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         "ICU4XGraphemeClusterSegmenter.h",  # ICU4X
  78         "ICU4XSentenceSegmenter.h",  # ICU4X
  79         "ICU4XWordSegmenter.h",  # ICU4X
  80         "mozmemory.h",  # included without a path
  81         "pratom.h",  # NSPR
  82         "prcvar.h",  # NSPR
  83         "prerror.h",  # NSPR
  84         "prinit.h",  # NSPR
  85         "prio.h",  # NSPR
  86         "private/pprio.h",  # NSPR
  87         "prlink.h",  # NSPR
  88         "prlock.h",  # NSPR
  89         "prprf.h",  # NSPR
  90         "prthread.h",  # NSPR
  91         "prtypes.h",  # NSPR
  92         "selfhosted.out.h",  # generated in $OBJDIR
  93         "shellmoduleloader.out.h",  # generated in $OBJDIR
  94         "unicode/locid.h",  # ICU
  95         "unicode/uchar.h",  # ICU
  96         "unicode/uniset.h",  # ICU
  97         "unicode/unistr.h",  # ICU
  98         "unicode/utypes.h",  # ICU
  99         "vtune/VTuneWrapper.h",  # VTune
 100         "wasm/WasmBuiltinModuleGenerated.h",  # generated in $OBJDIR"
 101         "zydis/ZydisAPI.h",  # Zydis
 102     ]
 103 )
 104
 105 deprecated_inclnames = {
 106     "mozilla/Unused.h": "Use [[nodiscard]] and (void)expr casts instead.",
 107 }
 108
 109 # JSAPI functions should be included through headers from js/public instead of
 110 # using the old, catch-all jsapi.h file.
 111 deprecated_inclnames_in_header = {
 112     "jsapi.h": "Prefer including headers from js/public.",
 113 }
 114
 115 # Temporary exclusions for files which still need to include jsapi.h.
 116 deprecated_inclnames_in_header_excludes = {
 117     "js/src/vm/Compartment-inl.h",  # for JS::InformalValueTypeName
 118     "js/src/jsapi-tests/tests.h",  # for JS_ValueToSource
 119 }
 120
 121 # These files have additional constraints on where they are #included, so we
 122 # ignore #includes of them when checking #include ordering.
 123 oddly_ordered_inclnames = set(
 124     [
 125         "ctypes/typedefs.h",  # Included multiple times in the body of ctypes/CTypes.h
 126         # Included in the body of frontend/TokenStream.h
 127         "frontend/ReservedWordsGenerated.h",
 128         "gc/StatsPhasesGenerated.h",  # Included in the body of gc/Statistics.h
 129         "gc/StatsPhasesGenerated.inc",  # Included in the body of gc/Statistics.cpp
 130         "psapi.h",  # Must be included after "util/WindowsWrapper.h" on Windows
 131         "machine/endian.h",  # Must be included after <sys/types.h> on BSD
 132         "process.h",  # Windows-specific
 133         "winbase.h",  # Must precede other system headers(?)
 134         "windef.h",  # Must precede other system headers(?)
 135         "windows.h",  # Must precede other system headers(?)
 136     ]
 137 )
 138
 139 # The files in tests/style/ contain code that fails this checking in various
 140 # ways.  Here is the output we expect.  If the actual output differs from
 141 # this, one of the following must have happened.
 142 # - New SpiderMonkey code violates one of the checked rules.
 143 # - The tests/style/ files have changed without expected_output being changed
 144 #   accordingly.
 145 # - This script has been broken somehow.
 146 #
 147 expected_output = """\
 148 js/src/tests/style/BadIncludes.h:3: error:
 149     the file includes itself
 150
 151 js/src/tests/style/BadIncludes.h:6: error:
 152     "BadIncludes2.h" is included using the wrong path;
 153     did you forget a prefix, or is the file not yet committed?
 154
 155 js/src/tests/style/BadIncludes.h:8: error:
 156     <tests/style/BadIncludes2.h> should be included using
 157     the #include "..." form
 158
 159 js/src/tests/style/BadIncludes.h:10: error:
 160     "stdio.h" is included using the wrong path;
 161     did you forget a prefix, or is the file not yet committed?
