config/check_spidermonkey_style.py

   1 # vim: set ts=8 sts=4 et sw=4 tw=99:
   2 # This Source Code Form is subject to the terms of the Mozilla Public
   3 # License, v. 2.0. If a copy of the MPL was not distributed with this
   4 # file, You can obtain one at http://mozilla.org/MPL/2.0/.
   5
   6 # ----------------------------------------------------------------------------
   7 # This script checks various aspects of SpiderMonkey code style.  The current checks are as
   8 # follows.
   9 #
  10 # We check the following things in headers.
  11 #
  12 # - No cyclic dependencies.
  13 #
  14 # - No normal header should #include a inlines.h/-inl.h file.
  15 #
  16 # - #ifndef wrappers should have the right form. (XXX: not yet implemented)
  17 #   - Every header file should have one.
  18 #   - The guard name used should be appropriate for the filename.
  19 #
  20 # We check the following things in all files.
  21 #
  22 # - #includes should have full paths, e.g. "jit/Ion.h", not "Ion.h".
  23 #
  24 # - #includes should use the appropriate form for system headers (<...>) and
  25 #   local headers ("...").
  26 #
  27 # - #includes should be ordered correctly.
  28 #   - Each one should be in the correct section.
  29 #   - Alphabetical order should be used within sections.
  30 #   - Sections should be in the right order.
  31 #   Note that the presence of #if/#endif blocks complicates things, to the
  32 #   point that it's not always clear where a conditionally-compiled #include
  33 #   statement should go, even to a human.  Therefore, we check the #include
  34 #   statements within each #if/#endif block (including nested ones) in
  35 #   isolation, but don't try to do any order checking between such blocks.
  36 # ----------------------------------------------------------------------------
  37
  38 from __future__ import absolute_import, print_function
  39
  40 import difflib
  41 import os
  42 import re
  43 import sys
  44
  45 # We don't bother checking files in these directories, because they're (a) auxiliary or (b)
  46 # imported code that doesn't follow our coding style.
  47 ignored_js_src_dirs = [
  48     "js/src/config/",  # auxiliary stuff
  49     "js/src/ctypes/libffi/",  # imported code
  50     "js/src/devtools/",  # auxiliary stuff
  51     "js/src/editline/",  # imported code
  52     "js/src/gdb/",  # auxiliary stuff
  53     "js/src/vtune/",  # imported code
  54     "js/src/zydis/",  # imported code
  55 ]
  56
  57 # We ignore #includes of these files, because they don't follow the usual rules.
  58 included_inclnames_to_ignore = set(
  59     [
  60         "ffi.h",  # generated in ctypes/libffi/
  61         "devtools/Instruments.h",  # we ignore devtools/ in general
  62         "double-conversion/double-conversion.h",  # strange MFBT case
  63         "javascript-trace.h",  # generated in $OBJDIR if HAVE_DTRACE is defined
  64         "frontend/ReservedWordsGenerated.h",  # generated in $OBJDIR
  65         "frontend/smoosh_generated.h",  # generated in $OBJDIR
  66         "gc/StatsPhasesGenerated.h",  # generated in $OBJDIR
  67         "gc/StatsPhasesGenerated.inc",  # generated in $OBJDIR
  68         "jit/CacheIROpsGenerated.h",  # generated in $OBJDIR
  69         "jit/LIROpsGenerated.h",  # generated in $OBJDIR
  70         "jit/MIROpsGenerated.h",  # generated in $OBJDIR
  71         "js/ProfilingCategoryList.h",  # comes from mozglue/baseprofiler
  72         "jscustomallocator.h",  # provided by embedders;  allowed to be missing
  73         "js-config.h",  # generated in $OBJDIR
  74         "fdlibm.h",  # fdlibm
  75         "FuzzerDefs.h",  # included without a path
  76         "FuzzingInterface.h",  # included without a path
  77         "mozmemory.h",  # included without a path
  78         "pratom.h",  # NSPR
  79         "prcvar.h",  # NSPR
  80         "prerror.h",  # NSPR
  81         "prinit.h",  # NSPR
