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 print_function
  39
  40 import difflib
  41 import os
  42 import re
  43 import subprocess
  44 import sys
  45 import traceback
  46
  47 # We don't bother checking files in these directories, because they're (a) auxiliary or (b)
  48 # imported code that doesn't follow our coding style.
  49 ignored_js_src_dirs = [
  50    'js/src/config/',            # auxiliary stuff
  51    'js/src/ctypes/libffi/',     # imported code
  52    'js/src/devtools/',          # auxiliary stuff
  53    'js/src/editline/',          # imported code
  54    'js/src/gdb/',               # auxiliary stuff
  55    'js/src/vtune/'              # imported code
  56 ]
  57
  58 # We ignore #includes of these files, because they don't follow the usual rules.
  59 included_inclnames_to_ignore = set([
  60     'ffi.h',                    # generated in ctypes/libffi/
  61     'devtools/sharkctl.h',      # we ignore devtools/ in general
  62     'devtools/Instruments.h',   # we ignore devtools/ in general
  63     'double-conversion.h',      # strange MFBT case
  64     'javascript-trace.h',       # generated in $OBJDIR if HAVE_DTRACE is defined
  65     'jsautokw.h',               # generated in $OBJDIR
  66     'jsautooplen.h',            # generated in $OBJDIR
  67     'jscustomallocator.h',      # provided by embedders;  allowed to be missing
  68     'js-config.h',              # generated in $OBJDIR
  69     'pratom.h',                 # NSPR
  70     'prcvar.h',                 # NSPR
  71     'prinit.h',                 # NSPR
  72     'prlink.h',                 # NSPR
  73     'prlock.h',                 # NSPR
  74     'prprf.h',                  # NSPR
  75     'prthread.h',               # NSPR
  76     'prtypes.h',                # NSPR
  77     'selfhosted.out.h',         # generated in $OBJDIR
  78     'unicode/locid.h',          # ICU
  79     'unicode/numsys.h',         # ICU
  80     'unicode/ucal.h',           # ICU
  81     'unicode/uclean.h',         # ICU
  82     'unicode/ucol.h',           # ICU
  83     'unicode/udat.h',           # ICU
  84     'unicode/udatpg.h',         # ICU
  85     'unicode/uenum.h',          # ICU
  86     'unicode/unum.h',           # ICU
  87     'unicode/ustring.h',        # ICU
  88     'unicode/utypes.h',         # ICU
  89     'vtune/VTuneWrapper.h'      # VTune
  90 ])
  91
  92 # These files have additional constraints on where they are #included, so we
  93 # ignore #includes of them when checking #include ordering.
  94 oddly_ordered_inclnames = set([
  95     'ctypes/typedefs.h',        # Included multiple times in the body of ctypes/CTypes.h
  96     'jsautokw.h',               # Included in the body of frontend/TokenStream.h
  97     'jswin.h',                  # Must be #included before <psapi.h>
  98     'machine/endian.h',         # Must be included after <sys/types.h> on BSD
  99     'winbase.h',                # Must precede other system headers(?)
 100     'windef.h'                  # Must precede other system headers(?)
 101 ])
 102
 103 # The files in tests/style/ contain code that fails this checking in various
 104 # ways.  Here is the output we expect.  If the actual output differs from
 105 # this, one of the following must have happened.
 106 # - New SpiderMonkey code violates one of the checked rules.
 107 # - The tests/style/ files have changed without expected_output being changed
 108 #   accordingly.
 109 # - This script has been broken somehow.
 110 #
 111 expected_output = '''\
 112 js/src/tests/style/BadIncludes2.h:1: error:
 113     vanilla header includes an inline-header file "tests/style/BadIncludes2-inl.h"
 114
 115 js/src/tests/style/BadIncludes.h:3: error:
 116     the file includes itself
 117
 118 js/src/tests/style/BadIncludes.h:6: error:
 119     "BadIncludes2.h" is included using the wrong path;
 120     did you forget a prefix, or is the file not yet committed?
 121
 122 js/src/tests/style/BadIncludes.h:8: error:
 123     <tests/style/BadIncludes2.h> should be included using
 124     the #include "..." form
 125
 126 js/src/tests/style/BadIncludes.h:10: error:
 127     "stdio.h" is included using the wrong path;
 128     did you forget a prefix, or is the file not yet committed?
