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/AtomicOperationsGenerated.h",  # generated in $OBJDIR
  67         "jit/CacheIROpsGenerated.h",  # generated in $OBJDIR
  68         "jit/LIROpsGenerated.h",  # generated in $OBJDIR
  69         "jit/MIROpsGenerated.h",  # generated in $OBJDIR
  70         "js/ProfilingCategoryList.h",  # comes from mozglue/baseprofiler
  71         "jscustomallocator.h",  # provided by embedders;  allowed to be missing
  72         "js-config.h",  # generated in $OBJDIR
  73         "fdlibm.h",  # fdlibm
  74         "FuzzerDefs.h",  # included without a path
  75         "FuzzingInterface.h",  # included without a path
  76         "mozmemory.h",  # included without a path
  77         "pratom.h",  # NSPR
  78         "prcvar.h",  # NSPR
  79         "prerror.h",  # NSPR
  80         "prinit.h",  # NSPR
  81         "prio.h",  # NSPR
  82         "private/pprio.h",  # NSPR
  83         "prlink.h",  # NSPR
  84         "prlock.h",  # NSPR
  85         "prprf.h",  # NSPR
  86         "prthread.h",  # NSPR
  87         "prtypes.h",  # NSPR
  88         "selfhosted.out.h",  # generated in $OBJDIR
  89         "shellmoduleloader.out.h",  # generated in $OBJDIR
  90         "unicode/locid.h",  # ICU
  91         "unicode/uchar.h",  # ICU
  92         "unicode/uniset.h",  # ICU
  93         "unicode/unistr.h",  # ICU
  94         "unicode/utypes.h",  # ICU
  95         "vtune/VTuneWrapper.h",  # VTune
  96         "wasm/WasmIntrinsicGenerated.h",  # generated in $OBJDIR"
  97         "zydis/ZydisAPI.h",  # Zydis
  98     ]
  99 )
 100
 101 deprecated_inclnames = {
 102     "mozilla/Unused.h": "Use [[nodiscard]] and (void)expr casts instead.",
 103 }
 104
 105 # JSAPI functions should be included through headers from js/public instead of
 106 # using the old, catch-all jsapi.h file.
 107 deprecated_inclnames_in_header = {
 108     "jsapi.h": "Prefer including headers from js/public.",
 109 }
 110
 111 # Temporary exclusions for files which still need to include jsapi.h.
 112 deprecated_inclnames_in_header_excludes = {
 113     "js/src/vm/Compartment-inl.h",  # for JS::InformalValueTypeName
 114     "js/src/jsapi-tests/tests.h",  # for JS_ValueToSource
 115 }
 116
 117 # These files have additional constraints on where they are #included, so we
 118 # ignore #includes of them when checking #include ordering.
 119 oddly_ordered_inclnames = set(
 120     [
 121         "ctypes/typedefs.h",  # Included multiple times in the body of ctypes/CTypes.h
 122         # Included in the body of frontend/TokenStream.h
 123         "frontend/ReservedWordsGenerated.h",
 124         "gc/StatsPhasesGenerated.h",  # Included in the body of gc/Statistics.h
 125         "gc/StatsPhasesGenerated.inc",  # Included in the body of gc/Statistics.cpp
 126         "psapi.h",  # Must be included after "util/WindowsWrapper.h" on Windows
 127         "machine/endian.h",  # Must be included after <sys/types.h> on BSD
 128         "winbase.h",  # Must precede other system headers(?)
 129         "windef.h",  # Must precede other system headers(?)
 130     ]
 131 )
 132
 133 # The files in tests/style/ contain code that fails this checking in various
 134 # ways.  Here is the output we expect.  If the actual output differs from
 135 # this, one of the following must have happened.
 136 # - New SpiderMonkey code violates one of the checked rules.
 137 # - The tests/style/ files have changed without expected_output being changed
 138 #   accordingly.
 139 # - This script has been broken somehow.
 140 #
 141 expected_output = """\
 142 js/src/tests/style/BadIncludes.h:3: error:
 143     the file includes itself
 144
 145 js/src/tests/style/BadIncludes.h:6: error:
 146     "BadIncludes2.h" is included using the wrong path;
 147     did you forget a prefix, or is the file not yet committed?