  82         "prio.h",  # NSPR
  83         "private/pprio.h",  # NSPR
  84         "prlink.h",  # NSPR
  85         "prlock.h",  # NSPR
  86         "prprf.h",  # NSPR
  87         "prthread.h",  # NSPR
  88         "prtypes.h",  # NSPR
  89         "selfhosted.out.h",  # generated in $OBJDIR
  90         "shellmoduleloader.out.h",  # generated in $OBJDIR
  91         "unicode/basictz.h",  # ICU
  92         "unicode/locid.h",  # ICU
  93         "unicode/plurrule.h",  # ICU
  94         "unicode/putil.h",  # ICU
  95         "unicode/timezone.h",  # ICU
  96         "unicode/ucal.h",  # ICU
  97         "unicode/uchar.h",  # ICU
  98         "unicode/uclean.h",  # ICU
  99         "unicode/ucol.h",  # ICU
 100         "unicode/ucurr.h",  # ICU
 101         "unicode/udat.h",  # ICU
 102         "unicode/udata.h",  # ICU
 103         "unicode/udateintervalformat.h",  # ICU
 104         "unicode/udatpg.h",  # ICU
 105         "unicode/udisplaycontext.h",  # ICU
 106         "unicode/uenum.h",  # ICU
 107         "unicode/ufieldpositer.h",  # ICU
 108         "unicode/uformattedvalue.h",  # ICU
 109         "unicode/ulistformatter.h",  # ICU
 110         "unicode/uldnames.h",  # ICU
 111         "unicode/uloc.h",  # ICU
 112         "unicode/umachine.h",  # ICU
 113         "unicode/uniset.h",  # ICU
 114         "unicode/unistr.h",  # ICU
 115         "unicode/unorm2.h",  # ICU
 116         "unicode/unum.h",  # ICU
 117         "unicode/unumberformatter.h",  # ICU
 118         "unicode/unumsys.h",  # ICU
 119         "unicode/upluralrules.h",  # ICU
 120         "unicode/ureldatefmt.h",  # ICU
 121         "unicode/ures.h",  # ICU
 122         "unicode/ustring.h",  # ICU
 123         "unicode/utypes.h",  # ICU
 124         "unicode/uversion.h",  # ICU
 125         "vtune/VTuneWrapper.h",  # VTune
 126         "wasm/WasmIntrinsicGenerated.h",  # generated in $OBJDIR"
 127         "zydis/ZydisAPI.h",  # Zydis
 128     ]
 129 )
 130
 131 deprecated_inclnames = {
 132     "mozilla/Unused.h": "Use [[nodiscard]] and (void)expr casts instead.",
 133 }
 134
 135 # JSAPI functions should be included through headers from js/public instead of
 136 # using the old, catch-all jsapi.h file.
 137 deprecated_inclnames_in_header = {
 138     "jsapi.h": "Prefer including headers from js/public.",
 139 }
 140
 141 # Temporary exclusions for files which still need to include jsapi.h.
 142 deprecated_inclnames_in_header_excludes = {
 143     "js/src/vm/Compartment-inl.h",  # for JS::InformalValueTypeName
 144     "js/src/jsapi-tests/tests.h",  # for JS_ValueToSource
 145 }
 146
 147 # These files have additional constraints on where they are #included, so we
 148 # ignore #includes of them when checking #include ordering.
 149 oddly_ordered_inclnames = set(
 150     [
 151         "ctypes/typedefs.h",  # Included multiple times in the body of ctypes/CTypes.h
 152         # Included in the body of frontend/TokenStream.h
 153         "frontend/ReservedWordsGenerated.h",
 154         "gc/StatsPhasesGenerated.h",  # Included in the body of gc/Statistics.h
 155         "gc/StatsPhasesGenerated.inc",  # Included in the body of gc/Statistics.cpp
 156         "psapi.h",  # Must be included after "util/WindowsWrapper.h" on Windows
 157         "machine/endian.h",  # Must be included after <sys/types.h> on BSD
 158         "winbase.h",  # Must precede other system headers(?)
 159         "windef.h",  # Must precede other system headers(?)
 160     ]
 161 )
 162
 163 # The files in tests/style/ contain code that fails this checking in various
 164 # ways.  Here is the output we expect.  If the actual output differs from
 165 # this, one of the following must have happened.
 166 # - New SpiderMonkey code violates one of the checked rules.
 167 # - The tests/style/ files have changed without expected_output being changed
 168 #   accordingly.
 169 # - This script has been broken somehow.