 162
 163 js/src/tests/style/BadIncludes.h:12: error:
 164     "mozilla/Unused.h" is deprecated: Use [[nodiscard]] and (void)expr casts instead.
 165
 166 js/src/tests/style/BadIncludes2.h:1: error:
 167     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 168
 169 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 170     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 171
 172 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 173     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 174
 175 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 176     "js/Value.h" should be included after "ds/LifoAlloc.h"
 177
 178 js/src/tests/style/BadIncludesOrder-inl.h:9: error:
 179     "jsapi.h" is deprecated: Prefer including headers from js/public.
 180
 181 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 182     "ds/LifoAlloc.h" should be included after "jsapi.h"
 183
 184 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 185     "jsapi.h" should be included after <stdio.h>
 186
 187 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 188     <stdio.h> should be included after "mozilla/HashFunctions.h"
 189
 190 js/src/tests/style/BadIncludesOrder-inl.h:20: error:
 191     "jsapi.h" is deprecated: Prefer including headers from js/public.
 192
 193 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 194     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 195
 196 (multiple files): error:
 197     header files form one or more cycles
 198
 199    tests/style/HeaderCycleA1.h
 200    -> tests/style/HeaderCycleA2.h
 201       -> tests/style/HeaderCycleA3.h
 202          -> tests/style/HeaderCycleA1.h
 203
 204    tests/style/HeaderCycleB1-inl.h
 205    -> tests/style/HeaderCycleB2-inl.h
 206       -> tests/style/HeaderCycleB3-inl.h
 207          -> tests/style/HeaderCycleB4-inl.h
 208             -> tests/style/HeaderCycleB1-inl.h
 209             -> tests/style/jsheadercycleB5inlines.h
 210                -> tests/style/HeaderCycleB1-inl.h
 211       -> tests/style/HeaderCycleB4-inl.h
 212
 213 """.splitlines(
 214     True
 215 )
 216
 217 actual_output = []
 218
 219
 220 def out(*lines):
 221     for line in lines:
 222         actual_output.append(line + "\n")
 223
 224
 225 def error(filename, linenum, *lines):
 226     location = filename
 227     if linenum is not None:
 228         location += ":" + str(linenum)
 229     out(location + ": error:")
 230     for line in lines:
 231         out("    " + line)
 232     out("")
 233
 234
 235 class FileKind(object):
 236     C = 1
 237     CPP = 2
 238     INL_H = 3
 239     H = 4
 240     TBL = 5
 241     MSG = 6
 242
 243     @staticmethod
 244     def get(filename):
 245         if filename.endswith(".c"):
 246             return FileKind.C
 247
 248         if filename.endswith(".cpp"):
 249             return FileKind.CPP
 250
 251         if filename.endswith(("inlines.h", "-inl.h")):
 252             return FileKind.INL_H
 253
 254         if filename.endswith(".h"):
 255             return FileKind.H
 256
 257         if filename.endswith(".tbl"):
 258             return FileKind.TBL
 259
 260         if filename.endswith(".msg"):
 261             return FileKind.MSG
 262
 263         error(filename, None, "unknown file kind")
 264
 265
 266 def check_style(enable_fixup):
 267     # We deal with two kinds of name.
 268     # - A "filename" is a full path to a file from the repository root.
 269     # - An "inclname" is how a file is referred to in a #include statement.
 270     #
 271     # Examples (filename -> inclname)
 272     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 273     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 274     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 275     # - "js/public/Vector.h"        -> "js/Vector.h"
 276     # - "js/src/vm/String.h"        -> "vm/String.h"
 277
 278     non_js_dirnames = (
 279         "mfbt/",
 280         "memory/mozalloc/",
 281         "mozglue/",
 282         "intl/components/",
 283     )  # type: tuple(str)
 284     non_js_inclnames = set()  # type: set(inclname)
 285     js_names = dict()  # type: dict(filename, inclname)
 286
 287     # Process files in js/src.
