Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(422)

Side by Side Diff: swarm_client/isolate.py

Issue 69143004: Delete swarm_client. (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/tools/
Patch Set: Created 7 years, 1 month ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
« no previous file with comments | « swarm_client/googletest/trace_test_cases.py ('k') | swarm_client/isolate_merge.py » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
(Empty)
1 #!/usr/bin/env python
2 # Copyright (c) 2012 The Chromium Authors. All rights reserved.
3 # Use of this source code is governed by a BSD-style license that can be
4 # found in the LICENSE file.
5
6 """Front end tool to operate on .isolate files.
7
8 This includes creating, merging or compiling them to generate a .isolated file.
9
10 See more information at
11 https://code.google.com/p/swarming/wiki/IsolateDesign
12 https://code.google.com/p/swarming/wiki/IsolateUserGuide
13 """
14 # Run ./isolate.py --help for more detailed information.
15
16 import ast
17 import copy
18 import itertools
19 import logging
20 import optparse
21 import os
22 import posixpath
23 import re
24 import stat
25 import subprocess
26 import sys
27
28 import isolateserver
29 import run_isolated
30 import trace_inputs
31
32 # Import here directly so isolate is easier to use as a library.
33 from run_isolated import get_flavor
34
35 from third_party import colorama
36 from third_party.depot_tools import fix_encoding
37 from third_party.depot_tools import subcommand
38
39 from utils import file_path
40 from utils import tools
41 from utils import short_expression_finder
42
43
44 __version__ = '0.1.1'
45
46
47 PATH_VARIABLES = ('DEPTH', 'PRODUCT_DIR')
48
49 # Files that should be 0-length when mapped.
50 KEY_TOUCHED = 'isolate_dependency_touched'
51 # Files that should be tracked by the build tool.
52 KEY_TRACKED = 'isolate_dependency_tracked'
53 # Files that should not be tracked by the build tool.
54 KEY_UNTRACKED = 'isolate_dependency_untracked'
55
56
57 class ExecutionError(Exception):
58 """A generic error occurred."""
59 def __str__(self):
60 return self.args[0]
61
62
63 ### Path handling code.
64
65
66 DEFAULT_BLACKLIST = (
67 # Temporary vim or python files.
68 r'^.+\.(?:pyc|swp)$',
69 # .git or .svn directory.
70 r'^(?:.+' + re.escape(os.path.sep) + r'|)\.(?:git|svn)$',
71 )
72
73
74 # Chromium-specific.
75 DEFAULT_BLACKLIST += (
76 r'^.+\.(?:run_test_cases)$',
77 r'^(?:.+' + re.escape(os.path.sep) + r'|)testserver\.log$',
78 )
79
80
81 def relpath(path, root):
82 """os.path.relpath() that keeps trailing os.path.sep."""
83 out = os.path.relpath(path, root)
84 if path.endswith(os.path.sep):
85 out += os.path.sep
86 return out
87
88
89 def safe_relpath(filepath, basepath):
90 """Do not throw on Windows when filepath and basepath are on different drives.
91
92 Different than relpath() above since this one doesn't keep the trailing
93 os.path.sep and it swallows exceptions on Windows and return the original
94 absolute path in the case of different drives.
95 """
96 try:
97 return os.path.relpath(filepath, basepath)
98 except ValueError:
99 assert sys.platform == 'win32'
100 return filepath
101
102
103 def normpath(path):
104 """os.path.normpath() that keeps trailing os.path.sep."""
105 out = os.path.normpath(path)
106 if path.endswith(os.path.sep):
107 out += os.path.sep
108 return out
109
110
111 def posix_relpath(path, root):
112 """posix.relpath() that keeps trailing slash."""
113 out = posixpath.relpath(path, root)
114 if path.endswith('/'):
115 out += '/'
116 return out
117
118
119 def cleanup_path(x):
120 """Cleans up a relative path. Converts any os.path.sep to '/' on Windows."""
121 if x:
122 x = x.rstrip(os.path.sep).replace(os.path.sep, '/')
123 if x == '.':
124 x = ''
125 if x:
126 x += '/'
127 return x
128
129
130 def is_url(path):
131 return bool(re.match(r'^https?://.+$', path))
132
133
134 def path_starts_with(prefix, path):
135 """Returns true if the components of the path |prefix| are the same as the
136 initial components of |path| (or all of the components of |path|). The paths
137 must be absolute.
138 """
139 assert os.path.isabs(prefix) and os.path.isabs(path)
140 prefix = os.path.normpath(prefix)
141 path = os.path.normpath(path)
142 assert prefix == file_path.get_native_path_case(prefix), prefix
143 assert path == file_path.get_native_path_case(path), path
144 prefix = prefix.rstrip(os.path.sep) + os.path.sep
145 path = path.rstrip(os.path.sep) + os.path.sep
146 return path.startswith(prefix)
147
148
149 def fix_native_path_case(root, path):
150 """Ensures that each component of |path| has the proper native case by
151 iterating slowly over the directory elements of |path|."""
152 native_case_path = root
153 for raw_part in path.split(os.sep):
154 if not raw_part or raw_part == '.':
155 break
156
157 part = file_path.find_item_native_case(native_case_path, raw_part)
158 if not part:
159 raise isolateserver.MappingError(
160 'Input file %s doesn\'t exist' %
161 os.path.join(native_case_path, raw_part))
162 native_case_path = os.path.join(native_case_path, part)
163
164 return os.path.normpath(native_case_path)
165
166
167 def expand_symlinks(indir, relfile):
168 """Follows symlinks in |relfile|, but treating symlinks that point outside the
169 build tree as if they were ordinary directories/files. Returns the final
170 symlink-free target and a list of paths to symlinks encountered in the
171 process.
172
173 The rule about symlinks outside the build tree is for the benefit of the
174 Chromium OS ebuild, which symlinks the output directory to an unrelated path
175 in the chroot.
176
177 Fails when a directory loop is detected, although in theory we could support
178 that case.
179 """
180 is_directory = relfile.endswith(os.path.sep)
181 done = indir
182 todo = relfile.strip(os.path.sep)
183 symlinks = []
184
185 while todo:
186 pre_symlink, symlink, post_symlink = file_path.split_at_symlink(
187 done, todo)
188 if not symlink:
189 todo = fix_native_path_case(done, todo)
190 done = os.path.join(done, todo)
191 break
192 symlink_path = os.path.join(done, pre_symlink, symlink)
193 post_symlink = post_symlink.lstrip(os.path.sep)
194 # readlink doesn't exist on Windows.
195 # pylint: disable=E1101
196 target = os.path.normpath(os.path.join(done, pre_symlink))
197 symlink_target = os.readlink(symlink_path)
198 if os.path.isabs(symlink_target):
199 # Absolute path are considered a normal directories. The use case is
200 # generally someone who puts the output directory on a separate drive.
201 target = symlink_target
202 else:
203 # The symlink itself could be using the wrong path case.
204 target = fix_native_path_case(target, symlink_target)
205
206 if not os.path.exists(target):
207 raise isolateserver.MappingError(
208 'Symlink target doesn\'t exist: %s -> %s' % (symlink_path, target))
209 target = file_path.get_native_path_case(target)
210 if not path_starts_with(indir, target):
211 done = symlink_path
212 todo = post_symlink
213 continue
214 if path_starts_with(target, symlink_path):
215 raise isolateserver.MappingError(
216 'Can\'t map recursive symlink reference %s -> %s' %
217 (symlink_path, target))
218 logging.info('Found symlink: %s -> %s', symlink_path, target)
219 symlinks.append(os.path.relpath(symlink_path, indir))
220 # Treat the common prefix of the old and new paths as done, and start
221 # scanning again.
222 target = target.split(os.path.sep)
223 symlink_path = symlink_path.split(os.path.sep)
224 prefix_length = 0
225 for target_piece, symlink_path_piece in zip(target, symlink_path):
226 if target_piece == symlink_path_piece:
227 prefix_length += 1
228 else:
229 break
230 done = os.path.sep.join(target[:prefix_length])
231 todo = os.path.join(
232 os.path.sep.join(target[prefix_length:]), post_symlink)
233
234 relfile = os.path.relpath(done, indir)
235 relfile = relfile.rstrip(os.path.sep) + is_directory * os.path.sep
236 return relfile, symlinks
237
238
239 def expand_directory_and_symlink(indir, relfile, blacklist, follow_symlinks):
240 """Expands a single input. It can result in multiple outputs.
241
242 This function is recursive when relfile is a directory.
243
244 Note: this code doesn't properly handle recursive symlink like one created
245 with:
246 ln -s .. foo
247 """
248 if os.path.isabs(relfile):
249 raise isolateserver.MappingError(
250 'Can\'t map absolute path %s' % relfile)
251
252 infile = normpath(os.path.join(indir, relfile))
253 if not infile.startswith(indir):
254 raise isolateserver.MappingError(
255 'Can\'t map file %s outside %s' % (infile, indir))
256
257 filepath = os.path.join(indir, relfile)
258 native_filepath = file_path.get_native_path_case(filepath)
259 if filepath != native_filepath:
260 # Special case './'.
261 if filepath != native_filepath + '.' + os.path.sep:
262 # Give up enforcing strict path case on OSX. Really, it's that sad. The
263 # case where it happens is very specific and hard to reproduce:
264 # get_native_path_case(
265 # u'Foo.framework/Versions/A/Resources/Something.nib') will return
266 # u'Foo.framework/Versions/A/resources/Something.nib', e.g. lowercase 'r'.
267 #
268 # Note that this is really something deep in OSX because running
269 # ls Foo.framework/Versions/A
270 # will print out 'Resources', while file_path.get_native_path_case()
271 # returns a lower case 'r'.
272 #
273 # So *something* is happening under the hood resulting in the command 'ls'
274 # and Carbon.File.FSPathMakeRef('path').FSRefMakePath() to disagree. We
275 # have no idea why.
276 if sys.platform != 'darwin':
277 raise isolateserver.MappingError(
278 'File path doesn\'t equal native file path\n%s != %s' %
279 (filepath, native_filepath))
280
281 symlinks = []
282 if follow_symlinks:
283 relfile, symlinks = expand_symlinks(indir, relfile)
284
285 if relfile.endswith(os.path.sep):
286 if not os.path.isdir(infile):
287 raise isolateserver.MappingError(
288 '%s is not a directory but ends with "%s"' % (infile, os.path.sep))
289
290 # Special case './'.
291 if relfile.startswith('.' + os.path.sep):
292 relfile = relfile[2:]
293 outfiles = symlinks
294 try:
295 for filename in os.listdir(infile):
296 inner_relfile = os.path.join(relfile, filename)
297 if blacklist(inner_relfile):
298 continue
299 if os.path.isdir(os.path.join(indir, inner_relfile)):
300 inner_relfile += os.path.sep
301 outfiles.extend(
302 expand_directory_and_symlink(indir, inner_relfile, blacklist,
303 follow_symlinks))
304 return outfiles
305 except OSError as e:
306 raise isolateserver.MappingError(
307 'Unable to iterate over directory %s.\n%s' % (infile, e))
308 else:
309 # Always add individual files even if they were blacklisted.
310 if os.path.isdir(infile):
311 raise isolateserver.MappingError(
312 'Input directory %s must have a trailing slash' % infile)
313
314 if not os.path.isfile(infile):
315 raise isolateserver.MappingError(
316 'Input file %s doesn\'t exist' % infile)
317
318 return symlinks + [relfile]
319
320
321 def expand_directories_and_symlinks(indir, infiles, blacklist,
322 follow_symlinks, ignore_broken_items):
323 """Expands the directories and the symlinks, applies the blacklist and
324 verifies files exist.
