XGraphemeClusterSegmenter.h",  # ICU4X
  79         "ICU4XSentenceSegmenter.h",  # ICU4X
  80         "ICU4XWordSegmenter.h",  # ICU4X
  81         "mozmemory.h",  # included without a path
  82         "pratom.h",  # NSPR
  83         "prcvar.h",  # NSPR
  84         "prerror.h",  # NSPR
  85         "prinit.h",  # NSPR
  86         "prio.h",  # NSPR
  87         "private/pprio.h",  # NSPR
  88         "prlink.h",  # NSPR
  89         "prlock.h",  # NSPR
  90         "prprf.h",  # NSPR
  91         "prthread.h",  # NSPR
  92         "prtypes.h",  # NSPR
  93         "selfhosted.out.h",  # generated in $OBJDIR
  94         "shellmoduleloader.out.h",  # generated in $OBJDIR
  95         "unicode/locid.h",  # ICU
  96         "unicode/uchar.h",  # ICU
  97         "unicode/uniset.h",  # ICU
  98         "unicode/unistr.h",  # ICU
  99         "unicode/utypes.h",  # ICU
 100         "vtune/VTuneWrapper.h",  # VTune
 101         "wasm/WasmBuiltinModuleGenerated.h",  # generated in $OBJDIR"
 102         "zydis/ZydisAPI.h",  # Zydis
 103     ]
 104 )
 105
 106 deprecated_inclnames = {
 107     "mozilla/Unused.h": "Use [[nodiscard]] and (void)expr casts instead.",
 108 }
 109
 110 # JSAPI functions should be included through headers from js/public instead of
 111 # using the old, catch-all jsapi.h file.
 112 deprecated_inclnames_in_header = {
 113     "jsapi.h": "Prefer including headers from js/public.",
 114 }
 115
 116 # Temporary exclusions for files which still need to include jsapi.h.
 117 deprecated_inclnames_in_header_excludes = {
 118     "js/src/vm/Compartment-inl.h",  # for JS::InformalValueTypeName
 119     "js/src/jsapi-tests/tests.h",  # for JS_ValueToSource
 120 }
 121
 122 # These files have additional constraints on where they are #included, so we
 123 # ignore #includes of them when checking #include ordering.
 124 oddly_ordered_inclnames = set(
 125     [
 126         "ctypes/typedefs.h",  # Included multiple times in the body of ctypes/CTypes.h
 127         # Included in the body of frontend/TokenStream.h
 128         "frontend/ReservedWordsGenerated.h",
 129         "gc/StatsPhasesGenerated.h",  # Included in the body of gc/Statistics.h
 130         "gc/StatsPhasesGenerated.inc",  # Included in the body of gc/Statistics.cpp
 131         "psapi.h",  # Must be included after "util/WindowsWrapper.h" on Windows
 132         "machine/endian.h",  # Must be included after <sys/types.h> on BSD
 133         "process.h",  # Windows-specific
 134         "winbase.h",  # Must precede other system headers(?)
 135         "windef.h",  # Must precede other system headers(?)
 136         "windows.h",  # Must precede other system headers(?)
 137     ]
 138 )
 139
 140 # The files in tests/style/ contain code that fails this checking in various
 141 # ways.  Here is the output we expect.  If the actual output differs from
 142 # this, one of the following must have happened.
 143 # - New SpiderMonkey code violates one of the checked rules.
 144 # - The tests/style/ files have changed without expected_output being changed
 145 #   accordingly.
 146 # - This script has been broken somehow.
 147 #
 148 expected_output = """\
 149 js/src/tests/style/BadIncludes.h:3: error:
 150     the file includes itself
 151
 152 js/src/tests/style/BadIncludes.h:6: error:
 153     "BadIncludes2.h" is included using the wrong path;
 154     did you forget a prefix, or is the file not yet committed?