 288     js_src_root = os.path.join("js", "src")
 289     for dirpath, dirnames, filenames in os.walk(js_src_root):
 290         if dirpath == js_src_root:
 291             # Skip any subdirectories that contain a config.status file
 292             # (likely objdirs).
 293             builddirs = []
 294             for dirname in dirnames:
 295                 path = os.path.join(dirpath, dirname, "config.status")
 296                 if os.path.isfile(path):
 297                     builddirs.append(dirname)
 298             for dirname in builddirs:
 299                 dirnames.remove(dirname)
 300         for filename in filenames:
 301             filepath = os.path.join(dirpath, filename).replace("\\", "/")
 302             if not filepath.startswith(
 303                 tuple(ignored_js_src_dirs)
 304             ) and filepath.endswith((".c", ".cpp", ".h", ".tbl", ".msg")):
 305                 inclname = filepath[len("js/src/") :]
 306                 js_names[filepath] = inclname
 307
 308     # Look for header files in directories in non_js_dirnames.
 309     for non_js_dir in non_js_dirnames:
 310         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 311             for filename in filenames:
 312                 if filename.endswith(".h"):
 313                     inclname = "mozilla/" + filename
 314                     if non_js_dir == "intl/components/":
 315                         inclname = "mozilla/intl/" + filename
 316                     non_js_inclnames.add(inclname)
 317
 318     # Look for header files in js/public.
 319     js_public_root = os.path.join("js", "public")
 320     for dirpath, dirnames, filenames in os.walk(js_public_root):
 321         for filename in filenames:
 322             if filename.endswith((".h", ".msg")):
 323                 filepath = os.path.join(dirpath, filename).replace("\\", "/")
 324                 inclname = "js/" + filepath[len("js/public/") :]
 325                 js_names[filepath] = inclname
 326
 327     all_inclnames = non_js_inclnames | set(js_names.values())
 328
 329     edges = dict()  # type: dict(inclname, set(inclname))
 330
 331     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 332     # are #included from JS files we have to add them to the inclusion graph.
 333     for inclname in non_js_inclnames:
 334         edges[inclname] = set()
 335
 336     # Process all the JS files.
 337     for filename in sorted(js_names.keys()):
 338         inclname = js_names[filename]
 339         file_kind = FileKind.get(filename)
 340         if (
 341             file_kind == FileKind.C
 342             or file_kind == FileKind.CPP
 343             or file_kind == FileKind.H
 344             or file_kind == FileKind.INL_H
 345         ):
 346             included_h_inclnames = set()  # type: set(inclname)
 347
 348             with open(filename, encoding="utf-8") as f:
 349                 code = read_file(f)
 350
 351             if enable_fixup:
 352                 code = code.sorted(inclname)
 353                 with open(filename, "w") as f:
 354                     f.write(code.to_source())
 355
 356             check_file(
 357                 filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 358             )
 359
 360         edges[inclname] = included_h_inclnames
 361
 362     find_cycles(all_inclnames, edges)
 363
 364     # Compare expected and actual output.
 365     difflines = difflib.unified_diff(
 366         expected_output,
 367         actual_output,
 368         fromfile="check_spidermonkey_style.py expected output",
 369         tofile="check_spidermonkey_style.py actual output",
 370     )
 371     ok = True
 372     for diffline in difflines:
 373         ok = False
 374         print(diffline, end="")
 375
 376     return ok
 377
 378
 379 def module_name(name):
 380     """Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename."""
 381
 382     return (
 383         name.replace("inlines.h", "")
 384         .replace("-inl.h", "")
 385         .replace(".h", "")
 386         .replace(".cpp", "")
 387     )  # NOQA: E501
 388
 389
 390 def is_module_header(enclosing_inclname, header_inclname):
 391     """Determine if an included name is the "module header", i.e. should be
 392     first in the file."""
 393
 394     module = module_name(enclosing_inclname)
 395
 396     # Normal case, for example:
 397     #   module == "vm/Runtime", header_inclname == "vm/Runtime.h".
