summaryrefslogtreecommitdiff
path: root/testing/tools/safetynet_compare.py
blob: 0cf1aec036e9076d9538d007f36081ed8143289f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
#!/usr/bin/env python
# Copyright 2017 The PDFium Authors. All rights reserved.
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.

"""Compares the performance of two versions of the pdfium code."""

import argparse
import functools
import json
import multiprocessing
import os
import re
import shutil
import subprocess
import sys
import tempfile

from common import GetBooleanGnArg
from common import PrintErr
from common import RunCommandPropagateErr
from githelper import GitHelper
from safetynet_conclusions import ComparisonConclusions
from safetynet_conclusions import PrintConclusionsDictHumanReadable
from safetynet_conclusions import RATING_IMPROVEMENT
from safetynet_conclusions import RATING_REGRESSION


def RunSingleTestCaseParallel(this, run_label, build_dir, test_case):
  result = this.RunSingleTestCase(run_label, build_dir, test_case)
  return (test_case, result)


class CompareRun(object):
  """A comparison between two branches of pdfium."""

  def __init__(self, args):
    self.git = GitHelper()
    self.args = args
    self._InitPaths()

  def _InitPaths(self):
    if self.args.this_repo:
      self.safe_script_dir = self.args.build_dir
    else:
      self.safe_script_dir = os.path.join('testing', 'tools')

    self.safe_measure_script_path = os.path.abspath(
        os.path.join(self.safe_script_dir,
                     'safetynet_measure.py'))

    input_file_re = re.compile('^.+[.]pdf$')
    self.test_cases = []
    for input_path in self.args.input_paths:
      if os.path.isfile(input_path):
        self.test_cases.append(input_path)
      elif os.path.isdir(input_path):
        for file_dir, _, filename_list in os.walk(input_path):
          for input_filename in filename_list:
            if input_file_re.match(input_filename):
              file_path = os.path.join(file_dir, input_filename)
              if os.path.isfile(file_path):
                self.test_cases.append(file_path)

    self.after_build_dir = self.args.build_dir
    if self.args.build_dir_before:
      self.before_build_dir = self.args.build_dir_before
    else:
      self.before_build_dir = self.after_build_dir

  def Run(self):
    """Runs comparison by checking out branches, building and measuring them.

    Returns:
      Exit code for the script.
    """
    if self.args.this_repo:
      self._FreezeMeasureScript()

    if self.args.branch_after:
      if self.args.this_repo:
        before, after = self._ProfileTwoOtherBranchesInThisRepo(
            self.args.branch_before,
            self.args.branch_after)
      else:
        before, after = self._ProfileTwoOtherBranches(
            self.args.branch_before,
            self.args.branch_after)
    elif self.args.branch_before:
      if self.args.this_repo:
        before, after = self._ProfileCurrentAndOtherBranchInThisRepo(
            self.args.branch_before)
      else:
        before, after = self._ProfileCurrentAndOtherBranch(
            self.args.branch_before)
    else:
      if self.args.this_repo:
        before, after = self._ProfileLocalChangesAndCurrentBranchInThisRepo()
      else:
        before, after = self._ProfileLocalChangesAndCurrentBranch()

    conclusions = self._DrawConclusions(before, after)
    conclusions_dict = conclusions.GetOutputDict()
    conclusions_dict.setdefault('metadata', {})['profiler'] = self.args.profiler

    self._PrintConclusions(conclusions_dict)

    self._CleanUp(conclusions)

    return 0

  def _FreezeMeasureScript(self):
    """Freezes a version of the measuring script.

    This is needed to make sure we are comparing the pdfium library changes and
    not script changes that may happen between the two branches.
    """
    self.__FreezeFile(os.path.join('testing', 'tools', 'safetynet_measure.py'))
    self.__FreezeFile(os.path.join('testing', 'tools', 'common.py'))

  def __FreezeFile(self, file):
    RunCommandPropagateErr(['cp', file, self.safe_script_dir],
                           exit_status_on_error=1)

  def _ProfileTwoOtherBranchesInThisRepo(self, before_branch, after_branch):
    """Profiles two branches that are not the current branch.

    This is done in the local repository and changes may not be restored if the
    script fails or is interrupted.

    after_branch does not need to descend from before_branch, they will be
    measured the same way

    Args:
      before_branch: One branch to profile.
      after_branch: Other branch to profile.