325
326 Files are specified in os native path separator.
327 """
328 outfiles = []
329 for relfile in infiles:
330 try:
331 outfiles.extend(expand_directory_and_symlink(indir, relfile, blacklist,
332 follow_symlinks))
333 except isolateserver.MappingError as e:
334 if ignore_broken_items:
335 logging.info('warning: %s', e)
336 else:
337 raise
338 return outfiles
339
340
341 def recreate_tree(outdir, indir, infiles, action, as_hash):
342 """Creates a new tree with only the input files in it.
343
344 Arguments:
345 outdir: Output directory to create the files in.
346 indir: Root directory the infiles are based in.
347 infiles: dict of files to map from |indir| to |outdir|.
348 action: One of accepted action of run_isolated.link_file().
349 as_hash: Output filename is the hash instead of relfile.
350 """
351 logging.info(
352 'recreate_tree(outdir=%s, indir=%s, files=%d, action=%s, as_hash=%s)' %
353 (outdir, indir, len(infiles), action, as_hash))
354
355 assert os.path.isabs(outdir) and outdir == os.path.normpath(outdir), outdir
356 if not os.path.isdir(outdir):
357 logging.info('Creating %s' % outdir)
358 os.makedirs(outdir)
359
360 for relfile, metadata in infiles.iteritems():
361 infile = os.path.join(indir, relfile)
362 if as_hash:
363 # Do the hashtable specific checks.
364 if 'l' in metadata:
365 # Skip links when storing a hashtable.
366 continue
367 outfile = os.path.join(outdir, metadata['h'])
368 if os.path.isfile(outfile):
369 # Just do a quick check that the file size matches. No need to stat()
370 # again the input file, grab the value from the dict.
371 if not 's' in metadata:
372 raise isolateserver.MappingError(
373 'Misconfigured item %s: %s' % (relfile, metadata))
374 if metadata['s'] == os.stat(outfile).st_size:
375 continue
376 else:
377 logging.warn('Overwritting %s' % metadata['h'])
378 os.remove(outfile)
379 else:
380 outfile = os.path.join(outdir, relfile)
381 outsubdir = os.path.dirname(outfile)
382 if not os.path.isdir(outsubdir):
383 os.makedirs(outsubdir)
384
385 # TODO(csharp): Fix crbug.com/150823 and enable the touched logic again.
386 # if metadata.get('T') == True:
387 # open(outfile, 'ab').close()
388 if 'l' in metadata:
389 pointed = metadata['l']
390 logging.debug('Symlink: %s -> %s' % (outfile, pointed))
391 # symlink doesn't exist on Windows.
392 os.symlink(pointed, outfile) # pylint: disable=E1101
393 else:
394 run_isolated.link_file(outfile, infile, action)
395
396
397 def process_input(filepath, prevdict, read_only, flavor, algo):
398 """Processes an input file, a dependency, and return meta data about it.
399
400 Arguments:
401 - filepath: File to act on.
402 - prevdict: the previous dictionary. It is used to retrieve the cached sha-1
403 to skip recalculating the hash.
404 - read_only: If True, the file mode is manipulated. In practice, only save
405 one of 4 modes: 0755 (rwx), 0644 (rw), 0555 (rx), 0444 (r). On
406 windows, mode is not set since all files are 'executable' by
407 default.
408 - algo: Hashing algorithm used.
409
410 Behaviors:
411 - Retrieves the file mode, file size, file timestamp, file link
412 destination if it is a file link and calcultate the SHA-1 of the file's
413 content if the path points to a file and not a symlink.
414 """
415 out = {}
416 # TODO(csharp): Fix crbug.com/150823 and enable the touched logic again.
417 # if prevdict.get('T') == True:
418 # # The file's content is ignored. Skip the time and hard code mode.
419 # if get_flavor() != 'win':
420 # out['m'] = stat.S_IRUSR | stat.S_IRGRP
421 # out['s'] = 0
422 # out['h'] = algo().hexdigest()
423 # out['T'] = True
424 # return out
425
426 # Always check the file stat and check if it is a link. The timestamp is used
427 # to know if the file's content/symlink destination should be looked into.
428 # E.g. only reuse from prevdict if the timestamp hasn't changed.
429 # There is the risk of the file's timestamp being reset to its last value
430 # manually while its content changed. We don't protect against that use case.
431 try:
432 filestats = os.lstat(filepath)
433 except OSError:
434 # The file is not present.
435 raise isolateserver.MappingError('%s is missing' % filepath)
436 is_link = stat.S_ISLNK(filestats.st_mode)
437
438 if flavor != 'win':
439 # Ignore file mode on Windows since it's not really useful there.
440 filemode = stat.S_IMODE(filestats.st_mode)
441 # Remove write access for group and all access to 'others'.
442 filemode &= ~(stat.S_IWGRP | stat.S_IRWXO)
443 if read_only:
444 filemode &= ~stat.S_IWUSR
445 if filemode & stat.S_IXUSR:
446 filemode |= stat.S_IXGRP
447 else:
448 filemode &= ~stat.S_IXGRP
449 if not is_link:
450 out['m'] = filemode
451
452 # Used to skip recalculating the hash or link destination. Use the most recent
453 # update time.
454 # TODO(maruel): Save it in the .state file instead of .isolated so the
455 # .isolated file is deterministic.
456 out['t'] = int(round(filestats.st_mtime))
457
458 if not is_link:
459 out['s'] = filestats.st_size
460 # If the timestamp wasn't updated and the file size is still the same, carry
461 # on the sha-1.
462 if (prevdict.get('t') == out['t'] and
463 prevdict.get('s') == out['s']):
464 # Reuse the previous hash if available.
465 out['h'] = prevdict.get('h')
466 if not out.get('h'):
467 out['h'] = isolateserver.hash_file(filepath, algo)
468 else:
469 # If the timestamp wasn't updated, carry on the link destination.
470 if prevdict.get('t') == out['t']:
471 # Reuse the previous link destination if available.
472 out['l'] = prevdict.get('l')
473 if out.get('l') is None:
474 # The link could be in an incorrect path case. In practice, this only
475 # happen on OSX on case insensitive HFS.
476 # TODO(maruel): It'd be better if it was only done once, in
477 # expand_directory_and_symlink(), so it would not be necessary to do again
478 # here.
479 symlink_value = os.readlink(filepath) # pylint: disable=E1101
480 filedir = file_path.get_native_path_case(os.path.dirname(filepath))
481 native_dest = fix_native_path_case(filedir, symlink_value)
482 out['l'] = os.path.relpath(native_dest, filedir)
483 return out
484
485
486 ### Variable stuff.
487
488
489 def isolatedfile_to_state(filename):
490 """Replaces the file's extension."""
491 return filename + '.state'
492
493
494 def determine_root_dir(relative_root, infiles):
495 """For a list of infiles, determines the deepest root directory that is
496 referenced indirectly.
497
498 All arguments must be using os.path.sep.
499 """
500 # The trick used to determine the root directory is to look at "how far" back
501 # up it is looking up.
502 deepest_root = relative_root
503 for i in infiles:
504 x = relative_root
505 while i.startswith('..' + os.path.sep):
506 i = i[3:]
507 assert not i.startswith(os.path.sep)
508 x = os.path.dirname(x)
509 if deepest_root.startswith(x):
510 deepest_root = x
511 logging.debug(
512 'determine_root_dir(%s, %d files) -> %s' % (
513 relative_root, len(infiles), deepest_root))
514 return deepest_root
515
516
517 def replace_variable(part, variables):
518 m = re.match(r'<\(([A-Z_]+)\)', part)
519 if m:
520 if m.group(1) not in variables:
521 raise ExecutionError(
522 'Variable "%s" was not found in %s.\nDid you forget to specify '
523 '--variable?' % (m.group(1), variables))
524 return variables[m.group(1)]
525 return part
526
527
528 def process_variables(cwd, variables, relative_base_dir):
529 """Processes path variables as a special case and returns a copy of the dict.
530
531 For each 'path' variable: first normalizes it based on |cwd|, verifies it
532 exists then sets it as relative to relative_base_dir.
533 """
534 relative_base_dir = file_path.get_native_path_case(relative_base_dir)
535 variables = variables.copy()
536 for i in PATH_VARIABLES:
537 if i not in variables:
538 continue
539 variable = variables[i].strip()
540 # Variables could contain / or \ on windows. Always normalize to
541 # os.path.sep.
542 variable = variable.replace('/', os.path.sep)
543 variable = os.path.join(cwd, variable)
544 variable = os.path.normpath(variable)
545 variable = file_path.get_native_path_case(variable)
546 if not os.path.isdir(variable):
547 raise ExecutionError('%s=%s is not a directory' % (i, variable))
548
549 # All variables are relative to the .isolate file.
550 variable = os.path.relpath(variable, relative_base_dir)
551 logging.debug(
552 'Translated variable %s from %s to %s', i, variables[i], variable)
553 variables[i] = variable
554 return variables
555
556
557 def eval_variables(item, variables):
558 """Replaces the .isolate variables in a string item.
559
560 Note that the .isolate format is a subset of the .gyp dialect.
561 """
562 return ''.join(
563 replace_variable(p, variables) for p in re.split(r'(<\([A-Z_]+\))', item))
564
565
566 def classify_files(root_dir, tracked, untracked):
567 """Converts the list of files into a .isolate 'variables' dictionary.
568
569 Arguments:
570 - tracked: list of files names to generate a dictionary out of that should
571 probably be tracked.
572 - untracked: list of files names that must not be tracked.
573 """
574 # These directories are not guaranteed to be always present on every builder.
575 OPTIONAL_DIRECTORIES = (
576 'test/data/plugin',
577 'third_party/WebKit/LayoutTests',
578 )
579
580 new_tracked = []
581 new_untracked = list(untracked)
582
583 def should_be_tracked(filepath):
584 """Returns True if it is a file without whitespace in a non-optional
585 directory that has no symlink in its path.
586 """
587 if filepath.endswith('/'):
588 return False
589 if ' ' in filepath:
590 return False
591 if any(i in filepath for i in OPTIONAL_DIRECTORIES):
592 return False
593 # Look if any element in the path is a symlink.
594 split = filepath.split('/')
595 for i in range(len(split)):
596 if os.path.islink(os.path.join(root_dir, '/'.join(split[:i+1]))):
597 return False
598 return True
599
600 for filepath in sorted(tracked):
601 if should_be_tracked(filepath):
602 new_tracked.append(filepath)
603 else:
604 # Anything else.
605 new_untracked.append(filepath)
606
607 variables = {}
608 if new_tracked:
609 variables[KEY_TRACKED] = sorted(new_tracked)
610 if new_untracked:
611 variables[KEY_UNTRACKED] = sorted(new_untracked)
612 return variables
613
614
615 def chromium_fix(f, variables):
616 """Fixes an isolate dependnecy with Chromium-specific fixes."""
617 # Skip log in PRODUCT_DIR. Note that these are applied on '/' style path
618 # separator.
619 LOG_FILE = re.compile(r'^\<\(PRODUCT_DIR\)\/[^\/]+\.log$')
620 # Ignored items.
621 IGNORED_ITEMS = (
622 # http://crbug.com/160539, on Windows, it's in chrome/.
623 'Media Cache/',
624 'chrome/Media Cache/',
625 # 'First Run' is not created by the compile, but by the test itself.
626 '<(PRODUCT_DIR)/First Run')
627
628 # Blacklist logs and other unimportant files.
