contrib/dg-extract-results.py

   1 #!/usr/bin/python
   2 #
   3 # Copyright (C) 2014 Free Software Foundation, Inc.
   4 #
   5 # This script is free software; you can redistribute it and/or modify
   6 # it under the terms of the GNU General Public License as published by
   7 # the Free Software Foundation; either version 3, or (at your option)
   8 # any later version.
   9
  10 import sys
  11 import getopt
  12 import re
  13 import io
  14 from datetime import datetime
  15 from operator import attrgetter
  16
  17 # True if unrecognised lines should cause a fatal error.  Might want to turn
  18 # this on by default later.
  19 strict = False
  20
  21 # True if the order of .log segments should match the .sum file, false if
  22 # they should keep the original order.
  23 sort_logs = True
  24
  25 # A version of open() that is safe against whatever binary output
  26 # might be added to the log.
  27 def safe_open (filename):
  28     if sys.version_info >= (3, 0):
  29         return open (filename, 'r', errors = 'surrogateescape')
  30     return open (filename, 'r')
  31
  32 # Force stdout to handle escape sequences from a safe_open file.
  33 if sys.version_info >= (3, 0):
  34     sys.stdout = io.TextIOWrapper (sys.stdout.buffer,
  35                                    errors = 'surrogateescape')
  36
  37 class Named:
  38     def __init__ (self, name):
  39         self.name = name
  40
  41 class ToolRun (Named):
  42     def __init__ (self, name):
  43         Named.__init__ (self, name)
  44         # The variations run for this tool, mapped by --target_board name.
  45         self.variations = dict()
  46
  47     # Return the VariationRun for variation NAME.
  48     def get_variation (self, name):
  49         if name not in self.variations:
  50             self.variations[name] = VariationRun (name)
  51         return self.variations[name]
  52
  53 class VariationRun (Named):
  54     def __init__ (self, name):
  55         Named.__init__ (self, name)
  56         # A segment of text before the harness runs start, describing which
  57         # baseboard files were loaded for the target.
  58         self.header = None
  59         # The harnesses run for this variation, mapped by filename.
  60         self.harnesses = dict()
  61         # A list giving the number of times each type of result has
  62         # been seen.
  63         self.counts = []
  64
  65     # Return the HarnessRun for harness NAME.
  66     def get_harness (self, name):
  67         if name not in self.harnesses:
  68             self.harnesses[name] = HarnessRun (name)
  69         return self.harnesses[name]
  70
  71 class HarnessRun (Named):
  72     def __init__ (self, name):
  73         Named.__init__ (self, name)
  74         # Segments of text that make up the harness run, mapped by a test-based
  75         # key that can be used to order them.
  76         self.segments = dict()
  77         # Segments of text that make up the harness run but which have
  78         # no recognized test results.  These are typically harnesses that
  79         # are completely skipped for the target.
  80         self.empty = []
  81         # A list of results.  Each entry is a pair in which the first element
  82         # is a unique sorting key and in which the second is the full
  83         # PASS/FAIL line.
  84         self.results = []
  85
  86     # Add a segment of text to the harness run.  If the segment includes
  87     # test results, KEY is an example of one of them, and can be used to
  88     # combine the individual segments in order.  If the segment has no
  89     # test results (e.g. because the harness doesn't do anything for the
  90     # current configuration) then KEY is None instead.  In that case
  91     # just collect the segments in the order that we see them.
  92     def add_segment (self, key, segment):
  93         if key:
  94             assert key not in self.segments
  95             self.segments[key] = segment
  96         else:
  97             self.empty.append (segment)
  98
  99 class Segment:
 100     def __init__ (self, filename, start):
 101         self.filename = filename
 102         self.start = start
 103         self.lines = 0
 104
 105 class Prog:
 106     def __init__ (self):
 107         # The variations specified on the command line.
 108         self.variations = []
 109         # The variations seen in the input files.
 110         self.known_variations = set()
 111         # The tools specified on the command line.
 112         self.tools = []
 113         # Whether to create .sum rather than .log output.
 114         self.do_sum = True
 115         # Regexps used while parsing.
 116         self.test_run_re = re.compile (r'^Test Run By (\S+) on (.*)$')
 117         self.tool_re = re.compile (r'^\t\t=== (.*) tests ===$')
 118         self.result_re = re.compile (r'^(PASS|XPASS|FAIL|XFAIL|UNRESOLVED'
 119                                      r'|WARNING|ERROR|UNSUPPORTED|UNTESTED'
 120                                      r'|KFAIL):\s*(.+)')
 121         self.completed_re = re.compile (r'.* completed at (.*)')
 122         # Pieces of text to write at the head of the output.