 155
 156 js/src/tests/style/BadIncludes.h:8: error:
 157     <tests/style/BadIncludes2.h> should be included using
 158     the #include "..." form
 159
 160 js/src/tests/style/BadIncludes.h:10: error:
 161     "stdio.h" is included using the wrong path;
 162     did you forget a prefix, or is the file not yet committed?
 163
 164 js/src/tests/style/BadIncludes.h:12: error:
 165     "mozilla/Unused.h" is deprecated: Use [[nodiscard]] and (void)expr casts instead.
 166
 167 js/src/tests/style/BadIncludes2.h:1: error:
 168     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 169
 170 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 171     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 172
 173 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 174     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 175
 176 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 177     "js/Value.h" should be included after "ds/LifoAlloc.h"
 178
 179 js/src/tests/style/BadIncludesOrder-inl.h:9: error:
 180     "jsapi.h" is deprecated: Prefer including headers from js/public.
 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:20: error:
 192     "jsapi.h" is deprecated: Prefer including headers from js/public.
 193
 194 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 195     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 196
 197 (multiple files): error:
 198     header files form one or more cycles
 199
 200    tests/style/HeaderCycleA1.h
 201    -> tests/style/HeaderCycleA2.h
 202       -> tests/style/HeaderCycleA3.h
 203          -> tests/style/HeaderCycleA1.h
 204
 205    tests/style/HeaderCycleB1-inl.h
 206    -> tests/style/HeaderCycleB2-inl.h
 207       -> tests/style/HeaderCycleB3-inl.h
 208          -> tests/style/HeaderCycleB4-inl.h
 209             -> tests/style/HeaderCycleB1-inl.h
 210             -> tests/style/jsheadercycleB5inlines.h
 211                -> tests/style/HeaderCycleB1-inl.h
 212       -> tests/style/HeaderCycleB4-inl.h
 213
 214 """.splitlines(
 215     True
 216 )
 217
 218 actual_output = []
 219
 220
 221 def out(*lines):
 222     for line in lines:
 223         actual_output.append(line + "\n")
 224
 225
 226 def error(filename, linenum, *lines):
 227     location = filename
 228     if linenum is not None:
 229         location += ":" + str(linenum)
 230     out(location + ": error:")
 231     for line in lines:
 232         out("    " + line)
 233     out("")
 234
 235
 236 class FileKind(object):
 237     C = 1
 238     CPP = 2
 239     INL_H = 3
 240     H = 4
 241     TBL = 5
 242     MSG = 6
 243
 244     @staticmethod
 245     def get(filename):
 246         if filename.endswith(".c"):
 247             return FileKind.C
 248
 249         if filename.endswith(".cpp"):
 250             return FileKind.CPP
 251
 252         if filename.endswith(("inlines.h", "-inl.h")):
 253             return FileKind.INL_H
 254
 255         if filename.endswith(".h"):
 256             return FileKind.H
 257
 258         if filename.endswith(".tbl"):
 259             return FileKind.TBL
 260
 261         if filename.endswith(".msg"):
 262             return FileKind.MSG
 263
 264         error(filename, None, "unknown file kind")
 265
 266
 267 def check_style(enable_fixup):
 268     # We deal with two kinds of name.
 269     # - A "filename" is a full path to a file from the repository root.
 270     # - An "inclname" is how a file is referred to in a #include statement.
 271     #
 272     # Examples (filename -> inclname)
 273     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 274     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 275     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 276     # - "js/public/Vector.h"        -> "js/Vector.h"
 277     # - "js/src/vm/String.h"        -> "vm/String.h"
 278
 279     non_js_dirnames = (
 280         "mfbt/",
 281         "memory/mozalloc/",
 282         "mozglue/",
 283         "intl/components/",
 284     )  # type: tuple(str)
 285     non_js_inclnames = set()  # type: set(inclname)
 286     js_names = dict()  # type: dict(filename, inclname)
 287
 288     # Process files in js/src.
 289     js_src_root = os.path.join("js", "src")
 290     for dirpath, dirnames, filenames in os.walk(js_src_root):
 291         if dirpath == js_src_root:
 292             # Skip any subdirectories that contain a config.status file
 293             # (likely objdirs).