 170 #
 171 expected_output = """\
 172 js/src/tests/style/BadIncludes.h:3: error:
 173     the file includes itself
 174
 175 js/src/tests/style/BadIncludes.h:6: error:
 176     "BadIncludes2.h" is included using the wrong path;
 177     did you forget a prefix, or is the file not yet committed?
 178
 179 js/src/tests/style/BadIncludes.h:8: error:
 180     <tests/style/BadIncludes2.h> should be included using
 181     the #include "..." form
 182
 183 js/src/tests/style/BadIncludes.h:10: error:
 184     "stdio.h" is included using the wrong path;
 185     did you forget a prefix, or is the file not yet committed?
 186
 187 js/src/tests/style/BadIncludes.h:12: error:
 188     "mozilla/Unused.h" is deprecated: Use [[nodiscard]] and (void)expr casts instead.
 189
 190 js/src/tests/style/BadIncludes2.h:1: error:
 191     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 192
 193 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 194     "vm/JSScript-inl.h" should be included after "vm/Interpreter-inl.h"
 195
 196 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 197     "vm/Interpreter-inl.h" should be included after "js/Value.h"
 198
 199 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 200     "js/Value.h" should be included after "ds/LifoAlloc.h"
 201
 202 js/src/tests/style/BadIncludesOrder-inl.h:9: error:
 203     "jsapi.h" is deprecated: Prefer including headers from js/public.
 204
 205 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 206     "ds/LifoAlloc.h" should be included after "jsapi.h"
 207
 208 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 209     "jsapi.h" should be included after <stdio.h>
 210
 211 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 212     <stdio.h> should be included after "mozilla/HashFunctions.h"
 213
 214 js/src/tests/style/BadIncludesOrder-inl.h:20: error:
 215     "jsapi.h" is deprecated: Prefer including headers from js/public.
 216
 217 js/src/tests/style/BadIncludesOrder-inl.h:28:29: error:
 218     "vm/JSScript.h" should be included after "vm/JSFunction.h"
 219
 220 (multiple files): error:
 221     header files form one or more cycles
 222
 223    tests/style/HeaderCycleA1.h
 224    -> tests/style/HeaderCycleA2.h
 225       -> tests/style/HeaderCycleA3.h
 226          -> tests/style/HeaderCycleA1.h
 227
 228    tests/style/HeaderCycleB1-inl.h
 229    -> tests/style/HeaderCycleB2-inl.h
 230       -> tests/style/HeaderCycleB3-inl.h
 231          -> tests/style/HeaderCycleB4-inl.h
 232             -> tests/style/HeaderCycleB1-inl.h
 233             -> tests/style/jsheadercycleB5inlines.h
 234                -> tests/style/HeaderCycleB1-inl.h
 235       -> tests/style/HeaderCycleB4-inl.h
 236
 237 """.splitlines(
 238     True
 239 )
 240
 241 actual_output = []
 242
 243
 244 def out(*lines):
 245     for line in lines:
 246         actual_output.append(line + "\n")
 247
 248
 249 def error(filename, linenum, *lines):
 250     location = filename
 251     if linenum is not None:
 252         location += ":" + str(linenum)
 253     out(location + ": error:")
 254     for line in lines:
 255         out("    " + line)
 256     out("")
 257
 258
 259 class FileKind(object):
 260     C = 1
 261     CPP = 2
 262     INL_H = 3
 263     H = 4
 264     TBL = 5
 265     MSG = 6
 266
 267     @staticmethod
 268     def get(filename):
 269         if filename.endswith(".c"):
 270             return FileKind.C
 271
 272         if filename.endswith(".cpp"):
 273             return FileKind.CPP
 274
 275         if filename.endswith(("inlines.h", "-inl.h")):
 276             return FileKind.INL_H
 277
 278         if filename.endswith(".h"):
 279             return FileKind.H
 280
 281         if filename.endswith(".tbl"):
 282             return FileKind.TBL
 283
 284         if filename.endswith(".msg"):
 285             return FileKind.MSG
 286
 287         error(filename, None, "unknown file kind")
 288
 289
 290 def check_style(enable_fixup):
 291     # We deal with two kinds of name.
 292     # - A "filename" is a full path to a file from the repository root.
 293     # - An "inclname" is how a file is referred to in a #include statement.