 148
 149 js/src/tests/style/BadIncludes.h:8: error:
 150     <tests/style/BadIncludes2.h> should be included using
 151     the #include "..." form
 152
 153 js/src/tests/style/BadIncludes.h:10: error:
 154     "stdio.h" is included using the wrong path;
 155     did you forget a prefix, or is the file not yet committed?
 156
 157 js/src/tests/style/BadIncludes.h:12: error:
 158     "mozilla/Unused.h" is deprecated: Use [[nodiscard]] and (void)expr casts instead.
 159
 160 js/src/tests/style/BadIncludes2.h:1: error:
 161     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 162
 163 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 164     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 165
 166 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 167     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 168
 169 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 170     "js/Value.h" should be included after "ds/LifoAlloc.h"
 171
 172 js/src/tests/style/BadIncludesOrder-inl.h:9: error:
 173     "jsapi.h" is deprecated: Prefer including headers from js/public.
 174
 175 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 176     "ds/LifoAlloc.h" should be included after "jsapi.h"
 177
 178 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 179     "jsapi.h" should be included after <stdio.h>
 180
 181 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 182     <stdio.h> should be included after "mozilla/HashFunctions.h"
 183
 184 js/src/tests/style/BadIncludesOrder-inl.h:20: error:
 185     "jsapi.h" is deprecated: Prefer including headers from js/public.
 186
 187 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 188     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 189
 190 (multiple files): error:
 191     header files form one or more cycles
 192
 193    tests/style/HeaderCycleA1.h
 194    -> tests/style/HeaderCycleA2.h
 195       -> tests/style/HeaderCycleA3.h
 196          -> tests/style/HeaderCycleA1.h
 197
 198    tests/style/HeaderCycleB1-inl.h
 199    -> tests/style/HeaderCycleB2-inl.h
 200       -> tests/style/HeaderCycleB3-inl.h
 201          -> tests/style/HeaderCycleB4-inl.h
 202             -> tests/style/HeaderCycleB1-inl.h
 203             -> tests/style/jsheadercycleB5inlines.h
 204                -> tests/style/HeaderCycleB1-inl.h
 205       -> tests/style/HeaderCycleB4-inl.h
 206
 207 """.splitlines(
 208     True
 209 )
 210
 211 actual_output = []
 212
 213
 214 def out(*lines):
 215     for line in lines:
 216         actual_output.append(line + "\n")
 217
 218
 219 def error(filename, linenum, *lines):
 220     location = filename
 221     if linenum is not None:
 222         location += ":" + str(linenum)
 223     out(location + ": error:")
 224     for line in lines:
 225         out("    " + line)
 226     out("")
 227
 228
 229 class FileKind(object):
 230     C = 1
 231     CPP = 2
 232     INL_H = 3
 233     H = 4
 234     TBL = 5
 235     MSG = 6
 236
 237     @staticmethod
 238     def get(filename):
 239         if filename.endswith(".c"):
 240             return FileKind.C
 241
 242         if filename.endswith(".cpp"):
 243             return FileKind.CPP
 244
 245         if filename.endswith(("inlines.h", "-inl.h")):
 246             return FileKind.INL_H
 247
 248         if filename.endswith(".h"):
 249             return FileKind.H
 250
 251         if filename.endswith(".tbl"):
 252             return FileKind.TBL
 253
 254         if filename.endswith(".msg"):
 255             return FileKind.MSG
 256
 257         error(filename, None, "unknown file kind")
 258
 259
 260 def check_style(enable_fixup):
 261     # We deal with two kinds of name.
 262     # - A "filename" is a full path to a file from the repository root.
 263     # - An "inclname" is how a file is referred to in a #include statement.
 264     #
 265     # Examples (filename -> inclname)
 266     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 267     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 268     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 269     # - "js/public/Vector.h"        -> "js/Vector.h"
 270     # - "js/src/vm/String.h"        -> "vm/String.h"
 271
 272     non_js_dirnames = (
 273         "mfbt/",
 274         "memory/mozalloc/",
 275         "mozglue/",
 276         "intl/components/",
 277     )  # type: tuple(str)
 278     non_js_inclnames = set()  # type: set(inclname)
 279     js_names = dict()  # type: dict(filename, inclname)
 280
 281     # Process files in js/src.