629 if LOG_FILE.match(f) or f in IGNORED_ITEMS:
630 logging.debug('Ignoring %s', f)
631 return None
632
633 EXECUTABLE = re.compile(
634 r'^(\<\(PRODUCT_DIR\)\/[^\/\.]+)' +
635 re.escape(variables.get('EXECUTABLE_SUFFIX', '')) +
636 r'$')
637 match = EXECUTABLE.match(f)
638 if match:
639 return match.group(1) + '<(EXECUTABLE_SUFFIX)'
640
641 if sys.platform == 'darwin':
642 # On OSX, the name of the output is dependent on gyp define, it can be
643 # 'Google Chrome.app' or 'Chromium.app', same for 'XXX
644 # Framework.framework'. Furthermore, they are versioned with a gyp
645 # variable. To lower the complexity of the .isolate file, remove all the
646 # individual entries that show up under any of the 4 entries and replace
647 # them with the directory itself. Overall, this results in a bit more
648 # files than strictly necessary.
649 OSX_BUNDLES = (
650 '<(PRODUCT_DIR)/Chromium Framework.framework/',
651 '<(PRODUCT_DIR)/Chromium.app/',
652 '<(PRODUCT_DIR)/Google Chrome Framework.framework/',
653 '<(PRODUCT_DIR)/Google Chrome.app/',
654 )
655 for prefix in OSX_BUNDLES:
656 if f.startswith(prefix):
657 # Note this result in duplicate values, so the a set() must be used to
658 # remove duplicates.
659 return prefix
660 return f
661
662
663 def generate_simplified(
664 tracked, untracked, touched, root_dir, variables, relative_cwd,
665 trace_blacklist):
666 """Generates a clean and complete .isolate 'variables' dictionary.
667
668 Cleans up and extracts only files from within root_dir then processes
669 variables and relative_cwd.
670 """
671 root_dir = os.path.realpath(root_dir)
672 logging.info(
673 'generate_simplified(%d files, %s, %s, %s)' %
674 (len(tracked) + len(untracked) + len(touched),
675 root_dir, variables, relative_cwd))
676
677 # Preparation work.
678 relative_cwd = cleanup_path(relative_cwd)
679 assert not os.path.isabs(relative_cwd), relative_cwd
680 # Creates the right set of variables here. We only care about PATH_VARIABLES.
681 path_variables = dict(
682 ('<(%s)' % k, variables[k].replace(os.path.sep, '/'))
683 for k in PATH_VARIABLES if k in variables)
684 variables = variables.copy()
685 variables.update(path_variables)
686
687 # Actual work: Process the files.
688 # TODO(maruel): if all the files in a directory are in part tracked and in
689 # part untracked, the directory will not be extracted. Tracked files should be
690 # 'promoted' to be untracked as needed.
691 tracked = trace_inputs.extract_directories(
692 root_dir, tracked, trace_blacklist)
693 untracked = trace_inputs.extract_directories(
694 root_dir, untracked, trace_blacklist)
695 # touched is not compressed, otherwise it would result in files to be archived
696 # that we don't need.
697
698 root_dir_posix = root_dir.replace(os.path.sep, '/')
699 def fix(f):
700 """Bases the file on the most restrictive variable."""
701 # Important, GYP stores the files with / and not \.
702 f = f.replace(os.path.sep, '/')
703 logging.debug('fix(%s)' % f)
704 # If it's not already a variable.
705 if not f.startswith('<'):
706 # relative_cwd is usually the directory containing the gyp file. It may be
707 # empty if the whole directory containing the gyp file is needed.
708 # Use absolute paths in case cwd_dir is outside of root_dir.
709 # Convert the whole thing to / since it's isolate's speak.
710 f = posix_relpath(
711 posixpath.join(root_dir_posix, f),
712 posixpath.join(root_dir_posix, relative_cwd)) or './'
713
714 for variable, root_path in path_variables.iteritems():
715 if f.startswith(root_path):
716 f = variable + f[len(root_path):]
717 logging.debug('Converted to %s' % f)
718 break
719 return f
720
721 def fix_all(items):
722 """Reduces the items to convert variables, removes unneeded items, apply
723 chromium-specific fixes and only return unique items.
724 """
725 variables_converted = (fix(f.path) for f in items)
726 chromium_fixed = (chromium_fix(f, variables) for f in variables_converted)
727 return set(f for f in chromium_fixed if f)
728
729 tracked = fix_all(tracked)
730 untracked = fix_all(untracked)
731 touched = fix_all(touched)
732 out = classify_files(root_dir, tracked, untracked)
733 if touched:
734 out[KEY_TOUCHED] = sorted(touched)
735 return out
736
737
738 def chromium_filter_flags(variables):
739 """Filters out build flags used in Chromium that we don't want to treat as
740 configuration variables.
741 """
742 # TODO(benrg): Need a better way to determine this.
743 blacklist = set(PATH_VARIABLES + ('EXECUTABLE_SUFFIX', 'FLAG'))
744 return dict((k, v) for k, v in variables.iteritems() if k not in blacklist)
745
746
747 def generate_isolate(
748 tracked, untracked, touched, root_dir, variables, relative_cwd,
749 trace_blacklist):
750 """Generates a clean and complete .isolate file."""
751 dependencies = generate_simplified(
752 tracked, untracked, touched, root_dir, variables, relative_cwd,
753 trace_blacklist)
754 config_variables = chromium_filter_flags(variables)
755 config_variable_names, config_values = zip(
756 *sorted(config_variables.iteritems()))
757 out = Configs(None)
758 # The new dependencies apply to just one configuration, namely config_values.
759 out.merge_dependencies(dependencies, config_variable_names, [config_values])
760 return out.make_isolate_file()
761
762
763 def split_touched(files):
764 """Splits files that are touched vs files that are read."""
765 tracked = []
766 touched = []
767 for f in files:
768 if f.size:
769 tracked.append(f)
770 else:
771 touched.append(f)
772 return tracked, touched
773
774
775 def pretty_print(variables, stdout):
776 """Outputs a gyp compatible list from the decoded variables.
777
778 Similar to pprint.print() but with NIH syndrome.
779 """
780 # Order the dictionary keys by these keys in priority.
781 ORDER = (
782 'variables', 'condition', 'command', 'relative_cwd', 'read_only',
783 KEY_TRACKED, KEY_UNTRACKED)
784
785 def sorting_key(x):
786 """Gives priority to 'most important' keys before the others."""
787 if x in ORDER:
788 return str(ORDER.index(x))
789 return x
790
791 def loop_list(indent, items):
792 for item in items:
793 if isinstance(item, basestring):
794 stdout.write('%s\'%s\',\n' % (indent, item))
795 elif isinstance(item, dict):
796 stdout.write('%s{\n' % indent)
797 loop_dict(indent + ' ', item)
798 stdout.write('%s},\n' % indent)
799 elif isinstance(item, list):
800 # A list inside a list will write the first item embedded.
801 stdout.write('%s[' % indent)
802 for index, i in enumerate(item):
803 if isinstance(i, basestring):
804 stdout.write(
805 '\'%s\', ' % i.replace('\\', '\\\\').replace('\'', '\\\''))
806 elif isinstance(i, dict):
807 stdout.write('{\n')
808 loop_dict(indent + ' ', i)
809 if index != len(item) - 1:
810 x = ', '
811 else:
812 x = ''
813 stdout.write('%s}%s' % (indent, x))
814 else:
815 assert False
816 stdout.write('],\n')
817 else:
818 assert False
819
820 def loop_dict(indent, items):
821 for key in sorted(items, key=sorting_key):
822 item = items[key]
823 stdout.write("%s'%s': " % (indent, key))
824 if isinstance(item, dict):
825 stdout.write('{\n')
826 loop_dict(indent + ' ', item)
827 stdout.write(indent + '},\n')
828 elif isinstance(item, list):
829 stdout.write('[\n')
830 loop_list(indent + ' ', item)
831 stdout.write(indent + '],\n')
832 elif isinstance(item, basestring):
833 stdout.write(
834 '\'%s\',\n' % item.replace('\\', '\\\\').replace('\'', '\\\''))
835 elif item in (True, False, None):
836 stdout.write('%s\n' % item)
837 else:
838 assert False, item
839
840 stdout.write('{\n')
841 loop_dict(' ', variables)
842 stdout.write('}\n')
843
844
845 def union(lhs, rhs):
846 """Merges two compatible datastructures composed of dict/list/set."""
847 assert lhs is not None or rhs is not None
848 if lhs is None:
849 return copy.deepcopy(rhs)
850 if rhs is None:
851 return copy.deepcopy(lhs)
852 assert type(lhs) == type(rhs), (lhs, rhs)
853 if hasattr(lhs, 'union'):
854 # Includes set, ConfigSettings and Configs.
855 return lhs.union(rhs)
856 if isinstance(lhs, dict):
857 return dict((k, union(lhs.get(k), rhs.get(k))) for k in set(lhs).union(rhs))
858 elif isinstance(lhs, list):
859 # Do not go inside the list.
860 return lhs + rhs
861 assert False, type(lhs)
862
863
864 def extract_comment(content):
865 """Extracts file level comment."""
866 out = []
867 for line in content.splitlines(True):
868 if line.startswith('#'):
869 out.append(line)
870 else:
871 break
872 return ''.join(out)
873
874
875 def eval_content(content):
876 """Evaluates a python file and return the value defined in it.
877
878 Used in practice for .isolate files.
879 """
880 globs = {'__builtins__': None}
881 locs = {}
882 try:
883 value = eval(content, globs, locs)
884 except TypeError as e:
885 e.args = list(e.args) + [content]
886 raise
887 assert locs == {}, locs
888 assert globs == {'__builtins__': None}, globs
889 return value
890
891
892 def match_configs(expr, config_variables, all_configs):
893 """Returns the configs from |all_configs| that match the |expr|, where
894 the elements of |all_configs| are tuples of values for the |config_variables|.
895 Example:
896 >>> match_configs(expr = "(foo==1 or foo==2) and bar=='b'",
897 config_variables = ["foo", "bar"],
898 all_configs = [(1, 'a'), (1, 'b'), (2, 'a'), (2, 'b')])
899 [(1, 'b'), (2, 'b')]
900 """
901 return [
902 config for config in all_configs
903 if eval(expr, dict(zip(config_variables, config)))
904 ]
905
906
907 def verify_variables(variables):
908 """Verifies the |variables| dictionary is in the expected format."""
909 VALID_VARIABLES = [
910 KEY_TOUCHED,
911 KEY_TRACKED,
912 KEY_UNTRACKED,
913 'command',
914 'read_only',
915 ]
916 assert isinstance(variables, dict), variables
917 assert set(VALID_VARIABLES).issuperset(set(variables)), variables.keys()
918 for name, value in variables.iteritems():
919 if name == 'read_only':
920 assert value in (True, False, None), value
921 else:
922 assert isinstance(value, list), value
923 assert all(isinstance(i, basestring) for i in value), value
924
925
926 def verify_ast(expr, variables_and_values):
927 """Verifies that |expr| is of the form
928 expr ::= expr ( "or" | "and" ) expr
929 | identifier "==" ( string | int )
930 Also collects the variable identifiers and string/int values in the dict
931 |variables_and_values|, in the form {'var': set([val1, val2, ...]), ...}.