 294     #
 295     # Examples (filename -> inclname)
 296     # - "mfbt/Attributes.h"         -> "mozilla/Attributes.h"
 297     # - "mozglue/misc/TimeStamp.h   -> "mozilla/TimeStamp.h"
 298     # - "memory/mozalloc/mozalloc.h -> "mozilla/mozalloc.h"
 299     # - "js/public/Vector.h"        -> "js/Vector.h"
 300     # - "js/src/vm/String.h"        -> "vm/String.h"
 301
 302     non_js_dirnames = (
 303         "mfbt/",
 304         "memory/mozalloc/",
 305         "mozglue/",
 306         "intl/components/",
 307     )  # type: tuple(str)
 308     non_js_inclnames = set()  # type: set(inclname)
 309     js_names = dict()  # type: dict(filename, inclname)
 310
 311     # Process files in js/src.
 312     js_src_root = os.path.join("js", "src")
 313     for dirpath, dirnames, filenames in os.walk(js_src_root):
 314         if dirpath == js_src_root:
 315             # Skip any subdirectories that contain a config.status file
 316             # (likely objdirs).
 317             builddirs = []
 318             for dirname in dirnames:
 319                 path = os.path.join(dirpath, dirname, "config.status")
 320                 if os.path.isfile(path):
 321                     builddirs.append(dirname)
 322             for dirname in builddirs:
 323                 dirnames.remove(dirname)
 324         for filename in filenames:
 325             filepath = os.path.join(dirpath, filename).replace("\\", "/")
 326             if not filepath.startswith(
 327                 tuple(ignored_js_src_dirs)
 328             ) and filepath.endswith((".c", ".cpp", ".h", ".tbl", ".msg")):
 329                 inclname = filepath[len("js/src/") :]
 330                 js_names[filepath] = inclname
 331
 332     # Look for header files in directories in non_js_dirnames.
 333     for non_js_dir in non_js_dirnames:
 334         for dirpath, dirnames, filenames in os.walk(non_js_dir):
 335             for filename in filenames:
 336                 if filename.endswith(".h"):
 337                     inclname = "mozilla/" + filename
 338                     if non_js_dir == "intl/components/":
 339                         inclname = "mozilla/intl/" + filename
 340                     non_js_inclnames.add(inclname)
 341
 342     # Look for header files in js/public.
 343     js_public_root = os.path.join("js", "public")
 344     for dirpath, dirnames, filenames in os.walk(js_public_root):
 345         for filename in filenames:
 346             if filename.endswith((".h", ".msg")):
 347                 filepath = os.path.join(dirpath, filename).replace("\\", "/")
 348                 inclname = "js/" + filepath[len("js/public/") :]
 349                 js_names[filepath] = inclname
 350
 351     all_inclnames = non_js_inclnames | set(js_names.values())
 352
 353     edges = dict()  # type: dict(inclname, set(inclname))
 354
 355     # We don't care what's inside the MFBT and MOZALLOC files, but because they
 356     # are #included from JS files we have to add them to the inclusion graph.
 357     for inclname in non_js_inclnames:
 358         edges[inclname] = set()
 359
 360     # Process all the JS files.
 361     for filename in sorted(js_names.keys()):
 362         inclname = js_names[filename]
 363         file_kind = FileKind.get(filename)
 364         if (
 365             file_kind == FileKind.C
 366             or file_kind == FileKind.CPP
 367             or file_kind == FileKind.H
 368             or file_kind == FileKind.INL_H
 369         ):
 370             included_h_inclnames = set()  # type: set(inclname)
 371
 372             with open(filename, encoding="utf-8") as f:
 373                 code = read_file(f)
 374
 375             if enable_fixup:
 376                 code = code.sorted(inclname)
 377                 with open(filename, "w") as f:
 378                     f.write(code.to_source())
 379
 380             check_file(
 381                 filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 382             )
 383
 384         edges[inclname] = included_h_inclnames
 385
 386     find_cycles(all_inclnames, edges)
 387
 388     # Compare expected and actual output.
 389     difflines = difflib.unified_diff(
 390         expected_output,
 391         actual_output,
 392         fromfile="check_spidermonkey_style.py expected output",
 393         tofile="check_spidermonkey_style.py actual output",
 394     )
 395     ok = True
 396     for diffline in difflines:
 397         ok = False
 398         print(diffline, end="")
 399
 400     return ok
 401
 402
 403 def module_name(name):
 404     """Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename."""