    Returns:
      A tuple (before, after), where each of before and after is a dict
      mapping a test case name to the profiling values for that test case
      in the given branch.
    """
    branch_to_restore = self.git.GetCurrentBranchName()

    self._StashLocalChanges()

    self._CheckoutBranch(after_branch)
    self._BuildCurrentBranch(self.after_build_dir)
    after = self._MeasureCurrentBranch('after', self.after_build_dir)

    self._CheckoutBranch(before_branch)
    self._BuildCurrentBranch(self.before_build_dir)
    before = self._MeasureCurrentBranch('before', self.before_build_dir)

    self._CheckoutBranch(branch_to_restore)
    self._RestoreLocalChanges()

    return before, after

  def _ProfileTwoOtherBranches(self, before_branch, after_branch):
    """Profiles two branches that are not the current branch.

    This is done in new, cloned repositories, therefore it is safer but slower
    and requires downloads.

    after_branch does not need to descend from before_branch, they will be
    measured the same way

    Args:
      before_branch: One branch to profile.
      after_branch: Other branch to profile.

    Returns:
      A tuple (before, after), where each of before and after is a dict
      mapping a test case name to the profiling values for that test case
      in the given branch.
    """
    after = self._ProfileSeparateRepo('after',
                                      self.after_build_dir,
                                      after_branch)
    before = self._ProfileSeparateRepo('before',
                                       self.before_build_dir,
                                       before_branch)
    return before, after

  def _ProfileCurrentAndOtherBranchInThisRepo(self, other_branch):
    """Profiles the current branch (with uncommitted changes) and another one.

    This is done in the local repository and changes may not be restored if the
    script fails or is interrupted.

    The current branch does not need to descend from other_branch.

    Args:
      other_branch: Other branch to profile that is not the current.

    Returns:
      A tuple (before, after), where each of before and after is a dict
      mapping a test case name to the profiling values for that test case
      in the given branch. The current branch is considered to be "after" and
      the other branch is considered to be "before".
    """
    branch_to_restore = self.git.GetCurrentBranchName()

    self._BuildCurrentBranch(self.after_build_dir)
    after = self._MeasureCurrentBranch('after', self.after_build_dir)

    self._StashLocalChanges()

    self._CheckoutBranch(other_branch)
    self._BuildCurrentBranch(self.before_build_dir)
    before = self._MeasureCurrentBranch('before', self.before_build_dir)

    self._CheckoutBranch(branch_to_restore)
    self._RestoreLocalChanges()

    return before, after

  def _ProfileCurrentAndOtherBranch(self, other_branch):
    """Profiles the current branch (with uncommitted changes) and another one.

    This is done in new, cloned repositories, therefore it is safer but slower
    and requires downloads.

    The current branch does not need to descend from other_branch.

    Args:
      other_branch: Other branch to profile that is not the current. None will
          compare to the same branch.

    Returns:
      A tuple (before, after), where each of before and after is a dict
      mapping a test case name to the profiling values for that test case
      in the given branch. The current branch is considered to be "after" and
      the other branch is considered to be "before".
    """
    self._BuildCurrentBranch(self.after_build_dir)
    after = self._MeasureCurrentBranch('after', self.after_build_dir)

    before = self._ProfileSeparateRepo('before',
                                       self.before_build_dir,
                                       other_branch)

    return before, after

  def _ProfileLocalChangesAndCurrentBranchInThisRepo(self):
    """Profiles the current branch with and without uncommitted changes.

    This is done in the local repository and changes may not be restored if the
    script fails or is interrupted.

    Returns:
      A tuple (before, after), where each of before and after is a dict
      mapping a test case name to the profiling values for that test case
      using the given version. The current branch without uncommitted changes is
      considered to be "before" and with uncommitted changes is considered to be
      "after".
    """
    self._BuildCurrentBranch(self.after_build_dir)
    after = self._MeasureCurrentBranch('after', self.after_build_dir)

    pushed = self._StashLocalChanges()
    if not pushed and not self.args.build_dir_before:
      PrintErr('Warning: No local changes to compare')

    before_build_dir = self.before_build_dir

    self._BuildCurrentBranch(before_build_dir)
    before = self._MeasureCurrentBranch('before', before_build_dir)

    self._RestoreLocalChanges()

    return before, after

  def _ProfileLocalChangesAndCurrentBranch(self):
    """Profiles the current branch with and without uncommitted changes.