 282     js_src_root = os.path.join("js", "src")
 283     for dirpath, dirnames, filenames in os.walk(js_src_root):
 284         if dirpath == js_src_root:
 285             # Skip any subdirectories that contain a config.status file
 286             # (likely objdirs).
 287             builddirs = []
 288             for dirname in dirnames:
 289                 path = os.path.join(dirpath, dirname, "config.status")
 290                 if os.path.isfile(path):
 291                     builddirs.append(dirname)
 292             for dirname in builddirs:
 293                 dirnames.remove(dirname)
 294         for filename in filenames:
 295             filepath = os.path.join(dirpath, filename).replace("\\", "/")
 296             if not filepath.startswith(
 297                 tuple(ignored_js_src_dirs)
 298             ) and filepath.endswith((".c", ".cpp", ".h", ".tbl", ".msg")):
 299                 inclname = filepath[len("js/src/") :]
 300                 js_names[filepath] = inclname
 301
 302     # Look for header files in directories in non_js_dirnames.
 303     for non_js_dir in non_js_dirnames:
 304         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 305             for filename in filenames:
 306                 if filename.endswith(".h"):
 307                     inclname = "mozilla/" + filename
 308                     if non_js_dir == "intl/components/":
 309                         inclname = "mozilla/intl/" + filename
 310                     non_js_inclnames.add(inclname)
 311
 312     # Look for header files in js/public.
 313     js_public_root = os.path.join("js", "public")
 314     for dirpath, dirnames, filenames in os.walk(js_public_root):
 315         for filename in filenames:
 316             if filename.endswith((".h", ".msg")):
 317                 filepath = os.path.join(dirpath, filename).replace("\\", "/")
 318                 inclname = "js/" + filepath[len("js/public/") :]
 319                 js_names[filepath] = inclname
 320
 321     all_inclnames = non_js_inclnames | set(js_names.values())
 322
 323     edges = dict()  # type: dict(inclname, set(inclname))
 324
 325     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 326     # are #included from JS files we have to add them to the inclusion graph.
 327     for inclname in non_js_inclnames:
 328         edges[inclname] = set()
 329
 330     # Process all the JS files.
 331     for filename in sorted(js_names.keys()):
 332         inclname = js_names[filename]
 333         file_kind = FileKind.get(filename)
 334         if (
 335             file_kind == FileKind.C
 336             or file_kind == FileKind.CPP
 337             or file_kind == FileKind.H
 338             or file_kind == FileKind.INL_H
 339         ):
 340             included_h_inclnames = set()  # type: set(inclname)
 341
 342             with open(filename, encoding="utf-8") as f:
 343                 code = read_file(f)
 344
 345             if enable_fixup:
 346                 code = code.sorted(inclname)
 347                 with open(filename, "w") as f:
 348                     f.write(code.to_source())
 349
 350             check_file(
 351                 filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 352             )
 353
 354         edges[inclname] = included_h_inclnames
 355
 356     find_cycles(all_inclnames, edges)
 357
 358     # Compare expected and actual output.
 359     difflines = difflib.unified_diff(
 360         expected_output,
 361         actual_output,
 362         fromfile="check_spidermonkey_style.py expected output",
 363         tofile="check_spidermonkey_style.py actual output",
 364     )
 365     ok = True
 366     for diffline in difflines:
 367         ok = False
 368         print(diffline, end="")
 369
 370     return ok
 371
 372
 373 def module_name(name):
 374     """Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename."""
 375
 376     return (
 377         name.replace("inlines.h", "")
 378         .replace("-inl.h", "")
 379         .replace(".h", "")
 380         .replace(".cpp", "")
 381     )  # NOQA: E501
 382
 383
 384 def is_module_header(enclosing_inclname, header_inclname):
 385     """Determine if an included name is the "module header", i.e. should be
 386     first in the file."""