 123         # start_line is a pair in which the first element is a datetime
 124         # and in which the second is the associated 'Test Run By' line.
 125         self.start_line = None
 126         self.native_line = ''
 127         self.target_line = ''
 128         self.host_line = ''
 129         self.acats_premable = ''
 130         # Pieces of text to write at the end of the output.
 131         # end_line is like start_line but for the 'runtest completed' line.
 132         self.acats_failures = []
 133         self.version_output = ''
 134         self.end_line = None
 135         # Known summary types.
 136         self.count_names = [
 137             '# of expected passes\t\t',
 138             '# of unexpected failures\t',
 139             '# of unexpected successes\t',
 140             '# of expected failures\t\t',
 141             '# of unknown successes\t\t',
 142             '# of known failures\t\t',
 143             '# of untested testcases\t\t',
 144             '# of unresolved testcases\t',
 145             '# of unsupported tests\t\t'
 146         ]
 147         self.runs = dict()
 148
 149     def usage (self):
 150         name = sys.argv[0]
 151         sys.stderr.write ('Usage: ' + name
 152                           + ''' [-t tool] [-l variant-list] [-L] log-or-sum-file ...
 153
 154     tool           The tool (e.g. g++, libffi) for which to create a
 155                    new test summary file.  If not specified then output
 156                    is created for all tools.
 157     variant-list   One or more test variant names.  If the list is
 158                    not specified then one is constructed from all
 159                    variants in the files for <tool>.
 160     sum-file       A test summary file with the format of those
 161                    created by runtest from DejaGnu.
 162     If -L is used, merge *.log files instead of *.sum.  In this
 163     mode the exact order of lines may not be preserved, just different
 164     Running *.exp chunks should be in correct order.
 165 ''')
 166         sys.exit (1)
 167
 168     def fatal (self, what, string):
 169         if not what:
 170             what = sys.argv[0]
 171         sys.stderr.write (what + ': ' + string + '\n')
 172         sys.exit (1)
 173
 174     # Parse the command-line arguments.
 175     def parse_cmdline (self):
 176         try:
 177             (options, self.files) = getopt.getopt (sys.argv[1:], 'l:t:L')
 178             if len (self.files) == 0:
 179                 self.usage()
 180             for (option, value) in options:
 181                 if option == '-l':
 182                     self.variations.append (value)
 183                 elif option == '-t':
 184                     self.tools.append (value)
 185                 else:
 186                     self.do_sum = False
 187         except getopt.GetoptError as e:
 188             self.fatal (None, e.msg)
 189
 190     # Try to parse time string TIME, returning an arbitrary time on failure.
 191     # Getting this right is just a nice-to-have so failures should be silent.
 192     def parse_time (self, time):
 193         try:
 194             return datetime.strptime (time, '%c')
 195         except ValueError:
 196             return datetime.now()
 197
 198     # Parse an integer and abort on failure.
 199     def parse_int (self, filename, value):
 200         try:
 201             return int (value)
 202         except ValueError:
 203             self.fatal (filename, 'expected an integer, got: ' + value)
 204
 205     # Return a list that represents no test results.
 206     def zero_counts (self):
 207         return [0 for x in self.count_names]
 208
 209     # Return the ToolRun for tool NAME.
 210     def get_tool (self, name):
 211         if name not in self.runs:
 212             self.runs[name] = ToolRun (name)
 213         return self.runs[name]
 214
 215     # Add the result counts in list FROMC to TOC.
 216     def accumulate_counts (self, toc, fromc):
 217         for i in range (len (self.count_names)):
 218             toc[i] += fromc[i]
 219
 220     # Parse the list of variations after 'Schedule of variations:'.
 221     # Return the number seen.
 222     def parse_variations (self, filename, file):
 223         num_variations = 0
 224         while True:
 225             line = file.readline()
 226             if line == '':
 227                 self.fatal (filename, 'could not parse variation list')
 228             if line == '\n':
 229                 break
 230             self.known_variations.add (line.strip())
 231             num_variations += 1
 232         return num_variations
 233
 234     # Parse from the first line after 'Running target ...' to the end
 235     # of the run's summary.