 405
 406     return (
 407         name.replace("inlines.h", "")
 408         .replace("-inl.h", "")
 409         .replace(".h", "")
 410         .replace(".cpp", "")
 411     )  # NOQA: E501
 412
 413
 414 def is_module_header(enclosing_inclname, header_inclname):
 415     """Determine if an included name is the "module header", i.e. should be
 416     first in the file."""
 417
 418     module = module_name(enclosing_inclname)
 419
 420     # Normal case, for example:
 421     #   module == "vm/Runtime", header_inclname == "vm/Runtime.h".
 422     if module == module_name(header_inclname):
 423         return True
 424
 425     # A public header, for example:
 426     #
 427     #   module == "vm/CharacterEncoding",
 428     #   header_inclname == "js/CharacterEncoding.h"
 429     #
 430     # or (for implementation files for js/public/*/*.h headers)
 431     #
 432     #   module == "vm/SourceHook",
 433     #   header_inclname == "js/experimental/SourceHook.h"
 434     m = re.match(r"js\/.*?([^\/]+)\.h", header_inclname)
 435     if m is not None and module.endswith("/" + m.group(1)):
 436         return True
 437
 438     return False
 439
 440
 441 class Include(object):
 442     """Important information for a single #include statement."""
 443
 444     def __init__(self, include_prefix, inclname, line_suffix, linenum, is_system):
 445         self.include_prefix = include_prefix
 446         self.line_suffix = line_suffix
 447         self.inclname = inclname
 448         self.linenum = linenum
 449         self.is_system = is_system
 450
 451     def is_style_relevant(self):
 452         # Includes are style-relevant; that is, they're checked by the pairwise
 453         # style-checking algorithm in check_file.
 454         return True
 455
 456     def section(self, enclosing_inclname):
 457         """Identify which section inclname belongs to.
 458
 459         The section numbers are as follows.
 460           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 461           1. mozilla/Foo.h
 462           2. <foo.h> or <foo>
 463           3. jsfoo.h, prmjtime.h, etc
 464           4. foo/Bar.h
 465           5. jsfooinlines.h
 466           6. foo/Bar-inl.h
 467           7. non-.h, e.g. *.tbl, *.msg (these can be scattered throughout files)
 468         """
 469
 470         if self.is_system:
 471             return 2
 472
 473         if not self.inclname.endswith(".h"):
 474             return 7
 475
 476         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 477         # special handling.
 478         if is_module_header(enclosing_inclname, self.inclname):
 479             return 0
 480
 481         if "/" in self.inclname:
 482             if self.inclname.startswith("mozilla/"):
 483                 return 1
 484
 485             if self.inclname.endswith("-inl.h"):
 486                 return 6
 487
 488             return 4
 489
 490         if self.inclname.endswith("inlines.h"):
 491             return 5
 492
 493         return 3
 494
 495     def quote(self):
 496         if self.is_system:
 497             return "<" + self.inclname + ">"
 498         else:
 499             return '"' + self.inclname + '"'
 500
 501     def sort_key(self, enclosing_inclname):
 502         return (self.section(enclosing_inclname), self.inclname.lower())
 503
 504     def to_source(self):
 505         return self.include_prefix + self.quote() + self.line_suffix + "\n"
 506
 507
 508 class CppBlock(object):
 509     """C preprocessor block: a whole file or a single #if/#elif/#else block.
 510
 511     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 512     The top-level block, which is not within a #if/#endif pair, is also
 513     considered a block.
 514
 515     Each kid is either an Include (representing a #include), OrdinaryCode, or
 516     a nested CppBlock."""
 517
 518     def __init__(self, start_line=""):
 519         self.start = start_line
 520         self.end = ""
 521         self.kids = []
 522
 523     def is_style_relevant(self):
 524         return True
 525
 526     def append_ordinary_line(self, line):
 527         if len(self.kids) == 0 or not isinstance(self.kids[-1], OrdinaryCode):
 528             self.kids.append(OrdinaryCode())
 529         self.kids[-1].lines.append(line)
 530
 531     def style_relevant_kids(self):
 532         """ Return a list of kids in this block that are style-relevant. """
 533         return [kid for kid in self.kids if kid.is_style_relevant()]
 534
 535     def sorted(self, enclosing_inclname):
 536         """Return a hopefully-sorted copy of this block. Implements --fixup.