 129
 130 js/src/tests/style/BadIncludesOrder-inl.h:5:6: error:
 131     "vm/Interpreter-inl.h" should be included after "jsscriptinlines.h"
 132
 133 js/src/tests/style/BadIncludesOrder-inl.h:6:7: error:
 134     "jsscriptinlines.h" should be included after "js/Value.h"
 135
 136 js/src/tests/style/BadIncludesOrder-inl.h:7:8: error:
 137     "js/Value.h" should be included after "ds/LifoAlloc.h"
 138
 139 js/src/tests/style/BadIncludesOrder-inl.h:8:9: error:
 140     "ds/LifoAlloc.h" should be included after "jsapi.h"
 141
 142 js/src/tests/style/BadIncludesOrder-inl.h:9:10: error:
 143     "jsapi.h" should be included after <stdio.h>
 144
 145 js/src/tests/style/BadIncludesOrder-inl.h:10:11: error:
 146     <stdio.h> should be included after "mozilla/HashFunctions.h"
 147
 148 js/src/tests/style/BadIncludesOrder-inl.h:27:28: error:
 149     "jsobj.h" should be included after "jsfun.h"
 150
 151 (multiple files): error:
 152     header files form one or more cycles
 153
 154    tests/style/HeaderCycleA1.h
 155    -> tests/style/HeaderCycleA2.h
 156       -> tests/style/HeaderCycleA3.h
 157          -> tests/style/HeaderCycleA1.h
 158
 159    tests/style/HeaderCycleB1-inl.h
 160    -> tests/style/HeaderCycleB2-inl.h
 161       -> tests/style/HeaderCycleB3-inl.h
 162          -> tests/style/HeaderCycleB4-inl.h
 163             -> tests/style/HeaderCycleB1-inl.h
 164             -> tests/style/jsheadercycleB5inlines.h
 165                -> tests/style/HeaderCycleB1-inl.h
 166       -> tests/style/HeaderCycleB4-inl.h
 167
 168 '''.splitlines(True)
 169
 170 actual_output = []
 171
 172
 173 def out(*lines):
 174     for line in lines:
 175         actual_output.append(line + '\n')
 176
 177
 178 def error(filename, linenum, *lines):
 179     location = filename
 180     if linenum is not None:
 181         location += ':' + str(linenum)
 182     out(location + ': error:')
 183     for line in (lines):
 184         out('    ' + line)
 185     out('')
 186
 187
 188 class FileKind(object):
 189     C = 1
 190     CPP = 2
 191     INL_H = 3
 192     H = 4
 193     TBL = 5
 194     MSG = 6
 195
 196     @staticmethod
 197     def get(filename):
 198         if filename.endswith('.c'):
 199             return FileKind.C
 200
 201         if filename.endswith('.cpp'):
 202             return FileKind.CPP
 203
 204         if filename.endswith(('inlines.h', '-inl.h', 'Inlines.h')):
 205             return FileKind.INL_H
 206
 207         if filename.endswith('.h'):
 208             return FileKind.H
 209
 210         if filename.endswith('.tbl'):
 211             return FileKind.TBL
 212
 213         if filename.endswith('.msg'):
 214             return FileKind.MSG
 215
 216         error(filename, None, 'unknown file kind')
 217
 218
 219 def get_all_filenames():
 220     '''Get a list of all the files in the (Mercurial or Git) repository.'''
 221     cmds = [['hg', 'manifest', '-q'], ['git', 'ls-files', '--full-name', '../..']]
 222     for cmd in cmds:
 223         try:
 224             all_filenames = subprocess.check_output(cmd, universal_newlines=True,
 225                                                     stderr=subprocess.PIPE).split('\n')
 226             return all_filenames
 227         except:
 228             continue
 229     else:
 230         raise Exception('failed to run any of the repo manifest commands', cmds)
 231
 232
 233 def check_style():
 234     # We deal with two kinds of name.
 235     # - A "filename" is a full path to a file from the repository root.
 236     # - An "inclname" is how a file is referred to in a #include statement.
 237     #
 238     # Examples (filename -> inclname)
 239     # - "mfbt/Attributes.h"  -> "mozilla/Attributes.h"
 240     # - "js/public/Vector.h" -> "js/Vector.h"
 241     # - "js/src/vm/String.h" -> "vm/String.h"
 242
 243     mfbt_inclnames = set()      # type: set(inclname)
 244     js_names = dict()           # type: dict(filename, inclname)
 245
 246     # Select the appropriate files.