 236     def parse_run (self, filename, file, tool, variation, num_variations):
 237         header = None
 238         harness = None
 239         segment = None
 240         final_using = 0
 241
 242         # If this is the first run for this variation, add any text before
 243         # the first harness to the header.
 244         if not variation.header:
 245             segment = Segment (filename, file.tell())
 246             variation.header = segment
 247
 248         # Parse up until the first line of the summary.
 249         if num_variations == 1:
 250             end = '\t\t=== ' + tool.name + ' Summary ===\n'
 251         else:
 252             end = ('\t\t=== ' + tool.name + ' Summary for '
 253                    + variation.name + ' ===\n')
 254         while True:
 255             line = file.readline()
 256             if line == '':
 257                 self.fatal (filename, 'no recognised summary line')
 258             if line == end:
 259                 break
 260
 261             # Look for the start of a new harness.
 262             if line.startswith ('Running ') and line.endswith (' ...\n'):
 263                 # Close off the current harness segment, if any.
 264                 if harness:
 265                     segment.lines -= final_using
 266                     harness.add_segment (first_key, segment)
 267                 name = line[len ('Running '):-len(' ...\n')]
 268                 harness = variation.get_harness (name)
 269                 segment = Segment (filename, file.tell())
 270                 first_key = None
 271                 final_using = 0
 272                 continue
 273
 274             # Record test results.  Associate the first test result with
 275             # the harness segment, so that if a run for a particular harness
 276             # has been split up, we can reassemble the individual segments
 277             # in a sensible order.
 278             #
 279             # dejagnu sometimes issues warnings about the testing environment
 280             # before running any tests.  Treat them as part of the header
 281             # rather than as a test result.
 282             match = self.result_re.match (line)
 283             if match and (harness or not line.startswith ('WARNING:')):
 284                 if not harness:
 285                     self.fatal (filename, 'saw test result before harness name')
 286                 name = match.group (2)
 287                 # Ugly hack to get the right order for gfortran.
 288                 if name.startswith ('gfortran.dg/g77/'):
 289                     name = 'h' + name
 290                 key = (name, len (harness.results))
 291                 harness.results.append ((key, line))
 292                 if not first_key and sort_logs:
 293                     first_key = key
 294
 295             # 'Using ...' lines are only interesting in a header.  Splitting
 296             # the test up into parallel runs leads to more 'Using ...' lines
 297             # than there would be in a single log.
 298             if line.startswith ('Using '):
 299                 final_using += 1
 300             else:
 301                 final_using = 0
 302
 303             # Add other text to the current segment, if any.
 304             if segment:
 305                 segment.lines += 1
 306
 307         # Close off the final harness segment, if any.
 308         if harness:
 309             segment.lines -= final_using
 310             harness.add_segment (first_key, segment)
 311
 312         # Parse the rest of the summary (the '# of ' lines).
 313         if len (variation.counts) == 0:
 314             variation.counts = self.zero_counts()
 315         while True:
 316             before = file.tell()
 317             line = file.readline()
 318             if line == '':
 319                 break
 320             if line == '\n':
 321                 continue
 322             if not line.startswith ('# '):
 323                 file.seek (before)
 324                 break
 325             found = False
 326             for i in range (len (self.count_names)):
 327                 if line.startswith (self.count_names[i]):
 328                     count = line[len (self.count_names[i]):-1].strip()
 329                     variation.counts[i] += self.parse_int (filename, count)
 330                     found = True
 331                     break
 332             if not found:
 333                 self.fatal (filename, 'unknown test result: ' + line[:-1])
 334
 335     # Parse an acats run, which uses a different format from dejagnu.
 336     # We have just skipped over '=== acats configuration ==='.
 337     def parse_acats_run (self, filename, file):
 338         # Parse the preamble, which describes the configuration and logs
 339         # the creation of support files.
 340         record = (self.acats_premable == '')
 341         if record:
 342             self.acats_premable = '\t\t=== acats configuration ===\n'
 343         while True:
 344             line = file.readline()
 345             if line == '':
 346                 self.fatal (filename, 'could not parse acats preamble')
 347             if line == '\t\t=== acats tests ===\n':
 348                 break
 349             if record:
 350                 self.acats_premable += line
 351
 352         # Parse the test results themselves, using a dummy variation name.
 353         tool = self.get_tool ('acats')
 354         variation = tool.get_variation ('none')
 355         self.parse_run (filename, file, tool, variation, 1)
 356
 357         # Parse the failure list.