 387
 388     module = module_name(enclosing_inclname)
 389
 390     # Normal case, for example:
 391     #   module == "vm/Runtime", header_inclname == "vm/Runtime.h".
 392     if module == module_name(header_inclname):
 393         return True
 394
 395     # A public header, for example:
 396     #
 397     #   module == "vm/CharacterEncoding",
 398     #   header_inclname == "js/CharacterEncoding.h"
 399     #
 400     # or (for implementation files for js/public/*/*.h headers)
 401     #
 402     #   module == "vm/SourceHook",
 403     #   header_inclname == "js/experimental/SourceHook.h"
 404     m = re.match(r"js\/.*?([^\/]+)\.h", header_inclname)
 405     if m is not None and module.endswith("/" + m.group(1)):
 406         return True
 407
 408     return False
 409
 410
 411 class Include(object):
 412     """Important information for a single #include statement."""
 413
 414     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 415         self.include_prefix = include_prefix
 416         self.line_suffix = line_suffix
 417         self.inclname = inclname
 418         self.linenum = linenum
 419         self.is_system = is_system
 420
 421     def is_style_relevant(self):
 422         # Includes are style-relevant; that is, they're checked by the pairwise
 423         # style-checking algorithm in check_file.
 424         return True
 425
 426     def section(self, enclosing_inclname):
 427         """Identify which section inclname belongs to.
 428
 429         The section numbers are as follows.
 430           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 431           1. mozilla/Foo.h
 432           2. <foo.h> or <foo>
 433           3. jsfoo.h, prmjtime.h, etc
 434           4. foo/Bar.h
 435           5. jsfooinlines.h
 436           6. foo/Bar-inl.h
 437           7. non-.h, e.g. *.tbl, *.msg (these can be scattered throughout files)
 438         """
 439
 440         if self.is_system:
 441             return 2
 442
 443         if not self.inclname.endswith(".h"):
 444             return 7
 445
 446         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 447         # special handling.
 448         if is_module_header(enclosing_inclname, self.inclname):
 449             return 0
 450
 451         if "/" in self.inclname:
 452             if self.inclname.startswith("mozilla/"):
 453                 return 1
 454
 455             if self.inclname.endswith("-inl.h"):
 456                 return 6
 457
 458             return 4
 459
 460         if self.inclname.endswith("inlines.h"):
 461             return 5
 462
 463         return 3
 464
 465     def quote(self):
 466         if self.is_system:
 467             return "<" + self.inclname + ">"
 468         else:
 469             return '"' + self.inclname + '"'
 470
 471     def sort_key(self, enclosing_inclname):
 472         return (self.section(enclosing_inclname), self.inclname.lower())
 473
 474     def to_source(self):
 475         return self.include_prefix + self.quote() + self.line_suffix + "\n"
 476
 477
 478 class CppBlock(object):
 479     """C preprocessor block: a whole file or a single #if/#elif/#else block.
 480
 481     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 482     The top-level block, which is not within a #if/#endif pair, is also
 483     considered a block.
 484
 485     Each kid is either an Include (representing a #include), OrdinaryCode, or
 486     a nested CppBlock."""
 487
 488     def __init__(self, start_line=""):
 489         self.start = start_line
 490         self.end = ""
 491         self.kids = []
 492
 493     def is_style_relevant(self):
 494         return True
 495
 496     def append_ordinary_line(self, line):
 497         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 498             self.kids.append(OrdinaryCode())
 499         self.kids[-1].lines.append(line)
 500
 501     def style_relevant_kids(self):
 502         """Return a list of kids in this block that are style-relevant."""
 503         return [kid for kid in self.kids if kid.is_style_relevant()]
 504
 505     def sorted(self, enclosing_inclname):
 506         """Return a hopefully-sorted copy of this block. Implements --fixup.
 507
 508         When in doubt, this leaves the code unchanged.
 509         """
 510
 511         def pretty_sorted_includes(includes):
 512             """Return a new list containing the given includes, in order,
 513             with blank lines separating sections."""