932 """
933 assert isinstance(expr, (ast.BoolOp, ast.Compare))
934 if isinstance(expr, ast.BoolOp):
935 assert isinstance(expr.op, (ast.And, ast.Or))
936 for subexpr in expr.values:
937 verify_ast(subexpr, variables_and_values)
938 else:
939 assert isinstance(expr.left.ctx, ast.Load)
940 assert len(expr.ops) == 1
941 assert isinstance(expr.ops[0], ast.Eq)
942 var_values = variables_and_values.setdefault(expr.left.id, set())
943 rhs = expr.comparators[0]
944 assert isinstance(rhs, (ast.Str, ast.Num))
945 var_values.add(rhs.n if isinstance(rhs, ast.Num) else rhs.s)
946
947
948 def verify_condition(condition, variables_and_values):
949 """Verifies the |condition| dictionary is in the expected format.
950 See verify_ast() for the meaning of |variables_and_values|.
951 """
952 VALID_INSIDE_CONDITION = ['variables']
953 assert isinstance(condition, list), condition
954 assert len(condition) == 2, condition
955 expr, then = condition
956
957 test_ast = compile(expr, '<condition>', 'eval', ast.PyCF_ONLY_AST)
958 verify_ast(test_ast.body, variables_and_values)
959
960 assert isinstance(then, dict), then
961 assert set(VALID_INSIDE_CONDITION).issuperset(set(then)), then.keys()
962 verify_variables(then['variables'])
963
964
965 def verify_root(value, variables_and_values):
966 """Verifies that |value| is the parsed form of a valid .isolate file.
967 See verify_ast() for the meaning of |variables_and_values|.
968 """
969 VALID_ROOTS = ['includes', 'conditions']
970 assert isinstance(value, dict), value
971 assert set(VALID_ROOTS).issuperset(set(value)), value.keys()
972
973 includes = value.get('includes', [])
974 assert isinstance(includes, list), includes
975 for include in includes:
976 assert isinstance(include, basestring), include
977
978 conditions = value.get('conditions', [])
979 assert isinstance(conditions, list), conditions
980 for condition in conditions:
981 verify_condition(condition, variables_and_values)
982
983
984 def remove_weak_dependencies(values, key, item, item_configs):
985 """Removes any configs from this key if the item is already under a
986 strong key.
987 """
988 if key == KEY_TOUCHED:
989 item_configs = set(item_configs)
990 for stronger_key in (KEY_TRACKED, KEY_UNTRACKED):
991 try:
992 item_configs -= values[stronger_key][item]
993 except KeyError:
994 pass
995
996 return item_configs
997
998
999 def remove_repeated_dependencies(folders, key, item, item_configs):
1000 """Removes any configs from this key if the item is in a folder that is
1001 already included."""
1002
1003 if key in (KEY_UNTRACKED, KEY_TRACKED, KEY_TOUCHED):
1004 item_configs = set(item_configs)
1005 for (folder, configs) in folders.iteritems():
1006 if folder != item and item.startswith(folder):
1007 item_configs -= configs
1008
1009 return item_configs
1010
1011
1012 def get_folders(values_dict):
1013 """Returns a dict of all the folders in the given value_dict."""
1014 return dict(
1015 (item, configs) for (item, configs) in values_dict.iteritems()
1016 if item.endswith('/')
1017 )
1018
1019
1020 def invert_map(variables):
1021 """Converts {config: {deptype: list(depvals)}} to
1022 {deptype: {depval: set(configs)}}.
1023 """
1024 KEYS = (
1025 KEY_TOUCHED,
1026 KEY_TRACKED,
1027 KEY_UNTRACKED,
1028 'command',
1029 'read_only',
1030 )
1031 out = dict((key, {}) for key in KEYS)
1032 for config, values in variables.iteritems():
1033 for key in KEYS:
1034 if key == 'command':
1035 items = [tuple(values[key])] if key in values else []
1036 elif key == 'read_only':
1037 items = [values[key]] if key in values else []
1038 else:
1039 assert key in (KEY_TOUCHED, KEY_TRACKED, KEY_UNTRACKED)
1040 items = values.get(key, [])
1041 for item in items:
1042 out[key].setdefault(item, set()).add(config)
1043 return out
1044
1045
1046 def reduce_inputs(values):
1047 """Reduces the output of invert_map() to the strictest minimum list.
1048
1049 Looks at each individual file and directory, maps where they are used and
1050 reconstructs the inverse dictionary.
1051
1052 Returns the minimized dictionary.
1053 """
1054 KEYS = (
1055 KEY_TOUCHED,
1056 KEY_TRACKED,
1057 KEY_UNTRACKED,
1058 'command',
1059 'read_only',
1060 )
1061
1062 # Folders can only live in KEY_UNTRACKED.
1063 folders = get_folders(values.get(KEY_UNTRACKED, {}))
1064
1065 out = dict((key, {}) for key in KEYS)
1066 for key in KEYS:
1067 for item, item_configs in values.get(key, {}).iteritems():
1068 item_configs = remove_weak_dependencies(values, key, item, item_configs)
1069 item_configs = remove_repeated_dependencies(
1070 folders, key, item, item_configs)
1071 if item_configs:
1072 out[key][item] = item_configs
1073 return out
1074
1075
1076 def convert_map_to_isolate_dict(values, config_variables):
1077 """Regenerates back a .isolate configuration dict from files and dirs
1078 mappings generated from reduce_inputs().
1079 """
1080 # Gather a list of configurations for set inversion later.
1081 all_mentioned_configs = set()
1082 for configs_by_item in values.itervalues():
1083 for configs in configs_by_item.itervalues():
1084 all_mentioned_configs.update(configs)
1085
1086 # Invert the mapping to make it dict first.
1087 conditions = {}
1088 for key in values:
1089 for item, configs in values[key].iteritems():
1090 then = conditions.setdefault(frozenset(configs), {})
1091 variables = then.setdefault('variables', {})
1092
1093 if item in (True, False):
1094 # One-off for read_only.
1095 variables[key] = item
1096 else:
1097 assert item
1098 if isinstance(item, tuple):
1099 # One-off for command.
1100 # Do not merge lists and do not sort!
1101 # Note that item is a tuple.
1102 assert key not in variables
1103 variables[key] = list(item)
1104 else:
1105 # The list of items (files or dirs). Append the new item and keep
1106 # the list sorted.
1107 l = variables.setdefault(key, [])
1108 l.append(item)
1109 l.sort()
1110
1111 if all_mentioned_configs:
1112 config_values = map(set, zip(*all_mentioned_configs))
1113 sef = short_expression_finder.ShortExpressionFinder(
1114 zip(config_variables, config_values))
1115
1116 conditions = sorted(
1117 [sef.get_expr(configs), then] for configs, then in conditions.iteritems())
1118 return {'conditions': conditions}
1119
1120
1121 ### Internal state files.
1122
1123
1124 class ConfigSettings(object):
1125 """Represents the dependency variables for a single build configuration.
1126 The structure is immutable.
1127 """
1128 def __init__(self, config, values):
1129 self.config = config
1130 verify_variables(values)
1131 self.touched = sorted(values.get(KEY_TOUCHED, []))
1132 self.tracked = sorted(values.get(KEY_TRACKED, []))
1133 self.untracked = sorted(values.get(KEY_UNTRACKED, []))
1134 self.command = values.get('command', [])[:]
1135 self.read_only = values.get('read_only')
1136
1137 def union(self, rhs):
1138 assert not (self.config and rhs.config) or (self.config == rhs.config)
1139 assert not (self.command and rhs.command) or (self.command == rhs.command)
1140 var = {
1141 KEY_TOUCHED: sorted(self.touched + rhs.touched),
1142 KEY_TRACKED: sorted(self.tracked + rhs.tracked),
1143 KEY_UNTRACKED: sorted(self.untracked + rhs.untracked),
1144 'command': self.command or rhs.command,
1145 'read_only': rhs.read_only if self.read_only is None else self.read_only,
1146 }
1147 return ConfigSettings(self.config or rhs.config, var)
1148
1149 def flatten(self):
1150 out = {}
1151 if self.command:
1152 out['command'] = self.command
1153 if self.touched:
1154 out[KEY_TOUCHED] = self.touched
1155 if self.tracked:
1156 out[KEY_TRACKED] = self.tracked
1157 if self.untracked:
1158 out[KEY_UNTRACKED] = self.untracked
1159 if self.read_only is not None:
1160 out['read_only'] = self.read_only
1161 return out
1162
1163
1164 class Configs(object):
1165 """Represents a processed .isolate file.
1166
1167 Stores the file in a processed way, split by configuration.
1168 """
1169 def __init__(self, file_comment):
1170 self.file_comment = file_comment
1171 # The keys of by_config are tuples of values for the configuration
1172 # variables. The names of the variables (which must be the same for
1173 # every by_config key) are kept in config_variables. Initially by_config
1174 # is empty and we don't know what configuration variables will be used,
1175 # so config_variables also starts out empty. It will be set by the first
1176 # call to union() or merge_dependencies().
1177 self.by_config = {}
1178 self.config_variables = ()
1179
1180 def union(self, rhs):
1181 """Adds variables from rhs (a Configs) to the existing variables.
1182 """
1183 config_variables = self.config_variables
1184 if not config_variables:
1185 config_variables = rhs.config_variables
1186 else:
1187 # We can't proceed if this isn't true since we don't know the correct
1188 # default values for extra variables. The variables are sorted so we
1189 # don't need to worry about permutations.
1190 if rhs.config_variables and rhs.config_variables != config_variables:
1191 raise ExecutionError(
1192 'Variables in merged .isolate files do not match: %r and %r' % (
1193 config_variables, rhs.config_variables))
1194
1195 # Takes the first file comment, prefering lhs.
1196 out = Configs(self.file_comment or rhs.file_comment)
1197 out.config_variables = config_variables
1198 for config in set(self.by_config) | set(rhs.by_config):
1199 out.by_config[config] = union(
1200 self.by_config.get(config), rhs.by_config.get(config))
1201 return out
1202
1203 def merge_dependencies(self, values, config_variables, configs):
1204 """Adds new dependencies to this object for the given configurations.
1205 Arguments:
1206 values: A variables dict as found in a .isolate file, e.g.,
1207 {KEY_TOUCHED: [...], 'command': ...}.
1208 config_variables: An ordered list of configuration variables, e.g.,
1209 ["OS", "chromeos"]. If this object already contains any dependencies,
1210 the configuration variables must match.
1211 configs: a list of tuples of values of the configuration variables,
1212 e.g., [("mac", 0), ("linux", 1)]. The dependencies in |values|
1213 are added to all of these configurations, and other configurations
1214 are unchanged.
1215 """
1216 if not values:
1217 return
1218
1219 if not self.config_variables:
1220 self.config_variables = config_variables
1221 else:
1222 # See comment in Configs.union().
1223 assert self.config_variables == config_variables
1224
1225 for config in configs:
1226 self.by_config[config] = union(
1227 self.by_config.get(config), ConfigSettings(config, values))
1228
1229 def flatten(self):
1230 """Returns a flat dictionary representation of the configuration.
1231 """
1232 return dict((k, v.flatten()) for k, v in self.by_config.iteritems())
1233
1234 def make_isolate_file(self):
1235 """Returns a dictionary suitable for writing to a .isolate file.
1236 """
1237 dependencies_by_config = self.flatten()
1238 configs_by_dependency = reduce_inputs(invert_map(dependencies_by_config))
1239 return convert_map_to_isolate_dict(configs_by_dependency,
1240 self.config_variables)
1241
1242
1243 # TODO(benrg): Remove this function when no old-format files are left.