 294             builddirs = []
 295             for dirname in dirnames:
 296                 path = os.path.join(dirpath, dirname, "config.status")
 297                 if os.path.isfile(path):
 298                     builddirs.append(dirname)
 299             for dirname in builddirs:
 300                 dirnames.remove(dirname)
 301         for filename in filenames:
 302             filepath = os.path.join(dirpath, filename).replace("\\", "/")
 303             if not filepath.startswith(
 304                 tuple(ignored_js_src_dirs)
 305             ) and filepath.endswith((".c", ".cpp", ".h", ".tbl", ".msg")):
 306                 inclname = filepath[len("js/src/") :]
 307                 js_names[filepath] = inclname
 308
 309     # Look for header files in directories in non_js_dirnames.
 310     for non_js_dir in non_js_dirnames:
 311         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 312             for filename in filenames:
 313                 if filename.endswith(".h"):
 314                     inclname = "mozilla/" + filename
 315                     if non_js_dir == "intl/components/":
 316                         inclname = "mozilla/intl/" + filename
 317                     non_js_inclnames.add(inclname)
 318
 319     # Look for header files in js/public.
 320     js_public_root = os.path.join("js", "public")
 321     for dirpath, dirnames, filenames in os.walk(js_public_root):
 322         for filename in filenames:
 323             if filename.endswith((".h", ".msg")):
 324                 filepath = os.path.join(dirpath, filename).replace("\\", "/")
 325                 inclname = "js/" + filepath[len("js/public/") :]
 326                 js_names[filepath] = inclname
 327
 328     all_inclnames = non_js_inclnames | set(js_names.values())
 329
 330     edges = dict()  # type: dict(inclname, set(inclname))
 331
 332     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 333     # are #included from JS files we have to add them to the inclusion graph.
 334     for inclname in non_js_inclnames:
 335         edges[inclname] = set()
 336
 337     # Process all the JS files.
 338     for filename in sorted(js_names.keys()):
 339         inclname = js_names[filename]
 340         file_kind = FileKind.get(filename)
 341         if (
 342             file_kind == FileKind.C
 343             or file_kind == FileKind.CPP
 344             or file_kind == FileKind.H
 345             or file_kind == FileKind.INL_H
 346         ):
 347             included_h_inclnames = set()  # type: set(inclname)
 348
 349             with open(filename, encoding="utf-8") as f:
 350                 code = read_file(f)
 351
 352             if enable_fixup:
 353                 code = code.sorted(inclname)
 354                 with open(filename, "w") as f:
 355                     f.write(code.to_source())
 356
 357             check_file(
 358                 filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 359             )
 360
 361         edges[inclname] = included_h_inclnames
 362
 363     find_cycles(all_inclnames, edges)
 364
 365     # Compare expected and actual output.
 366     difflines = difflib.unified_diff(
 367         expected_output,
 368         actual_output,
 369         fromfile="check_spidermonkey_style.py expected output",
 370         tofile="check_spidermonkey_style.py actual output",
 371     )
 372     ok = True
 373     for diffline in difflines:
 374         ok = False
 375         print(diffline, end="")
 376
 377     return ok
 378
 379
 380 def module_name(name):
 381     """Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename."""
 382
 383     return (
 384         name.replace("inlines.h", "")
 385         .replace("-inl.h", "")
 386         .replace(".h", "")
 387         .replace(".cpp", "")
 388     )  # NOQA: E501
 389
 390
 391 def is_module_header(enclosing_inclname, header_inclname):
 392     """Determine if an included name is the "module header", i.e. should be
 393     first in the file."""
 394
 395     module = module_name(enclosing_inclname)
 396
 397     # Normal case, for example:
 398     #   module == "vm/Runtime", header_inclname == "vm/Runtime.h".