 398     if module == module_name(header_inclname):
 399         return True
 400
 401     # A public header, for example:
 402     #
 403     #   module == "vm/CharacterEncoding",
 404     #   header_inclname == "js/CharacterEncoding.h"
 405     #
 406     # or (for implementation files for js/public/*/*.h headers)
 407     #
 408     #   module == "vm/SourceHook",
 409     #   header_inclname == "js/experimental/SourceHook.h"
 410     m = re.match(r"js\/.*?([^\/]+)\.h", header_inclname)
 411     if m is not None and module.endswith("/" + m.group(1)):
 412         return True
 413
 414     return False
 415
 416
 417 class Include(object):
 418     """Important information for a single #include statement."""
 419
 420     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 421         self.include_prefix = include_prefix
 422         self.line_suffix = line_suffix
 423         self.inclname = inclname
 424         self.linenum = linenum
 425         self.is_system = is_system
 426
 427     def is_style_relevant(self):
 428         # Includes are style-relevant; that is, they're checked by the pairwise
 429         # style-checking algorithm in check_file.
 430         return True
 431
 432     def section(self, enclosing_inclname):
 433         """Identify which section inclname belongs to.
 434
 435         The section numbers are as follows.
 436           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 437           1. mozilla/Foo.h
 438           2. <foo.h> or <foo>
 439           3. jsfoo.h, prmjtime.h, etc
 440           4. foo/Bar.h
 441           5. jsfooinlines.h
 442           6. foo/Bar-inl.h
 443           7. non-.h, e.g. *.tbl, *.msg (these can be scattered throughout files)
 444         """
 445
 446         if self.is_system:
 447             return 2
 448
 449         if not self.inclname.endswith(".h"):
 450             return 7
 451
 452         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 453         # special handling.
 454         if is_module_header(enclosing_inclname, self.inclname):
 455             return 0
 456
 457         if "/" in self.inclname:
 458             if self.inclname.startswith("mozilla/"):
 459                 return 1
 460
 461             if self.inclname.endswith("-inl.h"):
 462                 return 6
 463
 464             return 4
 465
 466         if self.inclname.endswith("inlines.h"):
 467             return 5
 468
 469         return 3
 470
 471     def quote(self):
 472         if self.is_system:
 473             return "<" + self.inclname + ">"
 474         else:
 475             return '"' + self.inclname + '"'
 476
 477     def sort_key(self, enclosing_inclname):
 478         return (self.section(enclosing_inclname), self.inclname.lower())
 479
 480     def to_source(self):
 481         return self.include_prefix + self.quote() + self.line_suffix + "\n"
 482
 483
 484 class CppBlock(object):
 485     """C preprocessor block: a whole file or a single #if/#elif/#else block.
 486
 487     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 488     The top-level block, which is not within a #if/#endif pair, is also
 489     considered a block.
 490
 491     Each kid is either an Include (representing a #include), OrdinaryCode, or
 492     a nested CppBlock."""
 493
 494     def __init__(self, start_line=""):
 495         self.start = start_line
 496         self.end = ""
 497         self.kids = []
 498
 499     def is_style_relevant(self):
 500         return True
 501
 502     def append_ordinary_line(self, line):
 503         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 504             self.kids.append(OrdinaryCode())
 505         self.kids[-1].lines.append(line)
 506
 507     def style_relevant_kids(self):
 508         """Return a list of kids in this block that are style-relevant."""
 509         return [kid for kid in self.kids if kid.is_style_relevant()]
 510
 511     def sorted(self, enclosing_inclname):
 512         """Return a hopefully-sorted copy of this block. Implements --fixup.
 513
 514         When in doubt, this leaves the code unchanged.
 515         """
 516
 517         def pretty_sorted_includes(includes):
 518             """Return a new list containing the given includes, in order,
 519             with blank lines separating sections."""