 247     for filename in get_all_filenames():
 248         if filename.startswith('mfbt/') and filename.endswith('.h'):
 249             inclname = 'mozilla/' + filename[len('mfbt/'):]
 250             mfbt_inclnames.add(inclname)
 251
 252         if filename.startswith('js/public/') and filename.endswith('.h'):
 253             inclname = 'js/' + filename[len('js/public/'):]
 254             js_names[filename] = inclname
 255
 256         if filename.startswith('js/src/') and \
 257            not filename.startswith(tuple(ignored_js_src_dirs)) and \
 258            filename.endswith(('.c', '.cpp', '.h', '.tbl', '.msg')):
 259             inclname = filename[len('js/src/'):]
 260             js_names[filename] = inclname
 261
 262     all_inclnames = mfbt_inclnames | set(js_names.values())
 263
 264     edges = dict()      # type: dict(inclname, set(inclname))
 265
 266     # We don't care what's inside the MFBT files, but because they are
 267     # #included from JS files we have to add them to the inclusion graph.
 268     for inclname in mfbt_inclnames:
 269         edges[inclname] = set()
 270
 271     # Process all the JS files.
 272     for filename in js_names.keys():
 273         inclname = js_names[filename]
 274         file_kind = FileKind.get(filename)
 275         if file_kind == FileKind.C or file_kind == FileKind.CPP or \
 276            file_kind == FileKind.H or file_kind == FileKind.INL_H:
 277             included_h_inclnames = set()    # type: set(inclname)
 278
 279             # This script is run in js/src/, so prepend '../../' to get to the root of the Mozilla
 280             # source tree.
 281             with open(os.path.join('../..', filename)) as f:
 282                 do_file(filename, inclname, file_kind, f, all_inclnames, included_h_inclnames)
 283
 284         edges[inclname] = included_h_inclnames
 285
 286     find_cycles(all_inclnames, edges)
 287
 288     # Compare expected and actual output.
 289     difflines = difflib.unified_diff(expected_output, actual_output,
 290                                      fromfile='check_spider_monkey_style.py expected output',
 291                                        tofile='check_spider_monkey_style.py actual output')
 292     ok = True
 293     for diffline in difflines:
 294         ok = False
 295         print(diffline, end='')
 296
 297     return ok
 298
 299
 300 def module_name(name):
 301     '''Strip the trailing .cpp, .h, inlines.h or -inl.h from a filename.'''
 302
 303     return name.replace('inlines.h', '').replace('-inl.h', '').replace('.h', '').replace('.cpp', '')
 304
 305
 306 def is_module_header(enclosing_inclname, header_inclname):
 307     '''Determine if an included name is the "module header", i.e. should be
 308     first in the file.'''
 309
 310     module = module_name(enclosing_inclname)
 311
 312     # Normal case, e.g. module == "foo/Bar", header_inclname == "foo/Bar.h".
 313     if module == module_name(header_inclname):
 314         return True
 315
 316     # A public header, e.g. module == "foo/Bar", header_inclname == "js/Bar.h".
 317     m = re.match(r'js\/(.*)\.h', header_inclname)
 318     if m is not None and module.endswith('/' + m.group(1)):
 319         return True
 320
 321     return False
 322
 323
 324 class Include(object):
 325     '''Important information for a single #include statement.'''
 326
 327     def __init__(self, inclname, linenum, is_system):
 328         self.inclname = inclname
 329         self.linenum = linenum
 330         self.is_system = is_system
 331
 332     def isLeaf(self):
 333         return True
 334
 335     def section(self, enclosing_inclname):
 336         '''Identify which section inclname belongs to.
 337
 338         The section numbers are as follows.
 339           0. Module header (e.g. jsfoo.h or jsfooinlines.h within jsfoo.cpp)
 340           1. mozilla/Foo.h
 341           2. <foo.h> or <foo>
 342           3. jsfoo.h, prmjtime.h, etc
 343           4. foo/Bar.h
 344           5. jsfooinlines.h
 345           6. foo/Bar-inl.h
 346           7. non-.h, e.g. *.tbl, *.msg
 347         '''
 348
 349         if self.is_system:
 350             return 2
 351
 352         if not self.inclname.endswith('.h'):
 353             return 7
 354
 355         # A couple of modules have the .h file in js/ and the .cpp file elsewhere and so need
 356         # special handling.
