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