 520             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 521             if sorted(keys) == keys:
 522                 return includes  # if nothing is out of order, don't touch anything
 523
 524             output = []
 525             current_section = None
 526             for (section, _), inc in sorted(zip(keys, includes)):
 527                 if current_section is not None and section != current_section:
 528                     output.append(OrdinaryCode(["\n"]))  # blank line
 529                 output.append(inc)
 530                 current_section = section
 531             return output
 532
 533         def should_try_to_sort(includes):
 534             if "tests/style/BadIncludes" in enclosing_inclname:
 535                 return False  # don't straighten the counterexample
 536             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 537                 return False  # don't sort batches containing odd includes
 538             if includes == sorted(
 539                 includes, key=lambda inc: inc.sort_key(enclosing_inclname)
 540             ):
 541                 return False  # it's already sorted, avoid whitespace-only fixups
 542             return True
 543
 544         # The content of the eventual output of this method.
 545         output = []
 546
 547         # The current batch of includes to sort. This list only ever contains Include objects
 548         # and whitespace OrdinaryCode objects.
 549         batch = []
 550
 551         def flush_batch():
 552             """Sort the contents of `batch` and move it to `output`."""
 553
 554             assert all(
 555                 isinstance(item, Include)
 556                 or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 557                 for item in batch
 558             )
 559
 560             # Here we throw away the blank lines.
 561             # `pretty_sorted_includes` puts them back.
 562             includes = []
 563             last_include_index = -1
 564             for i, item in enumerate(batch):
 565                 if isinstance(item, Include):
 566                     includes.append(item)
 567                     last_include_index = i
 568             cutoff = last_include_index + 1
 569
 570             if should_try_to_sort(includes):
 571                 output.extend(pretty_sorted_includes(includes) + batch[cutoff:])
 572             else:
 573                 output.extend(batch)
 574             del batch[:]
 575
 576         for kid in self.kids:
 577             if isinstance(kid, CppBlock):
 578                 flush_batch()
 579                 output.append(kid.sorted(enclosing_inclname))
 580             elif isinstance(kid, Include):
 581                 batch.append(kid)
 582             else:
 583                 assert isinstance(kid, OrdinaryCode)
 584                 if kid.to_source().isspace():
 585                     batch.append(kid)
 586                 else:
 587                     flush_batch()
 588                     output.append(kid)
 589         flush_batch()
 590
 591         result = CppBlock()
 592         result.start = self.start
 593         result.end = self.end
 594         result.kids = output
 595         return result
 596
 597     def to_source(self):
 598         return self.start + "".join(kid.to_source() for kid in self.kids) + self.end
 599
 600
 601 class OrdinaryCode(object):
 602     """A list of lines of code that aren't #include/#if/#else/#endif lines."""
 603
 604     def __init__(self, lines=None):
 605         self.lines = lines if lines is not None else []
 606
 607     def is_style_relevant(self):
 608         return False
 609
 610     def to_source(self):
 611         return "".join(self.lines)
 612
 613
 614 # A "snippet" is one of:
 615 #
 616 # *   Include - representing an #include line
 617 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 618 # *   OrdinaryCode - representing lines of non-#include-relevant code
 619
 620
 621 def read_file(f):
 622     block_stack = [CppBlock()]
 623
 624     # Extract the #include statements as a tree of snippets.
 625     for linenum, line in enumerate(f, start=1):
 626         if line.lstrip().startswith("#"):
 627             # Look for a |#include "..."| line.
 628             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 629             if m is not None:
 630                 prefix, inclname, suffix = m.groups()
 631                 block_stack[-1].kids.append(
 632                     Include(prefix, inclname, suffix, linenum, is_system=False)
 633                 )
 634                 continue
 635
 636             # Look for a |#include <...>| line.
 637             m = re.match(r"(\s*#\s*include\s+)<([^>]*)>(.*)", line)
 638             if m is not None:
 639                 prefix, inclname, suffix = m.groups()
 640                 block_stack[-1].kids.append(
 641                     Include(prefix, inclname, suffix, linenum, is_system=True)
 642                 )
 643                 continue
 644
 645             # Look for a |#{if,ifdef,ifndef}| line.
 646             m = re.match(r"\s*#\s*(if|ifdef|ifndef)\b", line)
 647             if m is not None:
 648                 # Open a new block.