1244 def convert_old_to_new_format(value):
1245 """Converts from the old .isolate format, which only has one variable (OS),
1246 always includes 'linux', 'mac' and 'win' in the set of valid values for OS,
1247 and allows conditions that depend on the set of all OSes, to the new format,
1248 which allows any set of variables, has no hardcoded values, and only allows
1249 explicit positive tests of variable values.
1250 """
1251 conditions = value.get('conditions', [])
1252 if 'variables' not in value and all(len(cond) == 2 for cond in conditions):
1253 return value # Nothing to change
1254
1255 def parse_condition(cond):
1256 return re.match(r'OS=="(\w+)"\Z', cond[0]).group(1)
1257
1258 oses = set(map(parse_condition, conditions))
1259 default_oses = set(['linux', 'mac', 'win'])
1260 oses = sorted(oses | default_oses)
1261
1262 def if_not_os(not_os, then):
1263 expr = ' or '.join('OS=="%s"' % os for os in oses if os != not_os)
1264 return [expr, then]
1265
1266 conditions = [
1267 cond[:2] for cond in conditions if cond[1]
1268 ] + [
1269 if_not_os(parse_condition(cond), cond[2])
1270 for cond in conditions if len(cond) == 3
1271 ]
1272
1273 if 'variables' in value:
1274 conditions.append(if_not_os(None, {'variables': value.pop('variables')}))
1275 conditions.sort()
1276
1277 value = value.copy()
1278 value['conditions'] = conditions
1279 return value
1280
1281
1282 def load_isolate_as_config(isolate_dir, value, file_comment):
1283 """Parses one .isolate file and returns a Configs() instance.
1284
1285 |value| is the loaded dictionary that was defined in the gyp file.
1286
1287 The expected format is strict, anything diverting from the format below will
1288 throw an assert:
1289 {
1290 'includes': [
1291 'foo.isolate',
1292 ],
1293 'conditions': [
1294 ['OS=="vms" and foo=42', {
1295 'variables': {
1296 'command': [
1297 ...
1298 ],
1299 'isolate_dependency_tracked': [
1300 ...
1301 ],
1302 'isolate_dependency_untracked': [
1303 ...
1304 ],
1305 'read_only': False,
1306 },
1307 }],
1308 ...
1309 ],
1310 }
1311 """
1312 value = convert_old_to_new_format(value)
1313
1314 variables_and_values = {}
1315 verify_root(value, variables_and_values)
1316 if variables_and_values:
1317 config_variables, config_values = zip(
1318 *sorted(variables_and_values.iteritems()))
1319 all_configs = list(itertools.product(*config_values))
1320 else:
1321 config_variables = None
1322 all_configs = []
1323
1324 isolate = Configs(file_comment)
1325
1326 # Add configuration-specific variables.
1327 for expr, then in value.get('conditions', []):
1328 configs = match_configs(expr, config_variables, all_configs)
1329 isolate.merge_dependencies(then['variables'], config_variables, configs)
1330
1331 # Load the includes.
1332 for include in value.get('includes', []):
1333 if os.path.isabs(include):
1334 raise ExecutionError(
1335 'Failed to load configuration; absolute include path \'%s\'' %
1336 include)
1337 included_isolate = os.path.normpath(os.path.join(isolate_dir, include))
1338 with open(included_isolate, 'r') as f:
1339 included_isolate = load_isolate_as_config(
1340 os.path.dirname(included_isolate),
1341 eval_content(f.read()),
1342 None)
1343 isolate = union(isolate, included_isolate)
1344
1345 return isolate
1346
1347
1348 def load_isolate_for_config(isolate_dir, content, variables):
1349 """Loads the .isolate file and returns the information unprocessed but
1350 filtered for the specific OS.
1351
1352 Returns the command, dependencies and read_only flag. The dependencies are
1353 fixed to use os.path.sep.
1354 """
1355 # Load the .isolate file, process its conditions, retrieve the command and
1356 # dependencies.
1357 isolate = load_isolate_as_config(isolate_dir, eval_content(content), None)
1358 try:
1359 config_name = tuple(variables[var] for var in isolate.config_variables)
1360 except KeyError:
1361 raise ExecutionError(
1362 'These configuration variables were missing from the command line: %s' %
1363 ', '.join(sorted(set(isolate.config_variables) - set(variables))))
1364 config = isolate.by_config.get(config_name)
1365 if not config:
1366 raise ExecutionError(
1367 'Failed to load configuration for variable \'%s\' for config(s) \'%s\''
1368 '\nAvailable configs: %s' %
1369 (', '.join(isolate.config_variables),
1370 ', '.join(config_name),
1371 ', '.join(str(s) for s in isolate.by_config)))
1372 # Merge tracked and untracked variables, isolate.py doesn't care about the
1373 # trackability of the variables, only the build tool does.
1374 dependencies = [
1375 f.replace('/', os.path.sep) for f in config.tracked + config.untracked
1376 ]
1377 touched = [f.replace('/', os.path.sep) for f in config.touched]
1378 return config.command, dependencies, touched, config.read_only
1379
1380
1381 def save_isolated(isolated, data):
1382 """Writes one or multiple .isolated files.
1383
1384 Note: this reference implementation does not create child .isolated file so it
1385 always returns an empty list.
1386
1387 Returns the list of child isolated files that are included by |isolated|.
1388 """
1389 trace_inputs.write_json(isolated, data, True)
1390 return []
1391
1392
1393 def chromium_save_isolated(isolated, data, variables, algo):
1394 """Writes one or many .isolated files.
1395
1396 This slightly increases the cold cache cost but greatly reduce the warm cache
1397 cost by splitting low-churn files off the master .isolated file. It also
1398 reduces overall isolateserver memcache consumption.
1399 """
1400 slaves = []
1401
1402 def extract_into_included_isolated(prefix):
1403 new_slave = {
1404 'algo': data['algo'],
1405 'files': {},
1406 'os': data['os'],
1407 'version': data['version'],
1408 }
1409 for f in data['files'].keys():
1410 if f.startswith(prefix):
1411 new_slave['files'][f] = data['files'].pop(f)
1412 if new_slave['files']:
1413 slaves.append(new_slave)
1414
1415 # Split test/data/ in its own .isolated file.
1416 extract_into_included_isolated(os.path.join('test', 'data', ''))
1417
1418 # Split everything out of PRODUCT_DIR in its own .isolated file.
1419 if variables.get('PRODUCT_DIR'):
1420 extract_into_included_isolated(variables['PRODUCT_DIR'])
1421
1422 files = []
1423 for index, f in enumerate(slaves):
1424 slavepath = isolated[:-len('.isolated')] + '.%d.isolated' % index
1425 trace_inputs.write_json(slavepath, f, True)
1426 data.setdefault('includes', []).append(
1427 isolateserver.hash_file(slavepath, algo))
1428 files.append(os.path.basename(slavepath))
1429
1430 files.extend(save_isolated(isolated, data))
1431 return files
1432
1433
1434 class Flattenable(object):
1435 """Represents data that can be represented as a json file."""
1436 MEMBERS = ()
1437
1438 def flatten(self):
1439 """Returns a json-serializable version of itself.
1440
1441 Skips None entries.
1442 """
1443 items = ((member, getattr(self, member)) for member in self.MEMBERS)
1444 return dict((member, value) for member, value in items if value is not None)
1445
1446 @classmethod
1447 def load(cls, data, *args, **kwargs):
1448 """Loads a flattened version."""
1449 data = data.copy()
1450 out = cls(*args, **kwargs)
1451 for member in out.MEMBERS:
1452 if member in data:
1453 # Access to a protected member XXX of a client class
1454 # pylint: disable=W0212
1455 out._load_member(member, data.pop(member))
1456 if data:
1457 raise ValueError(
1458 'Found unexpected entry %s while constructing an object %s' %
1459 (data, cls.__name__), data, cls.__name__)
1460 return out
1461
1462 def _load_member(self, member, value):
1463 """Loads a member into self."""
1464 setattr(self, member, value)
1465
1466 @classmethod
1467 def load_file(cls, filename, *args, **kwargs):
1468 """Loads the data from a file or return an empty instance."""
1469 try:
1470 out = cls.load(trace_inputs.read_json(filename), *args, **kwargs)
1471 logging.debug('Loaded %s(%s)', cls.__name__, filename)
1472 except (IOError, ValueError) as e:
1473 # On failure, loads the default instance.
1474 out = cls(*args, **kwargs)
1475 logging.warn('Failed to load %s: %s', filename, e)
1476 return out
1477
1478
1479 class SavedState(Flattenable):
1480 """Describes the content of a .state file.
1481
1482 This file caches the items calculated by this script and is used to increase
1483 the performance of the script. This file is not loaded by run_isolated.py.
1484 This file can always be safely removed.
1485
1486 It is important to note that the 'files' dict keys are using native OS path
1487 separator instead of '/' used in .isolate file.
1488 """
1489 MEMBERS = (
1490 # Algorithm used to generate the hash. The only supported value is at the
1491 # time of writting 'sha-1'.
1492 'algo',
1493 # Cache of the processed command. This value is saved because .isolated
1494 # files are never loaded by isolate.py so it's the only way to load the
1495 # command safely.
1496 'command',
1497 # Cache of the files found so the next run can skip hash calculation.
1498 'files',
1499 # Path of the original .isolate file. Relative path to isolated_basedir.
1500 'isolate_file',
1501 # List of included .isolated files. Used to support/remember 'slave'
1502 # .isolated files. Relative path to isolated_basedir.
1503 'child_isolated_files',
1504 # If the generated directory tree should be read-only.
1505 'read_only',
1506 # Relative cwd to use to start the command.
1507 'relative_cwd',
1508 # GYP variables used to generate the .isolated file. Variables are saved so
1509 # a user can use isolate.py after building and the GYP variables are still
1510 # defined.
1511 'variables',
1512 # Version of the file format in format 'major.minor'. Any non-breaking
1513 # change must update minor. Any breaking change must update major.
1514 'version',
1515 )
1516
1517 def __init__(self, isolated_basedir):
1518 """Creates an empty SavedState.
1519
1520 |isolated_basedir| is the directory where the .isolated and .isolated.state
1521 files are saved.
1522 """
1523 super(SavedState, self).__init__()
1524 assert os.path.isabs(isolated_basedir), isolated_basedir
1525 assert os.path.isdir(isolated_basedir), isolated_basedir
1526 self.isolated_basedir = isolated_basedir
1527
1528 # The default algorithm used.
1529 self.algo = isolateserver.SUPPORTED_ALGOS['sha-1']
1530 self.command = []
1531 self.files = {}
1532 self.isolate_file = None
1533 self.child_isolated_files = []
1534 self.read_only = None
1535 self.relative_cwd = None
1536 self.variables = {'OS': get_flavor()}
1537 # The current version.
1538 self.version = '1.0'
1539
1540 def update(self, isolate_file, variables):
1541 """Updates the saved state with new data to keep GYP variables and internal
1542 reference to the original .isolate file.
1543 """
1544 assert os.path.isabs(isolate_file)
1545 # Convert back to a relative path. On Windows, if the isolate and
1546 # isolated files are on different drives, isolate_file will stay an absolute
1547 # path.
1548 isolate_file = safe_relpath(isolate_file, self.isolated_basedir)
1549
1550 # The same .isolate file should always be used to generate the .isolated and
1551 # .isolated.state.
1552 assert isolate_file == self.isolate_file or not self.isolate_file, (
1553 isolate_file, self.isolate_file)
1554 self.isolate_file = isolate_file
1555 self.variables.update(variables)
1556
1557 def update_isolated(self, command, infiles, touched, read_only, relative_cwd):
1558 """Updates the saved state with data necessary to generate a .isolated file.