 399     if module == module_name(header_inclname):
 400         return True
 401
 402     # A public header, for example:
 403     #
 404     #   module == "vm/CharacterEncoding",
 405     #   header_inclname == "js/CharacterEncoding.h"
 406     #
 407     # or (for implementation files for js/public/*/*.h headers)
 408     #
 409     #   module == "vm/SourceHook",
 410     #   header_inclname == "js/experimental/SourceHook.h"
 411     m = re.match(r"js\/.*?([^\/]+)\.h", header_inclname)
 412     if m is not None and module.endswith("/" + m.group(1)):
 413         return True
 414
 415     return False
 416
 417
 418 class Include(object):
 419     """Important information for a single #include statement."""
 420
 421     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 422         self.include_prefix = include_prefix
 423         self.line_suffix = line_suffix
 424         self.inclname = inclname
 425         self.linenum = linenum
 426         self.is_system = is_system
 427
 428     def is_style_relevant(self):
 429         # Includes are style-relevant; that is, they're checked by the pairwise
 430         # style-checking algorithm in check_file.
 431         return True
 432
 433     def section(self, enclosing_inclname):
 434         """Identify which section inclname belongs to.
 435
 436         The section numbers are as follows.
 437           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 438           1. mozilla/Foo.h
 439           2. <foo.h> or <foo>
 440           3. jsfoo.h, prmjtime.h, etc
 441           4. foo/Bar.h
 442           5. jsfooinlines.h
 443           6. foo/Bar-inl.h
 444           7. non-.h, e.g. *.tbl, *.msg (these can be scattered throughout files)
 445         """
 446
 447         if self.is_system:
 448             return 2
 449
 450         if not self.inclname.endswith(".h"):
 451             return 7
 452
 453         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 454         # special handling.
 455         if is_module_header(enclosing_inclname, self.inclname):
 456             return 0
 457
 458         if "/" in self.inclname:
 459             if self.inclname.startswith("mozilla/"):
 460                 return 1
 461
 462             if self.inclname.endswith("-inl.h"):
 463                 return 6
 464
 465             return 4
 466
 467         if self.inclname.endswith("inlines.h"):
 468             return 5
 469
 470         return 3
 471
 472     def quote(self):
 473         if self.is_system:
 474             return "<" + self.inclname + ">"
 475         else:
 476             return '"' + self.inclname + '"'
 477
 478     def sort_key(self, enclosing_inclname):
 479         return (self.section(enclosing_inclname), self.inclname.lower())
 480
 481     def to_source(self):
 482         return self.include_prefix + self.quote() + self.line_suffix + "\n"
 483
 484
 485 class CppBlock(object):
 486     """C preprocessor block: a whole file or a single #if/#elif/#else block.
 487
 488     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 489     The top-level block, which is not within a #if/#endif pair, is also
 490     considered a block.
 491
 492     Each kid is either an Include (representing a #include), OrdinaryCode, or
 493     a nested CppBlock."""
 494
 495     def __init__(self, start_line=""):
 496         self.start = start_line
 497         self.end = ""
 498         self.kids = []
 499
 500     def is_style_relevant(self):
 501         return True
 502
 503     def append_ordinary_line(self, line):
 504         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 505             self.kids.append(OrdinaryCode())
 506         self.kids[-1].lines.append(line)
 507
 508     def style_relevant_kids(self):
 509         """Return a list of kids in this block that are style-relevant."""
 510         return [kid for kid in self.kids if kid.is_style_relevant()]
 511
 512     def sorted(self, enclosing_inclname):
 513         """Return a hopefully-sorted copy of this block. Implements --fixup.
 514
 515         When in doubt, this leaves the code unchanged.
 516         """
 517
 518         def pretty_sorted_includes(includes):
 519             """Return a new list containing the given includes, in order,
 520             with blank lines separating sections."""