 357         if is_module_header(enclosing_inclname, self.inclname):
 358             return 0
 359
 360         if '/' in self.inclname:
 361             if self.inclname.startswith('mozilla/'):
 362                 return 1
 363
 364             if self.inclname.endswith('-inl.h'):
 365                 return 6
 366
 367             return 4
 368
 369         if self.inclname.endswith('inlines.h'):
 370             return 5
 371
 372         return 3
 373
 374     def quote(self):
 375         if self.is_system:
 376             return '<' + self.inclname + '>'
 377         else:
 378             return '"' + self.inclname + '"'
 379
 380
 381 class HashIfBlock(object):
 382     '''Important information about a #if/#endif block.
 383
 384     A #if/#endif block is the contents of a #if/#endif (or similar) section.
 385     The top-level block, which is not within a #if/#endif pair, is also
 386     considered a block.
 387
 388     Each leaf is either an Include (representing a #include), or another
 389     nested HashIfBlock.'''
 390     def __init__(self):
 391         self.kids = []
 392
 393     def isLeaf(self):
 394         return False
 395
 396
 397 def do_file(filename, inclname, file_kind, f, all_inclnames, included_h_inclnames):
 398     block_stack = [HashIfBlock()]
 399
 400     # Extract the #include statements as a tree of IBlocks and IIncludes.
 401     for linenum, line in enumerate(f, start=1):
 402         # Look for a |#include "..."| line.
 403         m = re.match(r'\s*#\s*include\s+"([^"]*)"', line)
 404         if m is not None:
 405             block_stack[-1].kids.append(Include(m.group(1), linenum, False))
 406
 407         # Look for a |#include <...>| line.
 408         m = re.match(r'\s*#\s*include\s+<([^>]*)>', line)
 409         if m is not None:
 410             block_stack[-1].kids.append(Include(m.group(1), linenum, True))
 411
 412         # Look for a |#{if,ifdef,ifndef}| line.
 413         m = re.match(r'\s*#\s*(if|ifdef|ifndef)\b', line)
 414         if m is not None:
 415             # Open a new block.
 416             new_block = HashIfBlock()
 417             block_stack[-1].kids.append(new_block)
 418             block_stack.append(new_block)
 419
 420         # Look for a |#{elif,else}| line.
 421         m = re.match(r'\s*#\s*(elif|else)\b', line)
 422         if m is not None:
 423             # Close the current block, and open an adjacent one.
 424             block_stack.pop()
 425             new_block = HashIfBlock()
 426             block_stack[-1].kids.append(new_block)
 427             block_stack.append(new_block)
 428
 429         # Look for a |#endif| line.
 430         m = re.match(r'\s*#\s*endif\b', line)
 431         if m is not None:
 432             # Close the current block.
 433             block_stack.pop()
 434
 435     def check_include_statement(include):
 436         '''Check the style of a single #include statement.'''
 437
 438         if include.is_system:
 439             # Check it is not a known local file (in which case it's probably a system header).
 440             if include.inclname in included_inclnames_to_ignore or \
 441                include.inclname in all_inclnames:
 442                 error(filename, include.linenum,
 443                       include.quote() + ' should be included using',
 444                       'the #include "..." form')
 445
 446         else:
 447             if include.inclname not in included_inclnames_to_ignore:
 448                 included_kind = FileKind.get(include.inclname)
 449
 450                 # Check the #include path has the correct form.
 451                 if include.inclname not in all_inclnames:
 452                     error(filename, include.linenum,
 453                           include.quote() + ' is included ' + 'using the wrong path;',
 454                           'did you forget a prefix, or is the file not yet committed?')
 455
 456                 # Record inclusions of .h files for cycle detection later.
 457                 # (Exclude .tbl and .msg files.)
 458                 elif included_kind == FileKind.H or included_kind == FileKind.INL_H:
 459                     included_h_inclnames.add(include.inclname)
 460
 461                 # Check a H file doesn't #include an INL_H file.
 462                 if file_kind == FileKind.H and included_kind == FileKind.INL_H:
 463                     error(filename, include.linenum,
 464                           'vanilla header includes an inline-header file ' + include.quote())
 465
 466                 # Check a file doesn't #include itself.  (We do this here because the cycle
 467                 # detection below doesn't detect this case.)