1559
1560 The new files in |infiles| are added to self.files dict but their hash is
1561 not calculated here.
1562 """
1563 self.command = command
1564 # Add new files.
1565 for f in infiles:
1566 self.files.setdefault(f, {})
1567 for f in touched:
1568 self.files.setdefault(f, {})['T'] = True
1569 # Prune extraneous files that are not a dependency anymore.
1570 for f in set(self.files).difference(set(infiles).union(touched)):
1571 del self.files[f]
1572 if read_only is not None:
1573 self.read_only = read_only
1574 self.relative_cwd = relative_cwd
1575
1576 def to_isolated(self):
1577 """Creates a .isolated dictionary out of the saved state.
1578
1579 https://code.google.com/p/swarming/wiki/IsolatedDesign
1580 """
1581 def strip(data):
1582 """Returns a 'files' entry with only the whitelisted keys."""
1583 return dict((k, data[k]) for k in ('h', 'l', 'm', 's') if k in data)
1584
1585 out = {
1586 'algo': isolateserver.SUPPORTED_ALGOS_REVERSE[self.algo],
1587 'files': dict(
1588 (filepath, strip(data)) for filepath, data in self.files.iteritems()),
1589 'os': self.variables['OS'],
1590 'version': self.version,
1591 }
1592 if self.command:
1593 out['command'] = self.command
1594 if self.read_only is not None:
1595 out['read_only'] = self.read_only
1596 if self.relative_cwd:
1597 out['relative_cwd'] = self.relative_cwd
1598 return out
1599
1600 @property
1601 def isolate_filepath(self):
1602 """Returns the absolute path of self.isolate_file."""
1603 return os.path.normpath(
1604 os.path.join(self.isolated_basedir, self.isolate_file))
1605
1606 # Arguments number differs from overridden method
1607 @classmethod
1608 def load(cls, data, isolated_basedir): # pylint: disable=W0221
1609 """Special case loading to disallow different OS.
1610
1611 It is not possible to load a .isolated.state files from a different OS, this
1612 file is saved in OS-specific format.
1613 """
1614 out = super(SavedState, cls).load(data, isolated_basedir)
1615 if 'os' in data:
1616 out.variables['OS'] = data['os']
1617
1618 # Converts human readable form back into the proper class type.
1619 algo = data.get('algo', 'sha-1')
1620 if not algo in isolateserver.SUPPORTED_ALGOS:
1621 raise isolateserver.ConfigError('Unknown algo \'%s\'' % out.algo)
1622 out.algo = isolateserver.SUPPORTED_ALGOS[algo]
1623
1624 # For example, 1.1 is guaranteed to be backward compatible with 1.0 code.
1625 if not re.match(r'^(\d+)\.(\d+)$', out.version):
1626 raise isolateserver.ConfigError('Unknown version \'%s\'' % out.version)
1627 if out.version.split('.', 1)[0] != '1':
1628 raise isolateserver.ConfigError(
1629 'Unsupported version \'%s\'' % out.version)
1630
1631 # The .isolate file must be valid. It could be absolute on Windows if the
1632 # drive containing the .isolate and the drive containing the .isolated files
1633 # differ.
1634 assert not os.path.isabs(out.isolate_file) or sys.platform == 'win32'
1635 assert os.path.isfile(out.isolate_filepath), out.isolate_filepath
1636 return out
1637
1638 def flatten(self):
1639 """Makes sure 'algo' is in human readable form."""
1640 out = super(SavedState, self).flatten()
1641 out['algo'] = isolateserver.SUPPORTED_ALGOS_REVERSE[out['algo']]
1642 return out
1643
1644 def __str__(self):
1645 out = '%s(\n' % self.__class__.__name__
1646 out += ' command: %s\n' % self.command
1647 out += ' files: %d\n' % len(self.files)
1648 out += ' isolate_file: %s\n' % self.isolate_file
1649 out += ' read_only: %s\n' % self.read_only
1650 out += ' relative_cwd: %s\n' % self.relative_cwd
1651 out += ' child_isolated_files: %s\n' % self.child_isolated_files
1652 out += ' variables: %s' % ''.join(
1653 '\n %s=%s' % (k, self.variables[k]) for k in sorted(self.variables))
1654 out += ')'
1655 return out
1656
1657
1658 class CompleteState(object):
1659 """Contains all the state to run the task at hand."""
1660 def __init__(self, isolated_filepath, saved_state):
1661 super(CompleteState, self).__init__()
1662 assert isolated_filepath is None or os.path.isabs(isolated_filepath)
1663 self.isolated_filepath = isolated_filepath
1664 # Contains the data to ease developer's use-case but that is not strictly
1665 # necessary.
1666 self.saved_state = saved_state
1667
1668 @classmethod
1669 def load_files(cls, isolated_filepath):
1670 """Loads state from disk."""
1671 assert os.path.isabs(isolated_filepath), isolated_filepath
1672 isolated_basedir = os.path.dirname(isolated_filepath)
1673 return cls(
1674 isolated_filepath,
1675 SavedState.load_file(
1676 isolatedfile_to_state(isolated_filepath), isolated_basedir))
1677
1678 def load_isolate(self, cwd, isolate_file, variables, ignore_broken_items):
1679 """Updates self.isolated and self.saved_state with information loaded from a
1680 .isolate file.
1681
1682 Processes the loaded data, deduce root_dir, relative_cwd.
1683 """
1684 # Make sure to not depend on os.getcwd().
1685 assert os.path.isabs(isolate_file), isolate_file
1686 isolate_file = file_path.get_native_path_case(isolate_file)
1687 logging.info(
1688 'CompleteState.load_isolate(%s, %s, %s, %s)',
1689 cwd, isolate_file, variables, ignore_broken_items)
1690 relative_base_dir = os.path.dirname(isolate_file)
1691
1692 # Processes the variables and update the saved state.
1693 variables = process_variables(cwd, variables, relative_base_dir)
1694 self.saved_state.update(isolate_file, variables)
1695 variables = self.saved_state.variables
1696
1697 with open(isolate_file, 'r') as f:
1698 # At that point, variables are not replaced yet in command and infiles.
1699 # infiles may contain directory entries and is in posix style.
1700 command, infiles, touched, read_only = load_isolate_for_config(
1701 os.path.dirname(isolate_file), f.read(), variables)
1702 command = [eval_variables(i, variables) for i in command]
1703 infiles = [eval_variables(f, variables) for f in infiles]
1704 touched = [eval_variables(f, variables) for f in touched]
1705 # root_dir is automatically determined by the deepest root accessed with the
1706 # form '../../foo/bar'. Note that path variables must be taken in account
1707 # too, add them as if they were input files.
1708 path_variables = [variables[v] for v in PATH_VARIABLES if v in variables]
1709 root_dir = determine_root_dir(
1710 relative_base_dir, infiles + touched + path_variables)
1711 # The relative directory is automatically determined by the relative path
1712 # between root_dir and the directory containing the .isolate file,
1713 # isolate_base_dir.
1714 relative_cwd = os.path.relpath(relative_base_dir, root_dir)
1715 # Now that we know where the root is, check that the PATH_VARIABLES point
1716 # inside it.
1717 for i in PATH_VARIABLES:
1718 if i in variables:
1719 if not path_starts_with(
1720 root_dir, os.path.join(relative_base_dir, variables[i])):
1721 raise isolateserver.MappingError(
1722 'Path variable %s=%r points outside the inferred root directory'
1723 ' %s' % (i, variables[i], root_dir))
1724 # Normalize the files based to root_dir. It is important to keep the
1725 # trailing os.path.sep at that step.
1726 infiles = [
1727 relpath(normpath(os.path.join(relative_base_dir, f)), root_dir)
1728 for f in infiles
1729 ]
1730 touched = [
1731 relpath(normpath(os.path.join(relative_base_dir, f)), root_dir)
1732 for f in touched
1733 ]
1734 follow_symlinks = variables['OS'] != 'win'
1735 # Expand the directories by listing each file inside. Up to now, trailing
1736 # os.path.sep must be kept. Do not expand 'touched'.
1737 infiles = expand_directories_and_symlinks(
1738 root_dir,
1739 infiles,
1740 lambda x: re.match(r'.*\.(git|svn|pyc)$', x),
1741 follow_symlinks,
1742 ignore_broken_items)
1743
1744 # If we ignore broken items then remove any missing touched items.
1745 if ignore_broken_items:
1746 original_touched_count = len(touched)
1747 touched = [touch for touch in touched if os.path.exists(touch)]
1748
1749 if len(touched) != original_touched_count:
1750 logging.info('Removed %d invalid touched entries',
1751 len(touched) - original_touched_count)
1752
1753 # Finally, update the new data to be able to generate the foo.isolated file,
1754 # the file that is used by run_isolated.py.
1755 self.saved_state.update_isolated(
1756 command, infiles, touched, read_only, relative_cwd)
1757 logging.debug(self)
1758
1759 def process_inputs(self, subdir):
1760 """Updates self.saved_state.files with the files' mode and hash.
1761
1762 If |subdir| is specified, filters to a subdirectory. The resulting .isolated
1763 file is tainted.
1764
1765 See process_input() for more information.
1766 """
1767 for infile in sorted(self.saved_state.files):
1768 if subdir and not infile.startswith(subdir):
1769 self.saved_state.files.pop(infile)
1770 else:
1771 filepath = os.path.join(self.root_dir, infile)
1772 self.saved_state.files[infile] = process_input(
1773 filepath,
1774 self.saved_state.files[infile],
1775 self.saved_state.read_only,
1776 self.saved_state.variables['OS'],
1777 self.saved_state.algo)
1778
1779 def save_files(self):
1780 """Saves self.saved_state and creates a .isolated file."""
1781 logging.debug('Dumping to %s' % self.isolated_filepath)
1782 self.saved_state.child_isolated_files = chromium_save_isolated(
1783 self.isolated_filepath,
1784 self.saved_state.to_isolated(),
1785 self.saved_state.variables,
1786 self.saved_state.algo)
1787 total_bytes = sum(
1788 i.get('s', 0) for i in self.saved_state.files.itervalues())
1789 if total_bytes:
1790 # TODO(maruel): Stats are missing the .isolated files.
1791 logging.debug('Total size: %d bytes' % total_bytes)
1792 saved_state_file = isolatedfile_to_state(self.isolated_filepath)
1793 logging.debug('Dumping to %s' % saved_state_file)
1794 trace_inputs.write_json(saved_state_file, self.saved_state.flatten(), True)
1795
1796 @property
1797 def root_dir(self):
1798 """Returns the absolute path of the root_dir to reference the .isolate file
1799 via relative_cwd.
1800
1801 So that join(root_dir, relative_cwd, basename(isolate_file)) is equivalent
1802 to isolate_filepath.
1803 """
1804 if not self.saved_state.isolate_file:
1805 raise ExecutionError('Please specify --isolate')
1806 isolate_dir = os.path.dirname(self.saved_state.isolate_filepath)
1807 # Special case '.'.
1808 if self.saved_state.relative_cwd == '.':
1809 root_dir = isolate_dir
1810 else:
1811 if not isolate_dir.endswith(self.saved_state.relative_cwd):
1812 raise ExecutionError(
1813 ('Make sure the .isolate file is in the directory that will be '
1814 'used as the relative directory. It is currently in %s and should '
1815 'be in %s') % (isolate_dir, self.saved_state.relative_cwd))
1816 # Walk back back to the root directory.