 537
 538         When in doubt, this leaves the code unchanged.
 539         """
 540
 541         def pretty_sorted_includes(includes):
 542             """Return a new list containing the given includes, in order,
 543             with blank lines separating sections."""
 544             keys = [inc.sort_key(enclosing_inclname) for inc in includes]
 545             if sorted(keys) == keys:
 546                 return includes  # if nothing is out of order, don't touch anything
 547
 548             output = []
 549             current_section = None
 550             for (section, _), inc in sorted(zip(keys, includes)):
 551                 if current_section is not None and section != current_section:
 552                     output.append(OrdinaryCode(["\n"]))  # blank line
 553                 output.append(inc)
 554                 current_section = section
 555             return output
 556
 557         def should_try_to_sort(includes):
 558             if "tests/style/BadIncludes" in enclosing_inclname:
 559                 return False  # don't straighten the counterexample
 560             if any(inc.inclname in oddly_ordered_inclnames for inc in includes):
 561                 return False  # don't sort batches containing odd includes
 562             if includes == sorted(
 563                 includes, key=lambda inc: inc.sort_key(enclosing_inclname)
 564             ):
 565                 return False  # it's already sorted, avoid whitespace-only fixups
 566             return True
 567
 568         # The content of the eventual output of this method.
 569         output = []
 570
 571         # The current batch of includes to sort. This list only ever contains Include objects
 572         # and whitespace OrdinaryCode objects.
 573         batch = []
 574
 575         def flush_batch():
 576             """Sort the contents of `batch` and move it to `output`."""
 577
 578             assert all(
 579                 isinstance(item, Include)
 580                 or (isinstance(item, OrdinaryCode) and "".join(item.lines).isspace())
 581                 for item in batch
 582             )
 583
 584             # Here we throw away the blank lines.
 585             # `pretty_sorted_includes` puts them back.
 586             includes = []
 587             last_include_index = -1
 588             for i, item in enumerate(batch):
 589                 if isinstance(item, Include):
 590                     includes.append(item)
 591                     last_include_index = i
 592             cutoff = last_include_index + 1
 593
 594             if should_try_to_sort(includes):
 595                 output.extend(pretty_sorted_includes(includes) + batch[cutoff:])
 596             else:
 597                 output.extend(batch)
 598             del batch[:]
 599
 600         for kid in self.kids:
 601             if isinstance(kid, CppBlock):
 602                 flush_batch()
 603                 output.append(kid.sorted(enclosing_inclname))
 604             elif isinstance(kid, Include):
 605                 batch.append(kid)
 606             else:
 607                 assert isinstance(kid, OrdinaryCode)
 608                 if kid.to_source().isspace():
 609                     batch.append(kid)
 610                 else:
 611                     flush_batch()
 612                     output.append(kid)
 613         flush_batch()
 614
 615         result = CppBlock()
 616         result.start = self.start
 617         result.end = self.end
 618         result.kids = output
 619         return result
 620
 621     def to_source(self):
 622         return self.start + "".join(kid.to_source() for kid in self.kids) + self.end
 623
 624
 625 class OrdinaryCode(object):
 626     """ A list of lines of code that aren't #include/#if/#else/#endif lines. """
 627
 628     def __init__(self, lines=None):
 629         self.lines = lines if lines is not None else []
 630
 631     def is_style_relevant(self):
 632         return False
 633
 634     def to_source(self):
 635         return "".join(self.lines)
 636
 637
 638 # A "snippet" is one of:
 639 #
 640 # *   Include - representing an #include line
 641 # *   CppBlock - a whole file or #if/#elif/#else block; contains a list of snippets
 642 # *   OrdinaryCode - representing lines of non-#include-relevant code
 643
 644
 645 def read_file(f):
 646     block_stack = [CppBlock()]
 647
 648     # Extract the #include statements as a tree of snippets.
 649     for linenum, line in enumerate(f, start=1):
 650         if line.lstrip().startswith("#"):
 651             # Look for a |#include "..."| line.
 652             m = re.match(r'(\s*#\s*include\s+)"([^"]*)"(.*)', line)
 653             if m is not None:
 654                 prefix, inclname, suffix = m.groups()
 655                 block_stack[-1].kids.append(
 656                     Include(prefix, inclname, suffix, linenum, is_system=False)
 657                 )
 658                 continue
 659
 660             # Look for a |#include <...>| line.