 514             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 515             if sorted(keys) == keys:
 516                 return includes  # if nothing is out of order, don't touch anything
 517
 518             output = []
 519             current_section = None
 520             for (section, _), inc in sorted(zip(keys, includes)):
 521                 if current_section is not None and section != current_section:
 522                     output.append(OrdinaryCode(["\n"]))  # blank line
 523                 output.append(inc)
 524                 current_section = section
 525             return output
 526
 527         def should_try_to_sort(includes):
 528             if "tests/style/BadIncludes" in enclosing_inclname:
 529                 return False  # don't straighten the counterexample
 530             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 531                 return False  # don't sort batches containing odd includes
 532             if includes == sorted(
 533                 includes, key=lambda inc: inc.sort_key(enclosing_inclname)
 534             ):
 535                 return False  # it's already sorted, avoid whitespace-only fixups
 536             return True
 537
 538         # The content of the eventual output of this method.
 539         output = []
 540
 541         # The current batch of includes to sort. This list only ever contains Include objects
 542         # and whitespace OrdinaryCode objects.
 543         batch = []
 544
 545         def flush_batch():
 546             """Sort the contents of `batch` and move it to `output`."""
 547
 548             assert all(
 549                 isinstance(item, Include)
 550                 or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 551                 for item in batch
 552             )
 553
 554             # Here we throw away the blank lines.
 555             # `pretty_sorted_includes` puts them back.
 556             includes = []
 557             last_include_index = -1
 558             for i, item in enumerate(batch):
 559                 if isinstance(item, Include):
 560                     includes.append(item)
 561                     last_include_index = i
 562             cutoff = last_include_index + 1
 563
 564             if should_try_to_sort(includes):
 565                 output.extend(pretty_sorted_includes(includes) + batch[cutoff:])
 566             else:
 567                 output.extend(batch)
 568             del batch[:]
 569
 570         for kid in self.kids:
 571             if isinstance(kid, CppBlock):
 572                 flush_batch()
 573                 output.append(kid.sorted(enclosing_inclname))
 574             elif isinstance(kid, Include):
 575                 batch.append(kid)
 576             else:
 577                 assert isinstance(kid, OrdinaryCode)
 578                 if kid.to_source().isspace():
 579                     batch.append(kid)
 580                 else:
 581                     flush_batch()
 582                     output.append(kid)
 583         flush_batch()
 584
 585         result = CppBlock()
 586         result.start = self.start
 587         result.end = self.end
 588         result.kids = output
 589         return result
 590
 591     def to_source(self):
 592         return self.start + "".join(kid.to_source() for kid in self.kids) + self.end
 593
 594
 595 class OrdinaryCode(object):
 596     """A list of lines of code that aren't #include/#if/#else/#endif lines."""
 597
 598     def __init__(self, lines=None):
 599         self.lines = lines if lines is not None else []
 600
 601     def is_style_relevant(self):
 602         return False
 603
 604     def to_source(self):
 605         return "".join(self.lines)
 606
 607
 608 # A "snippet" is one of:
 609 #
 610 # *   Include - representing an #include line
 611 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 612 # *   OrdinaryCode - representing lines of non-#include-relevant code
 613
 614
 615 def read_file(f):
 616     block_stack = [CppBlock()]
 617
 618     # Extract the #include statements as a tree of snippets.
 619     for linenum, line in enumerate(f, start=1):
 620         if line.lstrip().startswith("#"):
 621             # Look for a |#include "..."| line.
 622             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 623             if m is not None:
 624                 prefix, inclname, suffix = m.groups()
 625                 block_stack[-1].kids.append(
 626                     Include(prefix, inclname, suffix, linenum, is_system=False)
 627                 )
 628                 continue
 629
 630             # Look for a |#include <...>| line.
 631             m = re.match(r"(\s*#\s*include\s+)<([^>]*)>(.*)", line)
 632             if m is not None:
 633                 prefix, inclname, suffix = m.groups()
 634                 block_stack[-1].kids.append(
 635                     Include(prefix, inclname, suffix, linenum, is_system=True)
 636                 )
 637                 continue
 638
 639             # Look for a |#{if,ifdef,ifndef}| line.