 649                 new_block = CppBlock(line)
 650                 block_stack[-1].kids.append(new_block)
 651                 block_stack.append(new_block)
 652                 continue
 653
 654             # Look for a |#{elif,else}| line.
 655             m = re.match(r"\s*#\s*(elif|else)\b", line)
 656             if m is not None:
 657                 # Close the current block, and open an adjacent one.
 658                 block_stack.pop()
 659                 new_block = CppBlock(line)
 660                 block_stack[-1].kids.append(new_block)
 661                 block_stack.append(new_block)
 662                 continue
 663
 664             # Look for a |#endif| line.
 665             m = re.match(r"\s*#\s*endif\b", line)
 666             if m is not None:
 667                 # Close the current block.
 668                 block_stack.pop().end = line
 669                 if len(block_stack) == 0:
 670                     raise ValueError("#endif without #if at line " + str(linenum))
 671                 continue
 672
 673         # Otherwise, we have an ordinary line.
 674         block_stack[-1].append_ordinary_line(line)
 675
 676     if len(block_stack) > 1:
 677         raise ValueError("unmatched #if")
 678     return block_stack[-1]
 679
 680
 681 def check_file(
 682     filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 683 ):
 684     def check_include_statement(include):
 685         """Check the style of a single #include statement."""
 686
 687         if include.is_system:
 688             # Check it is not a known local file (in which case it's probably a system header).
 689             if (
 690                 include.inclname in included_inclnames_to_ignore
 691                 or include.inclname in all_inclnames
 692             ):
 693                 error(
 694                     filename,
 695                     include.linenum,
 696                     include.quote() + " should be included using",
 697                     'the #include "..." form',
 698                 )
 699
 700         else:
 701             msg = deprecated_inclnames.get(include.inclname)
 702             if msg:
 703                 error(
 704                     filename,
 705                     include.linenum,
 706                     include.quote() + " is deprecated: " + msg,
 707                 )
 708
 709             if file_kind == FileKind.H or file_kind == FileKind.INL_H:
 710                 msg = deprecated_inclnames_in_header.get(include.inclname)
 711                 if msg and filename not in deprecated_inclnames_in_header_excludes:
 712                     error(
 713                         filename,
 714                         include.linenum,
 715                         include.quote() + " is deprecated: " + msg,
 716                     )
 717
 718             if include.inclname not in included_inclnames_to_ignore:
 719                 included_kind = FileKind.get(include.inclname)
 720
 721                 # Check the #include path has the correct form.
 722                 if include.inclname not in all_inclnames:
 723                     error(
 724                         filename,
 725                         include.linenum,
 726                         include.quote() + " is included using the wrong path;",
 727                         "did you forget a prefix, or is the file not yet committed?",
 728                     )
 729
 730                 # Record inclusions of .h files for cycle detection later.
 731                 # (Exclude .tbl and .msg files.)
 732                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 733                     included_h_inclnames.add(include.inclname)
 734
 735                 # Check a H file doesn't #include an INL_H file.
 736                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 737                     error(
 738                         filename,
 739                         include.linenum,
 740                         "vanilla header includes an inline-header file "
 741                         + include.quote(),
 742                     )
 743
 744                 # Check a file doesn't #include itself.  (We do this here because the cycle
 745                 # detection below doesn't detect this case.)
 746                 if inclname == include.inclname:
 747                     error(filename, include.linenum, "the file includes itself")
 748
 749     def check_includes_order(include1, include2):
 750         """Check the ordering of two #include statements."""
 751
 752         if (
 753             include1.inclname in oddly_ordered_inclnames
 754             or include2.inclname in oddly_ordered_inclnames
 755         ):
 756             return
 757
 758         section1 = include1.section(inclname)
 759         section2 = include2.section(inclname)
 760         if (section1 > section2) or (
 761             (section1 == section2)
 762             and (include1.inclname.lower() > include2.inclname.lower())
 763         ):
 764             error(
 765                 filename,
 766                 str(include1.linenum) + ":" + str(include2.linenum),
 767                 include1.quote() + " should be included after " + include2.quote(),
 768             )
 769
 770     # Check the extracted #include statements, both individually, and the ordering of
 771     # adjacent pairs that live in the same block.