 661             m = re.match(r"(\s*#\s*include\s+)<([^>]*)>(.*)", line)
 662             if m is not None:
 663                 prefix, inclname, suffix = m.groups()
 664                 block_stack[-1].kids.append(
 665                     Include(prefix, inclname, suffix, linenum, is_system=True)
 666                 )
 667                 continue
 668
 669             # Look for a |#{if,ifdef,ifndef}| line.
 670             m = re.match(r"\s*#\s*(if|ifdef|ifndef)\b", line)
 671             if m is not None:
 672                 # Open a new block.
 673                 new_block = CppBlock(line)
 674                 block_stack[-1].kids.append(new_block)
 675                 block_stack.append(new_block)
 676                 continue
 677
 678             # Look for a |#{elif,else}| line.
 679             m = re.match(r"\s*#\s*(elif|else)\b", line)
 680             if m is not None:
 681                 # Close the current block, and open an adjacent one.
 682                 block_stack.pop()
 683                 new_block = CppBlock(line)
 684                 block_stack[-1].kids.append(new_block)
 685                 block_stack.append(new_block)
 686                 continue
 687
 688             # Look for a |#endif| line.
 689             m = re.match(r"\s*#\s*endif\b", line)
 690             if m is not None:
 691                 # Close the current block.
 692                 block_stack.pop().end = line
 693                 if len(block_stack) == 0:
 694                     raise ValueError("#endif without #if at line " + str(linenum))
 695                 continue
 696
 697         # Otherwise, we have an ordinary line.
 698         block_stack[-1].append_ordinary_line(line)
 699
 700     if len(block_stack) > 1:
 701         raise ValueError("unmatched #if")
 702     return block_stack[-1]
 703
 704
 705 def check_file(
 706     filename, inclname, file_kind, code, all_inclnames, included_h_inclnames
 707 ):
 708     def check_include_statement(include):
 709         """Check the style of a single #include statement."""
 710
 711         if include.is_system:
 712             # Check it is not a known local file (in which case it's probably a system header).
 713             if (
 714                 include.inclname in included_inclnames_to_ignore
 715                 or include.inclname in all_inclnames
 716             ):
 717                 error(
 718                     filename,
 719                     include.linenum,
 720                     include.quote() + " should be included using",
 721                     'the #include "..." form',
 722                 )
 723
 724         else:
 725             msg = deprecated_inclnames.get(include.inclname)
 726             if msg:
 727                 error(
 728                     filename,
 729                     include.linenum,
 730                     include.quote() + " is deprecated: " + msg,
 731                 )
 732
 733             if file_kind == FileKind.H or file_kind == FileKind.INL_H:
 734                 msg = deprecated_inclnames_in_header.get(include.inclname)
 735                 if msg and filename not in deprecated_inclnames_in_header_excludes:
 736                     error(
 737                         filename,
 738                         include.linenum,
 739                         include.quote() + " is deprecated: " + msg,
 740                     )
 741
 742             if include.inclname not in included_inclnames_to_ignore:
 743                 included_kind = FileKind.get(include.inclname)
 744
 745                 # Check the #include path has the correct form.
 746                 if include.inclname not in all_inclnames:
 747                     error(
 748                         filename,
 749                         include.linenum,
 750                         include.quote() + " is included using the wrong path;",
 751                         "did you forget a prefix, or is the file not yet committed?",
 752                     )
 753
 754                 # Record inclusions of .h files for cycle detection later.
 755                 # (Exclude .tbl and .msg files.)
 756                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 757                     included_h_inclnames.add(include.inclname)
 758
 759                 # Check a H file doesn't #include an INL_H file.
 760                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 761                     error(
 762                         filename,
 763                         include.linenum,
 764                         "vanilla header includes an inline-header file "
 765                         + include.quote(),
 766                     )
 767
 768                 # Check a file doesn't #include itself.  (We do this here because the cycle
 769                 # detection below doesn't detect this case.)
 770                 if inclname == include.inclname:
 771                     error(filename, include.linenum, "the file includes itself")
 772
 773     def check_includes_order(include1, include2):
 774         """Check the ordering of two #include statements."""