    This is done in new, cloned repositories, therefore it is safer but slower
    and requires downloads.

    Returns:
      A tuple (before, after), where each of before and after is a dict
      mapping a test case name to the profiling values for that test case
      using the given version. The current branch without uncommitted changes is
      considered to be "before" and with uncommitted changes is considered to be
      "after".
    """
    return self._ProfileCurrentAndOtherBranch(other_branch=None)

  def _ProfileSeparateRepo(self, run_label, relative_build_dir, branch):
    """Profiles a branch in a a temporary git repository.

    Args:
      run_label: String to differentiate this version of the code in output
          files from other versions.
      relative_build_dir: Path to the build dir in the current working dir to
          clone build args from.
      branch: Branch to checkout in the new repository. None will
          profile the same branch checked out in the original repo.
    Returns:
      A dict mapping each test case name to the profiling values for that
      test case.
    """
    build_dir = self._CreateTempRepo('repo_%s' % run_label,
                                     relative_build_dir,
                                     branch)

    self._BuildCurrentBranch(build_dir)
    return self._MeasureCurrentBranch(run_label, build_dir)

  def _CreateTempRepo(self, dir_name, relative_build_dir, branch):
    """Clones a temporary git repository out of the current working dir.

    Args:
      dir_name: Name for the temporary repository directory
      relative_build_dir: Path to the build dir in the current working dir to
          clone build args from.
      branch: Branch to checkout in the new repository. None will keep checked
          out the same branch as the local repo.
    Returns:
      Path to the build directory of the new repository.
    """
    cwd = os.getcwd()

    repo_dir = tempfile.mkdtemp(suffix='-%s' % dir_name)
    src_dir = os.path.join(repo_dir, 'pdfium')

    self.git.CloneLocal(os.getcwd(), src_dir)

    if branch is not None:
      os.chdir(src_dir)
      self.git.Checkout(branch)

    os.chdir(repo_dir)
    PrintErr('Syncing...')

    cmd = ['gclient', 'config', '--unmanaged',
           'https://pdfium.googlesource.com/pdfium.git']
    if self.args.cache_dir:
      cmd.append('--cache-dir=%s' % self.args.cache_dir)
    RunCommandPropagateErr(cmd, exit_status_on_error=1)

    RunCommandPropagateErr(['gclient', 'sync'], exit_status_on_error=1)

    PrintErr('Done.')

    build_dir = os.path.join(src_dir, relative_build_dir)
    os.makedirs(build_dir)
    os.chdir(src_dir)

    source_gn_args = os.path.join(cwd, relative_build_dir, 'args.gn')
    dest_gn_args = os.path.join(build_dir, 'args.gn')
    shutil.copy(source_gn_args, dest_gn_args)

    RunCommandPropagateErr(['gn', 'gen', relative_build_dir],
                           exit_status_on_error=1)

    os.chdir(cwd)

    return build_dir


  def _CheckoutBranch(self, branch):
    PrintErr("Checking out branch '%s'" % branch)
    self.git.Checkout(branch)

  def _StashLocalChanges(self):
    PrintErr('Stashing local changes')
    return self.git.StashPush()

  def _RestoreLocalChanges(self):
    PrintErr('Restoring local changes')
    self.git.StashPopAll()

  def _BuildCurrentBranch(self, build_dir):
    """Synchronizes and builds the current version of pdfium.

    Args:
      build_dir: String with path to build directory
    """
    PrintErr('Syncing...')
    RunCommandPropagateErr(['gclient', 'sync'], exit_status_on_error=1)
    PrintErr('Done.')

    PrintErr('Building...')
    cmd = ['ninja', '-C', build_dir, 'pdfium_test']
    if GetBooleanGnArg('use_goma', build_dir):
      cmd.extend(['-j', '250'])
    RunCommandPropagateErr(cmd, stdout_has_errors=True, exit_status_on_error=1)
    PrintErr('Done.')

  def _MeasureCurrentBranch(self, run_label, build_dir):
    PrintErr('Measuring...')
    if self.args.num_workers > 1 and len(self.test_cases) > 1:
      results = self._RunAsync(run_label, build_dir)
    else:
      results = self._RunSync(run_label, build_dir)
    PrintErr('Done.')

    return results

  def _RunSync(self, run_label, build_dir):
    """Profiles the test cases synchronously.