1817 root_dir = isolate_dir[:-(len(self.saved_state.relative_cwd) + 1)]
1818 return file_path.get_native_path_case(root_dir)
1819
1820 @property
1821 def resultdir(self):
1822 """Returns the absolute path containing the .isolated file.
1823
1824 It is usually equivalent to the variable PRODUCT_DIR. Uses the .isolated
1825 path as the value.
1826 """
1827 return os.path.dirname(self.isolated_filepath)
1828
1829 def __str__(self):
1830 def indent(data, indent_length):
1831 """Indents text."""
1832 spacing = ' ' * indent_length
1833 return ''.join(spacing + l for l in str(data).splitlines(True))
1834
1835 out = '%s(\n' % self.__class__.__name__
1836 out += ' root_dir: %s\n' % self.root_dir
1837 out += ' saved_state: %s)' % indent(self.saved_state, 2)
1838 return out
1839
1840
1841 def load_complete_state(options, cwd, subdir, skip_update):
1842 """Loads a CompleteState.
1843
1844 This includes data from .isolate and .isolated.state files. Never reads the
1845 .isolated file.
1846
1847 Arguments:
1848 options: Options instance generated with OptionParserIsolate. For either
1849 options.isolate and options.isolated, if the value is set, it is an
1850 absolute path.
1851 cwd: base directory to be used when loading the .isolate file.
1852 subdir: optional argument to only process file in the subdirectory, relative
1853 to CompleteState.root_dir.
1854 skip_update: Skip trying to load the .isolate file and processing the
1855 dependencies. It is useful when not needed, like when tracing.
1856 """
1857 assert not options.isolate or os.path.isabs(options.isolate)
1858 assert not options.isolated or os.path.isabs(options.isolated)
1859 cwd = file_path.get_native_path_case(unicode(cwd))
1860 if options.isolated:
1861 # Load the previous state if it was present. Namely, "foo.isolated.state".
1862 # Note: this call doesn't load the .isolate file.
1863 complete_state = CompleteState.load_files(options.isolated)
1864 else:
1865 # Constructs a dummy object that cannot be saved. Useful for temporary
1866 # commands like 'run'.
1867 complete_state = CompleteState(None, SavedState())
1868
1869 if not options.isolate:
1870 if not complete_state.saved_state.isolate_file:
1871 if not skip_update:
1872 raise ExecutionError('A .isolate file is required.')
1873 isolate = None
1874 else:
1875 isolate = complete_state.saved_state.isolate_filepath
1876 else:
1877 isolate = options.isolate
1878 if complete_state.saved_state.isolate_file:
1879 rel_isolate = safe_relpath(
1880 options.isolate, complete_state.saved_state.isolated_basedir)
1881 if rel_isolate != complete_state.saved_state.isolate_file:
1882 raise ExecutionError(
1883 '%s and %s do not match.' % (
1884 options.isolate, complete_state.saved_state.isolate_file))
1885
1886 if not skip_update:
1887 # Then load the .isolate and expands directories.
1888 complete_state.load_isolate(
1889 cwd, isolate, options.variables, options.ignore_broken_items)
1890
1891 # Regenerate complete_state.saved_state.files.
1892 if subdir:
1893 subdir = unicode(subdir)
1894 subdir = eval_variables(subdir, complete_state.saved_state.variables)
1895 subdir = subdir.replace('/', os.path.sep)
1896
1897 if not skip_update:
1898 complete_state.process_inputs(subdir)
1899 return complete_state
1900
1901
1902 def read_trace_as_isolate_dict(complete_state, trace_blacklist):
1903 """Reads a trace and returns the .isolate dictionary.
1904
1905 Returns exceptions during the log parsing so it can be re-raised.
1906 """
1907 api = trace_inputs.get_api()
1908 logfile = complete_state.isolated_filepath + '.log'
1909 if not os.path.isfile(logfile):
1910 raise ExecutionError(
1911 'No log file \'%s\' to read, did you forget to \'trace\'?' % logfile)
1912 try:
1913 data = api.parse_log(logfile, trace_blacklist, None)
1914 exceptions = [i['exception'] for i in data if 'exception' in i]
1915 results = (i['results'] for i in data if 'results' in i)
1916 results_stripped = (i.strip_root(complete_state.root_dir) for i in results)
1917 files = set(sum((result.existent for result in results_stripped), []))
1918 tracked, touched = split_touched(files)
1919 value = generate_isolate(
1920 tracked,
1921 [],
1922 touched,
1923 complete_state.root_dir,
1924 complete_state.saved_state.variables,
1925 complete_state.saved_state.relative_cwd,
1926 trace_blacklist)
1927 return value, exceptions
1928 except trace_inputs.TracingFailure, e:
1929 raise ExecutionError(
1930 'Reading traces failed for: %s\n%s' %
1931 (' '.join(complete_state.saved_state.command), str(e)))
1932
1933
1934 def print_all(comment, data, stream):
1935 """Prints a complete .isolate file and its top-level file comment into a
1936 stream.
1937 """
1938 if comment:
1939 stream.write(comment)
1940 pretty_print(data, stream)
1941
1942
1943 def merge(complete_state, trace_blacklist):
1944 """Reads a trace and merges it back into the source .isolate file."""
1945 value, exceptions = read_trace_as_isolate_dict(
1946 complete_state, trace_blacklist)
1947
1948 # Now take that data and union it into the original .isolate file.
1949 with open(complete_state.saved_state.isolate_filepath, 'r') as f:
1950 prev_content = f.read()
1951 isolate_dir = os.path.dirname(complete_state.saved_state.isolate_filepath)
1952 prev_config = load_isolate_as_config(
1953 isolate_dir,
1954 eval_content(prev_content),
1955 extract_comment(prev_content))
1956 new_config = load_isolate_as_config(isolate_dir, value, '')
1957 config = union(prev_config, new_config)
1958 data = config.make_isolate_file()
1959 print('Updating %s' % complete_state.saved_state.isolate_file)
1960 with open(complete_state.saved_state.isolate_filepath, 'wb') as f:
1961 print_all(config.file_comment, data, f)
1962 if exceptions:
1963 # It got an exception, raise the first one.
1964 raise \
1965 exceptions[0][0], \
1966 exceptions[0][1], \
1967 exceptions[0][2]
1968
1969
1970 ### Commands.
1971
1972
1973 def CMDarchive(parser, args):
1974 """Creates a .isolated file and uploads the tree to an isolate server.
1975
1976 All the files listed in the .isolated file are put in the isolate server
1977 cache via isolateserver.py.
1978 """
1979 parser.add_option('--subdir', help='Filters to a subdirectory')
1980 options, args = parser.parse_args(args)
1981 if args:
1982 parser.error('Unsupported argument: %s' % args)
1983
1984 with tools.Profiler('GenerateHashtable'):
1985 success = False
1986 try:
1987 complete_state = load_complete_state(
1988 options, os.getcwd(), options.subdir, False)
1989 if not options.outdir:
1990 options.outdir = os.path.join(
1991 os.path.dirname(complete_state.isolated_filepath), 'hashtable')
1992 # Make sure that complete_state isn't modified until save_files() is
1993 # called, because any changes made to it here will propagate to the files
1994 # created (which is probably not intended).
1995 complete_state.save_files()
1996
1997 infiles = complete_state.saved_state.files
1998 # Add all the .isolated files.
1999 isolated_hash = []
2000 isolated_files = [
2001 options.isolated,
2002 ] + complete_state.saved_state.child_isolated_files
2003 for item in isolated_files:
2004 item_path = os.path.join(
2005 os.path.dirname(complete_state.isolated_filepath), item)
2006 # Do not use isolateserver.hash_file() here because the file is
2007 # likely smallish (under 500kb) and its file size is needed.
2008 with open(item_path, 'rb') as f:
2009 content = f.read()
2010 isolated_hash.append(
2011 complete_state.saved_state.algo(content).hexdigest())
2012 isolated_metadata = {
2013 'h': isolated_hash[-1],
2014 's': len(content),
2015 'priority': '0'
2016 }
2017 infiles[item_path] = isolated_metadata
2018
2019 logging.info('Creating content addressed object store with %d item',
2020 len(infiles))
2021
2022 if is_url(options.outdir):
2023 isolateserver.upload_tree(
2024 base_url=options.outdir,
2025 indir=complete_state.root_dir,
2026 infiles=infiles,
2027 namespace='default-gzip')
2028 else:
2029 recreate_tree(
2030 outdir=options.outdir,
2031 indir=complete_state.root_dir,
2032 infiles=infiles,
2033 action=run_isolated.HARDLINK_WITH_FALLBACK,
2034 as_hash=True)
2035 success = True
2036 print('%s %s' % (isolated_hash[0], os.path.basename(options.isolated)))
2037 finally:
2038 # If the command failed, delete the .isolated file if it exists. This is
2039 # important so no stale swarm job is executed.
2040 if not success and os.path.isfile(options.isolated):
2041 os.remove(options.isolated)
2042 return not success
2043
2044
2045 def CMDcheck(parser, args):
2046 """Checks that all the inputs are present and generates .isolated."""
2047 parser.add_option('--subdir', help='Filters to a subdirectory')
2048 options, args = parser.parse_args(args)
2049 if args:
2050 parser.error('Unsupported argument: %s' % args)
2051
2052 complete_state = load_complete_state(
2053 options, os.getcwd(), options.subdir, False)
2054
2055 # Nothing is done specifically. Just store the result and state.
2056 complete_state.save_files()
2057 return 0
2058
2059
2060 CMDhashtable = CMDarchive
2061
2062
2063 def CMDmerge(parser, args):
2064 """Reads and merges the data from the trace back into the original .isolate.
2065
2066 Ignores --outdir.
2067 """
2068 parser.require_isolated = False
2069 add_trace_option(parser)
2070 options, args = parser.parse_args(args)
2071 if args:
2072 parser.error('Unsupported argument: %s' % args)
2073
2074 complete_state = load_complete_state(options, os.getcwd(), None, False)
2075 blacklist = trace_inputs.gen_blacklist(options.trace_blacklist)
2076 merge(complete_state, blacklist)
2077 return 0
2078
2079
2080 def CMDread(parser, args):
2081 """Reads the trace file generated with command 'trace'.
2082
2083 Ignores --outdir.
2084 """
2085 parser.require_isolated = False
2086 add_trace_option(parser)
2087 parser.add_option(
2088 '--skip-refresh', action='store_true',
2089 help='Skip reading .isolate file and do not refresh the hash of '
2090 'dependencies')
2091 parser.add_option(
2092 '-m', '--merge', action='store_true',
2093 help='merge the results back in the .isolate file instead of printing')
2094 options, args = parser.parse_args(args)
2095 if args:
2096 parser.error('Unsupported argument: %s' % args)
2097
2098 complete_state = load_complete_state(
2099 options, os.getcwd(), None, options.skip_refresh)
2100 blacklist = trace_inputs.gen_blacklist(options.trace_blacklist)
2101 value, exceptions = read_trace_as_isolate_dict(complete_state, blacklist)
2102 if options.merge:
2103 merge(complete_state, blacklist)
2104 else:
2105 pretty_print(value, sys.stdout)
2106
2107 if exceptions:
2108 # It got an exception, raise the first one.
2109 raise \
2110 exceptions[0][0], \
2111 exceptions[0][1], \
2112 exceptions[0][2]
2113 return 0
2114
2115
2116 def CMDremap(parser, args):
2117 """Creates a directory with all the dependencies mapped into it.
2118
2119 Useful to test manually why a test is failing. The target executable is not
2120 run.