 640             m = re.match(r"\s*#\s*(if|ifdef|ifndef)\b", line)
 641             if m is not None:
 642                 # Open a new block.
 643                 new_block = CppBlock(line)
 644                 block_stack[-1].kids.append(new_block)
 645                 block_stack.append(new_block)
 646                 continue
 647
 648             # Look for a |#{elif,else}| line.
 649             m = re.match(r"\s*#\s*(elif|else)\b", line)
 650             if m is not None:
 651                 # Close the current block, and open an adjacent one.
 652                 block_stack.pop()
 653                 new_block = CppBlock(line)
 654                 block_stack[-1].kids.append(new_block)
 655                 block_stack.append(new_block)
 656                 continue
 657
 658             # Look for a |#endif| line.
 659             m = re.match(r"\s*#\s*endif\b", line)
 660             if m is not None:
 661                 # Close the current block.
 662                 block_stack.pop().end = line
 663                 if len(block_stack) == 0:
 664                     raise ValueError("#endif without #if at line " + str(linenum))
 665                 continue
 666
 667         # Otherwise, we have an ordinary line.
 668         block_stack[-1].append_ordinary_line(line)
 669
 670     if len(block_stack) > 1:
 671         raise ValueError("unmatched #if")
 672     return block_stack[-1]
 673
 674
 675 def check_file(
 676     filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 677 ):
 678     def check_include_statement(include):
 679         """Check the style of a single #include statement."""
 680
 681         if include.is_system:
 682             # Check it is not a known local file (in which case it's probably a system header).
 683             if (
 684                 include.inclname in included_inclnames_to_ignore
 685                 or include.inclname in all_inclnames
 686             ):
 687                 error(
 688                     filename,
 689                     include.linenum,
 690                     include.quote() + " should be included using",
 691                     'the #include "..." form',
 692                 )
 693
 694         else:
 695             msg = deprecated_inclnames.get(include.inclname)
 696             if msg:
 697                 error(
 698                     filename,
 699                     include.linenum,
 700                     include.quote() + " is deprecated: " + msg,
 701                 )
 702
 703             if file_kind == FileKind.H or file_kind == FileKind.INL_H:
 704                 msg = deprecated_inclnames_in_header.get(include.inclname)
 705                 if msg and filename not in deprecated_inclnames_in_header_excludes:
 706                     error(
 707                         filename,
 708                         include.linenum,
 709                         include.quote() + " is deprecated: " + msg,
 710                     )
 711
 712             if include.inclname not in included_inclnames_to_ignore:
 713                 included_kind = FileKind.get(include.inclname)
 714
 715                 # Check the #include path has the correct form.
 716                 if include.inclname not in all_inclnames:
 717                     error(
 718                         filename,
 719                         include.linenum,
 720                         include.quote() + " is included using the wrong path;",
 721                         "did you forget a prefix, or is the file not yet committed?",
 722                     )
 723
 724                 # Record inclusions of .h files for cycle detection later.
 725                 # (Exclude .tbl and .msg files.)
 726                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 727                     included_h_inclnames.add(include.inclname)
 728
 729                 # Check a H file doesn't #include an INL_H file.
 730                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 731                     error(
 732                         filename,
 733                         include.linenum,
 734                         "vanilla header includes an inline-header file "
 735                         + include.quote(),
 736                     )
 737
 738                 # Check a file doesn't #include itself.  (We do this here because the cycle
 739                 # detection below doesn't detect this case.)
 740                 if inclname == include.inclname:
 741                     error(filename, include.linenum, "the file includes itself")
 742
 743     def check_includes_order(include1, include2):
 744         """Check the ordering of two #include statements."""
 745
 746         if (
 747             include1.inclname in oddly_ordered_inclnames
 748             or include2.inclname in oddly_ordered_inclnames
 749         ):
 750             return
 751
 752         section1 = include1.section(inclname)
 753         section2 = include2.section(inclname)
 754         if (section1 > section2) or (
 755             (section1 == section2)
 756             and (include1.inclname.lower() > include2.inclname.lower())
 757         ):
 758             error(
 759                 filename,
 760                 str(include1.linenum) + ":" + str(include2.linenum),
 761                 include1.quote() + " should be included after " + include2.quote(),
 762             )
 763
 764     # Check the extracted #include statements, both individually, and the ordering of
 765     # adjacent pairs that live in the same block.