    Args:
      run_label: String to differentiate this version of the code in output
          files from other versions.
      build_dir: String with path to build directory

    Returns:
      A dict mapping each test case name to the profiling values for that
      test case.
    """
    results = {}

    for test_case in self.test_cases:
      result = self.RunSingleTestCase(run_label, build_dir, test_case)
      if result is not None:
        results[test_case] = result

    return results

  def _RunAsync(self, run_label, build_dir):
    """Profiles the test cases asynchronously.

    Uses as many workers as configured by --num-workers.

    Args:
      run_label: String to differentiate this version of the code in output
          files from other versions.
      build_dir: String with path to build directory

    Returns:
      A dict mapping each test case name to the profiling values for that
      test case.
    """
    results = {}
    pool = multiprocessing.Pool(self.args.num_workers)
    worker_func = functools.partial(
        RunSingleTestCaseParallel, self, run_label, build_dir)

    try:
      # The timeout is a workaround for http://bugs.python.org/issue8296
      # which prevents KeyboardInterrupt from working.
      one_year_in_seconds = 3600 * 24 * 365
      worker_results = (pool.map_async(worker_func, self.test_cases)
                        .get(one_year_in_seconds))
      for worker_result in worker_results:
        test_case, result = worker_result
        if result is not None:
          results[test_case] = result
    except KeyboardInterrupt:
      pool.terminate()
      sys.exit(1)
    else:
      pool.close()

    pool.join()

    return results

  def RunSingleTestCase(self, run_label, build_dir, test_case):
    """Profiles a single test case.

    Args:
      run_label: String to differentiate this version of the code in output
          files from other versions.
      build_dir: String with path to build directory
      test_case: Path to the test case.

    Returns:
      The measured profiling value for that test case.
    """
    command = [self.safe_measure_script_path, test_case,
               '--build-dir=%s' % build_dir]

    if self.args.interesting_section:
      command.append('--interesting-section')

    if self.args.profiler:
      command.append('--profiler=%s' % self.args.profiler)

    profile_file_path = self._GetProfileFilePath(run_label, test_case)
    if profile_file_path:
      command.append('--output-path=%s' % profile_file_path)

    output = RunCommandPropagateErr(command)

    if output is None:
      return None

    # Get the time number as output, making sure it's just a number
    output = output.strip()
    if re.match('^[0-9]+$', output):
      return int(output)

    return None

  def _GetProfileFilePath(self, run_label, test_case):
    if self.args.output_dir:
      output_filename = ('callgrind.out.%s.%s'
                         % (test_case.replace('/', '_'),
                            run_label))
      return os.path.join(self.args.output_dir, output_filename)
    else:
      return None

  def _DrawConclusions(self, times_before_branch, times_after_branch):
    """Draws conclusions comparing results of test runs in two branches.

    Args:
      times_before_branch: A dict mapping each test case name to the
          profiling values for that test case in the branch to be considered
          as the baseline.
      times_after_branch: A dict mapping each test case name to the
          profiling values for that test case in the branch to be considered
          as the new version.

    Returns:
      ComparisonConclusions with all test cases processed.
    """
    conclusions = ComparisonConclusions(self.args.threshold_significant)

    for test_case in sorted(self.test_cases):
      before = times_before_branch.get(test_case)
      after = times_after_branch.get(test_case)
      conclusions.ProcessCase(test_case, before, after)

    return conclusions

  def _PrintConclusions(self, conclusions_dict):
    """Prints the conclusions as the script output.

    Depending on the script args, this can output a human or a machine-readable
    version of the conclusions.

    Args:
      conclusions_dict: Dict to print returned from
          ComparisonConclusions.GetOutputDict().
    """
    if self.args.machine_readable:
      print json.dumps(conclusions_dict)
    else:
      PrintConclusionsDictHumanReadable(
          conclusions_dict, colored=True, key=self.args.case_order)

  def _CleanUp(self, conclusions):
    """Removes profile output files for uninteresting cases.

    Cases without significant regressions or improvements and considered
    uninteresting.