 775
 776         if (
 777             include1.inclname in oddly_ordered_inclnames
 778             or include2.inclname in oddly_ordered_inclnames
 779         ):
 780             return
 781
 782         section1 = include1.section(inclname)
 783         section2 = include2.section(inclname)
 784         if (section1 > section2) or (
 785             (section1 == section2)
 786             and (include1.inclname.lower() > include2.inclname.lower())
 787         ):
 788             error(
 789                 filename,
 790                 str(include1.linenum) + ":" + str(include2.linenum),
 791                 include1.quote() + " should be included after " + include2.quote(),
 792             )
 793
 794     # Check the extracted #include statements, both individually, and the ordering of
 795     # adjacent pairs that live in the same block.
 796     def pair_traverse(prev, this):
 797         if isinstance(this, Include):
 798             check_include_statement(this)
 799             if isinstance(prev, Include):
 800                 check_includes_order(prev, this)
 801         else:
 802             kids = this.style_relevant_kids()
 803             for prev2, this2 in zip([None] + kids[0:-1], kids):
 804                 pair_traverse(prev2, this2)
 805
 806     pair_traverse(None, code)
 807
 808
 809 def find_cycles(all_inclnames, edges):
 810     """Find and draw any cycles."""
 811
 812     SCCs = tarjan(all_inclnames, edges)
 813
 814     # The various sorted() calls below ensure the output is deterministic.
 815
 816     def draw_SCC(c):
 817         cset = set(c)
 818         drawn = set()
 819
 820         def draw(v, indent):
 821             out("   " * indent + ("-> " if indent else "   ") + v)
 822             if v in drawn:
 823                 return
 824             drawn.add(v)
 825             for succ in sorted(edges[v]):
 826                 if succ in cset:
 827                     draw(succ, indent + 1)
 828
 829         draw(sorted(c)[0], 0)
 830         out("")
 831
 832     have_drawn_an_SCC = False
 833     for scc in sorted(SCCs):
 834         if len(scc) != 1:
 835             if not have_drawn_an_SCC:
 836                 error("(multiple files)", None, "header files form one or more cycles")
 837                 have_drawn_an_SCC = True
 838
 839             draw_SCC(scc)
 840
 841
 842 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 843 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 844 def tarjan(V, E):
 845     vertex_index = {}
 846     vertex_lowlink = {}
 847     index = 0
 848     S = []
 849     all_SCCs = []
 850
 851     def strongconnect(v, index):
 852         # Set the depth index for v to the smallest unused index
 853         vertex_index[v] = index
 854         vertex_lowlink[v] = index
 855         index += 1
 856         S.append(v)
 857
 858         # Consider successors of v
 859         for w in E[v]:
 860             if w not in vertex_index:
 861                 # Successor w has not yet been visited; recurse on it
 862                 index = strongconnect(w, index)
 863                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 864             elif w in S:
 865                 # Successor w is in stack S and hence in the current SCC
 866                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 867
 868         # If v is a root node, pop the stack and generate an SCC
 869         if vertex_lowlink[v] == vertex_index[v]:
 870             i = S.index(v)
 871             scc = S[i:]
 872             del S[i:]
 873             all_SCCs.append(scc)
 874
 875         return index
 876
 877     for v in V:
 878         if v not in vertex_index:
 879             index = strongconnect(v, index)
 880
 881     return all_SCCs
 882
 883
 884 def main():
 885     if sys.argv[1:] == ["--fixup"]:
 886         # Sort #include directives in-place.  Fixup mode doesn't solve
 887         # all possible silliness that the script checks for; it's just a
 888         # hack for the common case where renaming a header causes style
 889         # errors.
 890         fixup = True
 891     elif sys.argv[1:] == []:
 892         fixup = False
 893     else:
 894         print(
 895             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | unexpected command "
 896             "line options: " + repr(sys.argv[1:])
 897         )
 898         sys.exit(1)
 899
 900     ok = check_style(fixup)
 901
 902     if ok:
 903         print("TEST-PASS | check_spidermonkey_style.py | ok")
 904     else:
 905         print(
 906             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 907             + "actual output does not match expected output;  diff is above."
 908         )
 909         print(
 910             "TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | "
 911             + "Hint: If the problem is that you renamed a header, and many #includes "
 912             + "are no longer in alphabetical order, commit your work and then try "
 913             + "`check_spidermonkey_style.py --fixup`. "
 914             + "You need to commit first because --fixup modifies your files in place."
 915         )
 916
 917     sys.exit(0 if ok else 1)
 918
 919
 920 if __name__ == "__main__":
 921     main()