 521             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 522             if sorted(keys) == keys:
 523                 return includes  # if nothing is out of order, don't touch anything
 524
 525             output = []
 526             current_section = None
 527             for (section, _), inc in sorted(zip(keys, includes)):
 528                 if current_section is not None and section != current_section:
 529                     output.append(OrdinaryCode(["\n"]))  # blank line
 530                 output.append(inc)
 531                 current_section = section
 532             return output
 533
 534         def should_try_to_sort(includes):
 535             if "tests/style/BadIncludes" in enclosing_inclname:
 536                 return False  # don't straighten the counterexample
 537             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 538                 return False  # don't sort batches containing odd includes
 539             if includes == sorted(
 540                 includes, key=lambda inc: inc.sort_key(enclosing_inclname)
 541             ):
 542                 return False  # it's already sorted, avoid whitespace-only fixups
 543             return True
 544
 545         # The content of the eventual output of this method.
 546         output = []
 547
 548         # The current batch of includes to sort. This list only ever contains Include objects
 549         # and whitespace OrdinaryCode objects.
 550         batch = []
 551
 552         def flush_batch():
 553             """Sort the contents of `batch` and move it to `output`."""
 554
 555             assert all(
 556                 isinstance(item, Include)
 557                 or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 558                 for item in batch
 559             )
 560
 561             # Here we throw away the blank lines.
 562             # `pretty_sorted_includes` puts them back.
 563             includes = []
 564             last_include_index = -1
 565             for i, item in enumerate(batch):
 566                 if isinstance(item, Include):
 567                     includes.append(item)
 568                     last_include_index = i
 569             cutoff = last_include_index + 1
 570
 571             if should_try_to_sort(includes):
 572                 output.extend(pretty_sorted_includes(includes) + batch[cutoff:])
 573             else:
 574                 output.extend(batch)
 575             del batch[:]
 576
 577         for kid in self.kids:
 578             if isinstance(kid, CppBlock):
 579                 flush_batch()
 580                 output.append(kid.sorted(enclosing_inclname))
 581             elif isinstance(kid, Include):
 582                 batch.append(kid)
 583             else:
 584                 assert isinstance(kid, OrdinaryCode)
 585                 if kid.to_source().isspace():
 586                     batch.append(kid)
 587                 else:
 588                     flush_batch()
 589                     output.append(kid)
 590         flush_batch()
 591
 592         result = CppBlock()
 593         result.start = self.start
 594         result.end = self.end
 595         result.kids = output
 596         return result
 597
 598     def to_source(self):
 599         return self.start + "".join(kid.to_source() for kid in self.kids) + self.end
 600
 601
 602 class OrdinaryCode(object):
 603     """A list of lines of code that aren't #include/#if/#else/#endif lines."""
 604
 605     def __init__(self, lines=None):
 606         self.lines = lines if lines is not None else []
 607
 608     def is_style_relevant(self):
 609         return False
 610
 611     def to_source(self):
 612         return "".join(self.lines)
 613
 614
 615 # A "snippet" is one of:
 616 #
 617 # *   Include - representing an #include line
 618 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 619 # *   OrdinaryCode - representing lines of non-#include-relevant code
 620
 621
 622 def read_file(f):
 623     block_stack = [CppBlock()]
 624
 625     # Extract the #include statements as a tree of snippets.
 626     for linenum, line in enumerate(f, start=1):
 627         if line.lstrip().startswith("#"):
 628             # Look for a |#include "..."| line.
 629             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 630             if m is not None:
 631                 prefix, inclname, suffix = m.groups()
 632                 block_stack[-1].kids.append(
 633                     Include(prefix, inclname, suffix, linenum, is_system=False)
 634                 )
 635                 continue
 636
 637             # Look for a |#include <...>| line.
 638             m = re.match(r"(\s*#\s*include\s+)<([^>]*)>(.*)", line)
 639             if m is not None:
 640                 prefix, inclname, suffix = m.groups()
 641                 block_stack[-1].kids.append(
 642                     Include(prefix, inclname, suffix, linenum, is_system=True)
 643                 )
 644                 continue
 645
 646             # Look for a |#{if,ifdef,ifndef}| line.
 647             m = re.match(r"\s*#\s*(if|ifdef|ifndef)\b", line)
 648             if m is not None:
 649                 # Open a new block.