 358         while True:
 359             before = file.tell()
 360             line = file.readline()
 361             if line.startswith ('*** FAILURES: '):
 362                 self.acats_failures.append (line[len ('*** FAILURES: '):-1])
 363                 continue
 364             file.seek (before)
 365             break
 366
 367     # Parse the final summary at the end of a log in order to capture
 368     # the version output that follows it.
 369     def parse_final_summary (self, filename, file):
 370         record = (self.version_output == '')
 371         while True:
 372             line = file.readline()
 373             if line == '':
 374                 break
 375             if line.startswith ('# of '):
 376                 continue
 377             if record:
 378                 self.version_output += line
 379             if line == '\n':
 380                 break
 381
 382     # Parse a .log or .sum file.
 383     def parse_file (self, filename, file):
 384         tool = None
 385         target = None
 386         num_variations = 1
 387         while True:
 388             line = file.readline()
 389             if line == '':
 390                 return
 391
 392             # Parse the list of variations, which comes before the test
 393             # runs themselves.
 394             if line.startswith ('Schedule of variations:'):
 395                 num_variations = self.parse_variations (filename, file)
 396                 continue
 397
 398             # Parse a testsuite run for one tool/variation combination.
 399             if line.startswith ('Running target '):
 400                 name = line[len ('Running target '):-1]
 401                 if not tool:
 402                     self.fatal (filename, 'could not parse tool name')
 403                 if name not in self.known_variations:
 404                     self.fatal (filename, 'unknown target: ' + name)
 405                 self.parse_run (filename, file, tool,
 406                                 tool.get_variation (name),
 407                                 num_variations)
 408                 # If there is only one variation then there is no separate
 409                 # summary for it.  Record any following version output.
 410                 if num_variations == 1:
 411                     self.parse_final_summary (filename, file)
 412                 continue
 413
 414             # Parse the start line.  In the case where several files are being
 415             # parsed, pick the one with the earliest time.
 416             match = self.test_run_re.match (line)
 417             if match:
 418                 time = self.parse_time (match.group (2))
 419                 if not self.start_line or self.start_line[0] > time:
 420                     self.start_line = (time, line)
 421                 continue
 422
 423             # Parse the form used for native testing.
 424             if line.startswith ('Native configuration is '):
 425                 self.native_line = line
 426                 continue
 427
 428             # Parse the target triplet.
 429             if line.startswith ('Target is '):
 430                 self.target_line = line
 431                 continue
 432
 433             # Parse the host triplet.
 434             if line.startswith ('Host   is '):
 435                 self.host_line = line
 436                 continue
 437
 438             # Parse the acats premable.
 439             if line == '\t\t=== acats configuration ===\n':
 440                 self.parse_acats_run (filename, file)
 441                 continue
 442
 443             # Parse the tool name.
 444             match = self.tool_re.match (line)
 445             if match:
 446                 tool = self.get_tool (match.group (1))
 447                 continue
 448
 449             # Skip over the final summary (which we instead create from
 450             # individual runs) and parse the version output.
 451             if tool and line == '\t\t=== ' + tool.name + ' Summary ===\n':
 452                 if file.readline() != '\n':
 453                     self.fatal (filename, 'expected blank line after summary')
 454                 self.parse_final_summary (filename, file)
 455                 continue
 456
 457             # Parse the completion line.  In the case where several files
 458             # are being parsed, pick the one with the latest time.
 459             match = self.completed_re.match (line)
 460             if match:
 461                 time = self.parse_time (match.group (1))
 462                 if not self.end_line or self.end_line[0] < time:
 463                     self.end_line = (time, line)
 464                 continue
 465
 466             # Sanity check to make sure that important text doesn't get
 467             # dropped accidentally.
 468             if strict and line.strip() != '':
 469                 self.fatal (filename, 'unrecognised line: ' + line[:-1])
 470
 471     # Output a segment of text.
 472     def output_segment (self, segment):
 473         with safe_open (segment.filename) as file:
 474             file.seek (segment.start)
 475             for i in range (segment.lines):
 476                 sys.stdout.write (file.readline())
 477
 478     # Output a summary giving the number of times each type of result has
 479     # been seen.
 480     def output_summary (self, tool, counts):
 481         for i in range (len (self.count_names)):
 482             name = self.count_names[i]
 483             # dejagnu only prints result types that were seen at least once,
 484             # but acats always prints a number of unexpected failures.