 468                 if inclname == include.inclname:
 469                     error(filename, include.linenum, 'the file includes itself')
 470
 471     def check_includes_order(include1, include2):
 472         '''Check the ordering of two #include statements.'''
 473
 474         if include1.inclname in oddly_ordered_inclnames or \
 475            include2.inclname in oddly_ordered_inclnames:
 476             return
 477
 478         section1 = include1.section(inclname)
 479         section2 = include2.section(inclname)
 480         if (section1 > section2) or \
 481            ((section1 == section2) and (include1.inclname.lower() > include2.inclname.lower())):
 482             error(filename, str(include1.linenum) + ':' + str(include2.linenum),
 483                   include1.quote() + ' should be included after ' + include2.quote())
 484
 485     # The #include statements in the files in assembler/ and yarr/ have all manner of implicit
 486     # ordering requirements.  Boo.  Ignore them.
 487     skip_order_checking = inclname.startswith(('assembler/', 'yarr/'))
 488
 489     # Check the extracted #include statements, both individually, and the ordering of
 490     # adjacent pairs that live in the same block.
 491     def pair_traverse(prev, this):
 492         if this.isLeaf():
 493             check_include_statement(this)
 494             if prev is not None and prev.isLeaf() and not skip_order_checking:
 495                 check_includes_order(prev, this)
 496         else:
 497             for prev2, this2 in zip([None] + this.kids[0:-1], this.kids):
 498                 pair_traverse(prev2, this2)
 499
 500     pair_traverse(None, block_stack[-1])
 501
 502
 503 def find_cycles(all_inclnames, edges):
 504     '''Find and draw any cycles.'''
 505
 506     SCCs = tarjan(all_inclnames, edges)
 507
 508     # The various sorted() calls below ensure the output is deterministic.
 509
 510     def draw_SCC(c):
 511         cset = set(c)
 512         drawn = set()
 513         def draw(v, indent):
 514             out('   ' * indent + ('-> ' if indent else '   ') + v)
 515             if v in drawn:
 516                 return
 517             drawn.add(v)
 518             for succ in sorted(edges[v]):
 519                 if succ in cset:
 520                     draw(succ, indent + 1)
 521         draw(sorted(c)[0], 0)
 522         out('')
 523
 524     have_drawn_an_SCC = False
 525     for scc in sorted(SCCs):
 526         if len(scc) != 1:
 527             if not have_drawn_an_SCC:
 528                 error('(multiple files)', None, 'header files form one or more cycles')
 529                 have_drawn_an_SCC = True
 530
 531             draw_SCC(scc)
 532
 533
 534 # Tarjan's algorithm for finding the strongly connected components (SCCs) of a graph.
 535 # https://en.wikipedia.org/wiki/Tarjan%27s_strongly_connected_components_algorithm
 536 def tarjan(V, E):
 537     vertex_index = {}
 538     vertex_lowlink = {}
 539     index = 0
 540     S = []
 541     all_SCCs = []
 542
 543     def strongconnect(v, index):
 544         # Set the depth index for v to the smallest unused index
 545         vertex_index[v] = index
 546         vertex_lowlink[v] = index
 547         index += 1
 548         S.append(v)
 549
 550         # Consider successors of v
 551         for w in E[v]:
 552             if w not in vertex_index:
 553                 # Successor w has not yet been visited; recurse on it
 554                 index = strongconnect(w, index)
 555                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_lowlink[w])
 556             elif w in S:
 557                 # Successor w is in stack S and hence in the current SCC
 558                 vertex_lowlink[v] = min(vertex_lowlink[v], vertex_index[w])
 559
 560         # If v is a root node, pop the stack and generate an SCC
 561         if vertex_lowlink[v] == vertex_index[v]:
 562             i = S.index(v)
 563             scc = S[i:]
 564             del S[i:]
 565             all_SCCs.append(scc)
 566
 567         return index
 568
 569     for v in V:
 570         if v not in vertex_index:
 571             index = strongconnect(v, index)
 572
 573     return all_SCCs
 574
 575
 576 def main():
 577     ok = check_style()
 578
 579     if ok:
 580         print('TEST-PASS | check_spidermonkey_style.py | ok')
 581     else:
 582         print('TEST-UNEXPECTED-FAIL | check_spidermonkey_style.py | actual output does not match expected output;  diff is above')
 583
 584     sys.exit(0 if ok else 1)
 585
 586
 587 if __name__ == '__main__':
 588     main()