 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 |#{elif,else}| line.
 656             m = re.match(r"\s*#\s*(elif|else)\b", line)
 657             if m is not None:
 658                 # Close the current block, and open an adjacent one.
 659                 block_stack.pop()
 660                 new_block = CppBlock(line)
 661                 block_stack[-1].kids.append(new_block)
 662                 block_stack.append(new_block)
 663                 continue
 664
 665             # Look for a |#endif| line.
 666             m = re.match(r"\s*#\s*endif\b", line)
 667             if m is not None:
 668                 # Close the current block.
 669                 block_stack.pop().end = line
 670                 if len(block_stack) == 0:
 671                     raise ValueError("#endif without #if at line " + str(linenum))
 672                 continue
 673
 674         # Otherwise, we have an ordinary line.
 675         block_stack[-1].append_ordinary_line(line)
 676
 677     if len(block_stack) > 1:
 678         raise ValueError("unmatched #if")
 679     return block_stack[-1]
 680
 681
 682 def check_file(
 683     filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 684 ):
 685     def check_include_statement(include):
 686         """Check the style of a single #include statement."""
 687
 688         if include.is_system:
 689             # Check it is not a known local file (in which case it's probably a system header).
 690             if (
 691                 include.inclname in included_inclnames_to_ignore
 692                 or include.inclname in all_inclnames
 693             ):
 694                 error(
 695                     filename,
 696                     include.linenum,
 697                     include.quote() + " should be included using",
 698                     'the #include "..." form',
 699                 )
 700
 701         else:
 702             msg = deprecated_inclnames.get(include.inclname)
 703             if msg:
 704                 error(
 705                     filename,
 706                     include.linenum,
 707                     include.quote() + " is deprecated: " + msg,
 708                 )
 709
 710             if file_kind == FileKind.H or file_kind == FileKind.INL_H:
 711                 msg = deprecated_inclnames_in_header.get(include.inclname)
 712                 if msg and filename not in deprecated_inclnames_in_header_excludes:
 713                     error(
 714                         filename,
 715                         include.linenum,
 716                         include.quote() + " is deprecated: " + msg,
 717                     )
 718
 719             if include.inclname not in included_inclnames_to_ignore:
 720                 included_kind = FileKind.get(include.inclname)
 721
 722                 # Check the #include path has the correct form.
 723                 if include.inclname not in all_inclnames:
 724                     error(
 725                         filename,
 726                         include.linenum,
 727                         include.quote() + " is included using the wrong path;",
 728                         "did you forget a prefix, or is the file not yet committed?",
 729                     )
 730
 731                 # Record inclusions of .h files for cycle detection later.
 732                 # (Exclude .tbl and .msg files.)
 733                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 734                     included_h_inclnames.add(include.inclname)
 735
 736                 # Check a H file doesn't #include an INL_H file.
 737                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 738                     error(
 739                         filename,
 740                         include.linenum,
 741                         "vanilla header includes an inline-header file "
 742                         + include.quote(),
 743                     )
 744
 745                 # Check a file doesn't #include itself.  (We do this here because the cycle
 746                 # detection below doesn't detect this case.)
 747                 if inclname == include.inclname:
 748                     error(filename, include.linenum, "the file includes itself")
 749
 750     def check_includes_order(include1, include2):
 751         """Check the ordering of two #include statements."""
 752
 753         if (
 754             include1.inclname in oddly_ordered_inclnames
 755             or include2.inclname in oddly_ordered_inclnames
 756         ):
 757             return
 758
 759         section1 = include1.section(inclname)
 760         section2 = include2.section(inclname)
 761         if (section1 > section2) or (
 762             (section1 == section2)
 763             and (include1.inclname.lower() > include2.inclname.lower())
 764         ):
 765             error(
 766                 filename,
 767                 str(include1.linenum) + ":" + str(include2.linenum),
 768                 include1.quote() + " should be included after " + include2.quote(),
 769             )
 770
 771     # Check the extracted #include statements, both individually, and the ordering of
 772     # adjacent pairs that live in the same block.