 772     def pair_traverse(prev, this):
 773         if isinstance(this, Include):
 774             check_include_statement(this)
 775             if isinstance(prev, Include):
 776                 check_includes_order(prev, this)
 777         else:
 778             kids = this.style_relevant_kids()
 779             for prev2, this2 in zip([None] + kids[0:-1], kids):
 780                 pair_traverse(prev2, this2)
 781
 782     pair_traverse(None, code)
 783
 784
 785 def find_cycles(all_inclnames, edges):
 786     """Find and draw any cycles."""
 787
 788     SCCs = tarjan(all_inclnames, edges)
 789
 790     # The various sorted() calls below ensure the output is deterministic.
 791
 792     def draw_SCC(c):
 793         cset = set(c)
 794         drawn = set()
 795
 796         def draw(v, indent):
 797             out("   " * indent + ("-> " if indent else "   ") + v)
 798             if v in drawn:
 799                 return
 800             drawn.add(v)
 801             for succ in sorted(edges[v]):
 802                 if succ in cset:
 803                     draw(succ, indent + 1)
 804
 805         draw(sorted(c)[0], 0)
 806         out("")
 807
 808     have_drawn_an_SCC = False
 809     for scc in sorted(SCCs):
 810         if len(scc) != 1:
 811             if not have_drawn_an_SCC:
 812                 error("(multiple files)", None, "header files form one or more cycles")
 813                 have_drawn_an_SCC = True
 814
 815             draw_SCC(scc)
 816
 817
 818 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 819 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 820 def tarjan(V, E):
 821     vertex_index = {}
 822     vertex_lowlink = {}
 823     index = 0
 824     S = []
 825     all_SCCs = []
 826
 827     def strongconnect(v, index):
 828         # Set the depth index for v to the smallest unused index
 829         vertex_index[v] = index
 830         vertex_lowlink[v] = index
 831         index += 1
 832         S.append(v)
 833
 834         # Consider successors of v
 835         for w in E[v]:
 836             if w not in vertex_index:
 837                 # Successor w has not yet been visited; recurse on it
 838                 index = strongconnect(w, index)
 839                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 840             elif w in S:
 841                 # Successor w is in stack S and hence in the current SCC
 842                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 843
 844         # If v is a root node, pop the stack and generate an SCC
 845         if vertex_lowlink[v] == vertex_index[v]:
 846             i = S.index(v)
 847             scc = S[i:]
 848             del S[i:]
 849             all_SCCs.append(scc)
 850
 851         return index
 852
 853     for v in V:
 854         if v not in vertex_index:
 855             index = strongconnect(v, index)
 856
 857     return all_SCCs
 858
 859
 860 def main():
 861     if sys.argv[1:] == ["--fixup"]:
 862         # Sort #include directives in-place.  Fixup mode doesn't solve
 863         # all possible silliness that the script checks for; it's just a
 864         # hack for the common case where renaming a header causes style
 865         # errors.
 866         fixup = True
 867     elif sys.argv[1:] == []:
 868         fixup = False
 869     else:
 870         print(
 871             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 872             "line options: " + repr(sys.argv[1:])
 873         )
 874         sys.exit(1)
 875
 876     ok = check_style(fixup)
 877
 878     if ok:
 879         print("TEST-PASS | check_spidermonkey_style.py | ok")
 880     else:
 881         print(
 882             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 883             + "actual output does not match expected output;  diff is above."
 884         )
 885         print(
 886             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 887             + "Hint: If the problem is that you renamed a header, and many #includes "
 888             + "are no longer in alphabetical order, commit your work and then try "
 889             + "`check_spidermonkey_style.py --fixup`. "
 890             + "You need to commit first because --fixup modifies your files in place."
 891         )
 892
 893     sys.exit(0 if ok else 1)
 894
 895
 896 if __name__ == "__main__":
 897     main()