 485             if (counts[i] > 0
 486                 or (tool.name == 'acats'
 487                     and name.startswith ('# of unexpected failures'))):
 488                 sys.stdout.write ('%s%d\n' % (name, counts[i]))
 489
 490     # Output unified .log or .sum information for a particular variation,
 491     # with a summary at the end.
 492     def output_variation (self, tool, variation):
 493         self.output_segment (variation.header)
 494         for harness in sorted (variation.harnesses.values(),
 495                                key = attrgetter ('name')):
 496             sys.stdout.write ('Running ' + harness.name + ' ...\n')
 497             if self.do_sum:
 498                 harness.results.sort()
 499                 for (key, line) in harness.results:
 500                     sys.stdout.write (line)
 501             else:
 502                 # Rearrange the log segments into test order (but without
 503                 # rearranging text within those segments).
 504                 for key in sorted (harness.segments.keys()):
 505                     self.output_segment (harness.segments[key])
 506                 for segment in harness.empty:
 507                     self.output_segment (segment)
 508         if len (self.variations) > 1:
 509             sys.stdout.write ('\t\t=== ' + tool.name + ' Summary for '
 510                               + variation.name + ' ===\n\n')
 511             self.output_summary (tool, variation.counts)
 512
 513     # Output unified .log or .sum information for a particular tool,
 514     # with a summary at the end.
 515     def output_tool (self, tool):
 516         counts = self.zero_counts()
 517         if tool.name == 'acats':
 518             # acats doesn't use variations, so just output everything.
 519             # It also has a different approach to whitespace.
 520             sys.stdout.write ('\t\t=== ' + tool.name + ' tests ===\n')
 521             for variation in tool.variations.values():
 522                 self.output_variation (tool, variation)
 523                 self.accumulate_counts (counts, variation.counts)
 524             sys.stdout.write ('\t\t=== ' + tool.name + ' Summary ===\n')
 525         else:
 526             # Output the results in the usual dejagnu runtest format.
 527             sys.stdout.write ('\n\t\t=== ' + tool.name + ' tests ===\n\n'
 528                               'Schedule of variations:\n')
 529             for name in self.variations:
 530                 if name in tool.variations:
 531                     sys.stdout.write ('    ' + name + '\n')
 532             sys.stdout.write ('\n')
 533             for name in self.variations:
 534                 if name in tool.variations:
 535                     variation = tool.variations[name]
 536                     sys.stdout.write ('Running target '
 537                                       + variation.name + '\n')
 538                     self.output_variation (tool, variation)
 539                     self.accumulate_counts (counts, variation.counts)
 540             sys.stdout.write ('\n\t\t=== ' + tool.name + ' Summary ===\n\n')
 541         self.output_summary (tool, counts)
 542
 543     def main (self):
 544         self.parse_cmdline()
 545         try:
 546             # Parse the input files.
 547             for filename in self.files:
 548                 with safe_open (filename) as file:
 549                     self.parse_file (filename, file)
 550
 551             # Decide what to output.
 552             if len (self.variations) == 0:
 553                 self.variations = sorted (self.known_variations)
 554             else:
 555                 for name in self.variations:
 556                     if name not in self.known_variations:
 557                         self.fatal (None, 'no results for ' + name)
 558             if len (self.tools) == 0:
 559                 self.tools = sorted (self.runs.keys())
 560
 561             # Output the header.
 562             if self.start_line:
 563                 sys.stdout.write (self.start_line[1])
 564             sys.stdout.write (self.native_line)
 565             sys.stdout.write (self.target_line)
 566             sys.stdout.write (self.host_line)
 567             sys.stdout.write (self.acats_premable)
 568
 569             # Output the main body.
 570             for name in self.tools:
 571                 if name not in self.runs:
 572                     self.fatal (None, 'no results for ' + name)
 573                 self.output_tool (self.runs[name])
 574
 575             # Output the footer.
 576             if len (self.acats_failures) > 0:
 577                 sys.stdout.write ('*** FAILURES: '
 578                                   + ' '.join (self.acats_failures) + '\n')
 579             sys.stdout.write (self.version_output)
 580             if self.end_line:
 581                 sys.stdout.write (self.end_line[1])
 582         except IOError as e:
 583             self.fatal (e.filename, e.strerror)
 584
 585 Prog().main()