 773     def pair_traverse(prev, this):
 774         if isinstance(this, Include):
 775             check_include_statement(this)
 776             if isinstance(prev, Include):
 777                 check_includes_order(prev, this)
 778         else:
 779             kids = this.style_relevant_kids()
 780             for prev2, this2 in zip([None] + kids[0:-1], kids):
 781                 pair_traverse(prev2, this2)
 782
 783     pair_traverse(None, code)
 784
 785
 786 def find_cycles(all_inclnames, edges):
 787     """Find and draw any cycles."""
 788
 789     SCCs = tarjan(all_inclnames, edges)
 790
 791     # The various sorted() calls below ensure the output is deterministic.
 792
 793     def draw_SCC(c):
 794         cset = set(c)
 795         drawn = set()
 796
 797         def draw(v, indent):
 798             out("   " * indent + ("-> " if indent else "   ") + v)
 799             if v in drawn:
 800                 return
 801             drawn.add(v)
 802             for succ in sorted(edges[v]):
 803                 if succ in cset:
 804                     draw(succ, indent + 1)
 805
 806         draw(sorted(c)[0], 0)
 807         out("")
 808
 809     have_drawn_an_SCC = False
 810     for scc in sorted(SCCs):
 811         if len(scc) != 1:
 812             if not have_drawn_an_SCC:
 813                 error("(multiple files)", None, "header files form one or more cycles")
 814                 have_drawn_an_SCC = True
 815
 816             draw_SCC(scc)
 817
 818
 819 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 820 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 821 def tarjan(V, E):
 822     vertex_index = {}
 823     vertex_lowlink = {}
 824     index = 0
 825     S = []
 826     all_SCCs = []
 827
 828     def strongconnect(v, index):
 829         # Set the depth index for v to the smallest unused index
 830         vertex_index[v] = index
 831         vertex_lowlink[v] = index
 832         index += 1
 833         S.append(v)
 834
 835         # Consider successors of v
 836         for w in E[v]:
 837             if w not in vertex_index:
 838                 # Successor w has not yet been visited; recurse on it
 839                 index = strongconnect(w, index)
 840                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 841             elif w in S:
 842                 # Successor w is in stack S and hence in the current SCC
 843                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 844
 845         # If v is a root node, pop the stack and generate an SCC
 846         if vertex_lowlink[v] == vertex_index[v]:
 847             i = S.index(v)
 848             scc = S[i:]
 849             del S[i:]
 850             all_SCCs.append(scc)
 851
 852         return index
 853
 854     for v in V:
 855         if v not in vertex_index:
 856             index = strongconnect(v, index)
 857
 858     return all_SCCs
 859
 860
 861 def main():
 862     if sys.argv[1:] == ["--fixup"]:
 863         # Sort #include directives in-place.  Fixup mode doesn't solve
 864         # all possible silliness that the script checks for; it's just a
 865         # hack for the common case where renaming a header causes style
 866         # errors.
 867         fixup = True
 868     elif sys.argv[1:] == []:
 869         fixup = False
 870     else:
 871         print(
 872             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 873             "line options: " + repr(sys.argv[1:])
 874         )
 875         sys.exit(1)
 876
 877     ok = check_style(fixup)
 878
 879     if ok:
 880         print("TEST-PASS | check_spidermonkey_style.py | ok")
 881     else:
 882         print(
 883             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 884             + "actual output does not match expected output;  diff is above."
 885         )
 886         print(
 887             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 888             + "Hint: If the problem is that you renamed a header, and many #includes "
 889             + "are no longer in alphabetical order, commit your work and then try "
 890             + "`check_spidermonkey_style.py --fixup`. "
 891             + "You need to commit first because --fixup modifies your files in place."
 892         )
 893
 894     sys.exit(0 if ok else 1)
 895
 896
 897 if __name__ == "__main__":
 898     main()