    Args:
      conclusions: A ComparisonConclusions.
    """
    if not self.args.output_dir:
      return

    if self.args.profiler != 'callgrind':
      return

    for case_result in conclusions.GetCaseResults().values():
      if case_result.rating not in [RATING_REGRESSION, RATING_IMPROVEMENT]:
        self._CleanUpOutputFile('before', case_result.case_name)
        self._CleanUpOutputFile('after', case_result.case_name)

  def _CleanUpOutputFile(self, run_label, case_name):
    """Removes one profile output file.

    If the output file does not exist, fails silently.

    Args:
      run_label: String to differentiate a version of the code in output
          files from other versions.
      case_name: String identifying test case for which to remove the output
          file.
    """
    try:
      os.remove(self._GetProfileFilePath(run_label, case_name))
    except OSError:
      pass


def main():
  parser = argparse.ArgumentParser()
  parser.add_argument('input_paths', nargs='+',
                      help='pdf files or directories to search for pdf files '
                           'to run as test cases')
  parser.add_argument('--branch-before',
                      help='git branch to use as "before" for comparison. '
                           'Omitting this will use the current branch '
                           'without uncommitted changes as the baseline.')
  parser.add_argument('--branch-after',
                      help='git branch to use as "after" for comparison. '
                           'Omitting this will use the current branch '
                           'with uncommitted changes.')
  parser.add_argument('--build-dir', default=os.path.join('out', 'Release'),
                      help='relative path from the base source directory '
                           'to the build directory')
  parser.add_argument('--build-dir-before',
                      help='relative path from the base source directory '
                           'to the build directory for the "before" branch, if '
                           'different from the build directory for the '
                           '"after" branch')
  parser.add_argument('--cache-dir', default=None,
                      help='directory with a new or preexisting cache for '
                           'downloads. Default is to not use a cache.')
  parser.add_argument('--this-repo', action='store_true',
                      help='use the repository where the script is instead of '
                           'checking out a temporary one. This is faster and '
                           'does not require downloads, but although it '
                           'restores the state of the local repo, if the '
                           'script is killed or crashes the changes can remain '
                           'stashed and you may be on another branch.')
  parser.add_argument('--profiler', default='callgrind',
                      help='which profiler to use. Supports callgrind and '
                           'perfstat for now. Default is callgrind.')
  parser.add_argument('--interesting-section', action='store_true',
                      help='whether to measure just the interesting section or '
                           'the whole test harness. Limiting to only the '
                           'interesting section does not work on Release since '
                           'the delimiters are optimized out')
  parser.add_argument('--num-workers', default=multiprocessing.cpu_count(),
                      type=int, help='run NUM_WORKERS jobs in parallel')
  parser.add_argument('--output-dir',
                      help='directory to write the profile data output files')
  parser.add_argument('--threshold-significant', default=0.02, type=float,
                      help='variations in performance above this factor are '
                           'considered significant')
  parser.add_argument('--machine-readable', action='store_true',
                      help='whether to get output for machines. If enabled the '
                           'output will be a json with the format specified in '
                           'ComparisonConclusions.GetOutputDict(). Default is '
                           'human-readable.')
  parser.add_argument('--case-order', default=None,
                      help='what key to use when sorting test cases in the '
                           'output. Accepted values are "after", "before", '
                           '"ratio" and "rating". Default is sorting by test '
                           'case path.')

  args = parser.parse_args()

  # Always start at the pdfium src dir, which is assumed to be two level above
  # this script.
  pdfium_src_dir = os.path.join(
      os.path.dirname(__file__),
      os.path.pardir,
      os.path.pardir)
  os.chdir(pdfium_src_dir)

  git = GitHelper()

  if args.branch_after and not args.branch_before:
    PrintErr('--branch-after requires --branch-before to be specified.')
    return 1

  if args.branch_after and not git.BranchExists(args.branch_after):
    PrintErr('Branch "%s" does not exist' % args.branch_after)
    return 1

  if args.branch_before and not git.BranchExists(args.branch_before):
    PrintErr('Branch "%s" does not exist' % args.branch_before)
    return 1

  if args.output_dir:
    args.output_dir = os.path.expanduser(args.output_dir)
    if not os.path.isdir(args.output_dir):
      PrintErr('"%s" is not a directory' % args.output_dir)
      return 1

  if args.threshold_significant <= 0.0:
    PrintErr('--threshold-significant should receive a positive float')
    return 1

  run = CompareRun(args)
  return run.Run()


if __name__ == '__main__':
  sys.exit(main())