 766     def pair_traverse(prev, this):
 767         if isinstance(this, Include):
 768             check_include_statement(this)
 769             if isinstance(prev, Include):
 770                 check_includes_order(prev, this)
 771         else:
 772             kids = this.style_relevant_kids()
 773             for prev2, this2 in zip([None] + kids[0:-1], kids):
 774                 pair_traverse(prev2, this2)
 775
 776     pair_traverse(None, code)
 777
 778
 779 def find_cycles(all_inclnames, edges):
 780     """Find and draw any cycles."""
 781
 782     SCCs = tarjan(all_inclnames, edges)
 783
 784     # The various sorted() calls below ensure the output is deterministic.
 785
 786     def draw_SCC(c):
 787         cset = set(c)
 788         drawn = set()
 789
 790         def draw(v, indent):
 791             out("   " * indent + ("-> " if indent else "   ") + v)
 792             if v in drawn:
 793                 return
 794             drawn.add(v)
 795             for succ in sorted(edges[v]):
 796                 if succ in cset:
 797                     draw(succ, indent + 1)
 798
 799         draw(sorted(c)[0], 0)
 800         out("")
 801
 802     have_drawn_an_SCC = False
 803     for scc in sorted(SCCs):
 804         if len(scc) != 1:
 805             if not have_drawn_an_SCC:
 806                 error("(multiple files)", None, "header files form one or more cycles")
 807                 have_drawn_an_SCC = True
 808
 809             draw_SCC(scc)
 810
 811
 812 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 813 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 814 def tarjan(V, E):
 815     vertex_index = {}
 816     vertex_lowlink = {}
 817     index = 0
 818     S = []
 819     all_SCCs = []
 820
 821     def strongconnect(v, index):
 822         # Set the depth index for v to the smallest unused index
 823         vertex_index[v] = index
 824         vertex_lowlink[v] = index
 825         index += 1
 826         S.append(v)
 827
 828         # Consider successors of v
 829         for w in E[v]:
 830             if w not in vertex_index:
 831                 # Successor w has not yet been visited; recurse on it
 832                 index = strongconnect(w, index)
 833                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 834             elif w in S:
 835                 # Successor w is in stack S and hence in the current SCC
 836                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 837
 838         # If v is a root node, pop the stack and generate an SCC
 839         if vertex_lowlink[v] == vertex_index[v]:
 840             i = S.index(v)
 841             scc = S[i:]
 842             del S[i:]
 843             all_SCCs.append(scc)
 844
 845         return index
 846
 847     for v in V:
 848         if v not in vertex_index:
 849             index = strongconnect(v, index)
 850
 851     return all_SCCs
 852
 853
 854 def main():
 855     if sys.argv[1:] == ["--fixup"]:
 856         # Sort #include directives in-place.  Fixup mode doesn't solve
 857         # all possible silliness that the script checks for; it's just a
 858         # hack for the common case where renaming a header causes style
 859         # errors.
 860         fixup = True
 861     elif sys.argv[1:] == []:
 862         fixup = False
 863     else:
 864         print(
 865             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 866             "line options: " + repr(sys.argv[1:])
 867         )
 868         sys.exit(1)
 869
 870     ok = check_style(fixup)
 871
 872     if ok:
 873         print("TEST-PASS | check_spidermonkey_style.py | ok")
 874     else:
 875         print(
 876             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 877             + "actual output does not match expected output;  diff is above."
 878         )
 879         print(
 880             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 881             + "Hint: If the problem is that you renamed a header, and many #includes "
 882             + "are no longer in alphabetical order, commit your work and then try "
 883             + "`check_spidermonkey_style.py --fixup`. "
 884             + "You need to commit first because --fixup modifies your files in place."
 885         )
 886
 887     sys.exit(0 if ok else 1)
 888
 889
 890 if __name__ == "__main__":
 891     main()