2121 """
2122 parser.require_isolated = False
2123 options, args = parser.parse_args(args)
2124 if args:
2125 parser.error('Unsupported argument: %s' % args)
2126 complete_state = load_complete_state(options, os.getcwd(), None, False)
2127
2128 if not options.outdir:
2129 options.outdir = run_isolated.make_temp_dir(
2130 'isolate', complete_state.root_dir)
2131 else:
2132 if is_url(options.outdir):
2133 parser.error('Can\'t use url for --outdir with mode remap.')
2134 if not os.path.isdir(options.outdir):
2135 os.makedirs(options.outdir)
2136 print('Remapping into %s' % options.outdir)
2137 if len(os.listdir(options.outdir)):
2138 raise ExecutionError('Can\'t remap in a non-empty directory')
2139 recreate_tree(
2140 outdir=options.outdir,
2141 indir=complete_state.root_dir,
2142 infiles=complete_state.saved_state.files,
2143 action=run_isolated.HARDLINK_WITH_FALLBACK,
2144 as_hash=False)
2145 if complete_state.saved_state.read_only:
2146 run_isolated.make_writable(options.outdir, True)
2147
2148 if complete_state.isolated_filepath:
2149 complete_state.save_files()
2150 return 0
2151
2152
2153 def CMDrewrite(parser, args):
2154 """Rewrites a .isolate file into the canonical format."""
2155 parser.require_isolated = False
2156 options, args = parser.parse_args(args)
2157 if args:
2158 parser.error('Unsupported argument: %s' % args)
2159
2160 if options.isolated:
2161 # Load the previous state if it was present. Namely, "foo.isolated.state".
2162 complete_state = CompleteState.load_files(options.isolated)
2163 isolate = options.isolate or complete_state.saved_state.isolate_filepath
2164 else:
2165 isolate = options.isolate
2166 if not isolate:
2167 parser.error('--isolate is required.')
2168
2169 with open(isolate, 'r') as f:
2170 content = f.read()
2171 config = load_isolate_as_config(
2172 os.path.dirname(os.path.abspath(isolate)),
2173 eval_content(content),
2174 extract_comment(content))
2175 data = config.make_isolate_file()
2176 print('Updating %s' % isolate)
2177 with open(isolate, 'wb') as f:
2178 print_all(config.file_comment, data, f)
2179 return 0
2180
2181
2182 @subcommand.usage('-- [extra arguments]')
2183 def CMDrun(parser, args):
2184 """Runs the test executable in an isolated (temporary) directory.
2185
2186 All the dependencies are mapped into the temporary directory and the
2187 directory is cleaned up after the target exits. Warning: if --outdir is
2188 specified, it is deleted upon exit.
2189
2190 Argument processing stops at -- and these arguments are appended to the
2191 command line of the target to run. For example, use:
2192 isolate.py run --isolated foo.isolated -- --gtest_filter=Foo.Bar
2193 """
2194 parser.require_isolated = False
2195 parser.add_option(
2196 '--skip-refresh', action='store_true',
2197 help='Skip reading .isolate file and do not refresh the hash of '
2198 'dependencies')
2199 options, args = parser.parse_args(args)
2200 if options.outdir and is_url(options.outdir):
2201 parser.error('Can\'t use url for --outdir with mode run.')
2202
2203 complete_state = load_complete_state(
2204 options, os.getcwd(), None, options.skip_refresh)
2205 cmd = complete_state.saved_state.command + args
2206 if not cmd:
2207 raise ExecutionError('No command to run.')
2208
2209 cmd = tools.fix_python_path(cmd)
2210 try:
2211 root_dir = complete_state.root_dir
2212 if not options.outdir:
2213 if not os.path.isabs(root_dir):
2214 root_dir = os.path.join(os.path.dirname(options.isolated), root_dir)
2215 options.outdir = run_isolated.make_temp_dir('isolate', root_dir)
2216 else:
2217 if not os.path.isdir(options.outdir):
2218 os.makedirs(options.outdir)
2219 recreate_tree(
2220 outdir=options.outdir,
2221 indir=root_dir,
2222 infiles=complete_state.saved_state.files,
2223 action=run_isolated.HARDLINK_WITH_FALLBACK,
2224 as_hash=False)
2225 cwd = os.path.normpath(
2226 os.path.join(options.outdir, complete_state.saved_state.relative_cwd))
2227 if not os.path.isdir(cwd):
2228 # It can happen when no files are mapped from the directory containing the
2229 # .isolate file. But the directory must exist to be the current working
2230 # directory.
2231 os.makedirs(cwd)
2232 if complete_state.saved_state.read_only:
2233 run_isolated.make_writable(options.outdir, True)
2234 logging.info('Running %s, cwd=%s' % (cmd, cwd))
2235 result = subprocess.call(cmd, cwd=cwd)
2236 finally:
2237 if options.outdir:
2238 run_isolated.rmtree(options.outdir)
2239
2240 if complete_state.isolated_filepath:
2241 complete_state.save_files()
2242 return result
2243
2244
2245 @subcommand.usage('-- [extra arguments]')
2246 def CMDtrace(parser, args):
2247 """Traces the target using trace_inputs.py.
2248
2249 It runs the executable without remapping it, and traces all the files it and
2250 its child processes access. Then the 'merge' command can be used to generate
2251 an updated .isolate file out of it or the 'read' command to print it out to
2252 stdout.
2253
2254 Argument processing stops at -- and these arguments are appended to the
2255 command line of the target to run. For example, use:
2256 isolate.py trace --isolated foo.isolated -- --gtest_filter=Foo.Bar
2257 """
2258 add_trace_option(parser)
2259 parser.add_option(
2260 '-m', '--merge', action='store_true',
2261 help='After tracing, merge the results back in the .isolate file')
2262 parser.add_option(
2263 '--skip-refresh', action='store_true',
2264 help='Skip reading .isolate file and do not refresh the hash of '
2265 'dependencies')
2266 options, args = parser.parse_args(args)
2267
2268 complete_state = load_complete_state(
2269 options, os.getcwd(), None, options.skip_refresh)
2270 cmd = complete_state.saved_state.command + args
2271 if not cmd:
2272 raise ExecutionError('No command to run.')
2273 cmd = tools.fix_python_path(cmd)
2274 cwd = os.path.normpath(os.path.join(
2275 unicode(complete_state.root_dir),
2276 complete_state.saved_state.relative_cwd))
2277 cmd[0] = os.path.normpath(os.path.join(cwd, cmd[0]))
2278 if not os.path.isfile(cmd[0]):
2279 raise ExecutionError(
2280 'Tracing failed for: %s\nIt doesn\'t exit' % ' '.join(cmd))
2281 logging.info('Running %s, cwd=%s' % (cmd, cwd))
2282 api = trace_inputs.get_api()
2283 logfile = complete_state.isolated_filepath + '.log'
2284 api.clean_trace(logfile)
2285 out = None
2286 try:
2287 with api.get_tracer(logfile) as tracer:
2288 result, out = tracer.trace(
2289 cmd,
2290 cwd,
2291 'default',
2292 True)
2293 except trace_inputs.TracingFailure, e:
2294 raise ExecutionError('Tracing failed for: %s\n%s' % (' '.join(cmd), str(e)))
2295
2296 if result:
2297 logging.error(
2298 'Tracer exited with %d, which means the tests probably failed so the '
2299 'trace is probably incomplete.', result)
2300 logging.info(out)
2301
2302 complete_state.save_files()
2303
2304 if options.merge:
2305 blacklist = trace_inputs.gen_blacklist(options.trace_blacklist)
2306 merge(complete_state, blacklist)
2307
2308 return result
2309
2310
2311 def _process_variable_arg(_option, _opt, _value, parser):
2312 if not parser.rargs:
2313 raise optparse.OptionValueError(
2314 'Please use --variable FOO=BAR or --variable FOO BAR')
2315 k = parser.rargs.pop(0)
2316 if '=' in k:
2317 parser.values.variables.append(tuple(k.split('=', 1)))
2318 else:
2319 if not parser.rargs:
2320 raise optparse.OptionValueError(
2321 'Please use --variable FOO=BAR or --variable FOO BAR')
2322 v = parser.rargs.pop(0)
2323 parser.values.variables.append((k, v))
2324
2325
2326 def add_variable_option(parser):
2327 """Adds --isolated and --variable to an OptionParser."""
2328 parser.add_option(
2329 '-s', '--isolated',
2330 metavar='FILE',
2331 help='.isolated file to generate or read')
2332 # Keep for compatibility. TODO(maruel): Remove once not used anymore.
2333 parser.add_option(
2334 '-r', '--result',
2335 dest='isolated',
2336 help=optparse.SUPPRESS_HELP)
2337 default_variables = [('OS', get_flavor())]
2338 if sys.platform in ('win32', 'cygwin'):
2339 default_variables.append(('EXECUTABLE_SUFFIX', '.exe'))
2340 else:
2341 default_variables.append(('EXECUTABLE_SUFFIX', ''))
2342 parser.add_option(
2343 '-V', '--variable',
2344 action='callback',
2345 callback=_process_variable_arg,
2346 default=default_variables,
2347 dest='variables',
2348 metavar='FOO BAR',
2349 help='Variables to process in the .isolate file, default: %default. '
2350 'Variables are persistent accross calls, they are saved inside '
2351 '<.isolated>.state')
2352
2353
2354 def add_trace_option(parser):
2355 """Adds --trace-blacklist to the parser."""
2356 parser.add_option(
2357 '--trace-blacklist',
2358 action='append', default=list(DEFAULT_BLACKLIST),
2359 help='List of regexp to use as blacklist filter for files to consider '
2360 'important, not to be confused with --blacklist which blacklists '
2361 'test case.')
2362
2363
2364 def parse_isolated_option(parser, options, cwd, require_isolated):
2365 """Processes --isolated."""
2366 if options.isolated:
2367 options.isolated = os.path.normpath(
2368 os.path.join(cwd, options.isolated.replace('/', os.path.sep)))
2369 if require_isolated and not options.isolated:
2370 parser.error('--isolated is required.')
2371 if options.isolated and not options.isolated.endswith('.isolated'):
2372 parser.error('--isolated value must end with \'.isolated\'')
2373
2374
2375 def parse_variable_option(options):
2376 """Processes --variable."""
2377 # TODO(benrg): Maybe we should use a copy of gyp's NameValueListToDict here,
2378 # but it wouldn't be backward compatible.
2379 def try_make_int(s):
2380 """Converts a value to int if possible, converts to unicode otherwise."""
2381 try:
2382 return int(s)
2383 except ValueError:
2384 return s.decode('utf-8')
2385 options.variables = dict((k, try_make_int(v)) for k, v in options.variables)
2386
2387
2388 class OptionParserIsolate(tools.OptionParserWithLogging):
2389 """Adds automatic --isolate, --isolated, --out and --variable handling."""
2390 # Set it to False if it is not required, e.g. it can be passed on but do not
2391 # fail if not given.
2392 require_isolated = True
2393
2394 def __init__(self, **kwargs):
2395 tools.OptionParserWithLogging.__init__(
2396 self,
2397 verbose=int(os.environ.get('ISOLATE_DEBUG', 0)),
2398 **kwargs)
2399 group = optparse.OptionGroup(self, "Common options")
2400 group.add_option(
2401 '-i', '--isolate',
2402 metavar='FILE',
2403 help='.isolate file to load the dependency data from')
2404 add_variable_option(group)
2405 group.add_option(
2406 '-o', '--outdir', metavar='DIR',
2407 help='Directory used to recreate the tree or store the hash table. '
2408 'Defaults: run|remap: a /tmp subdirectory, others: '
2409 'defaults to the directory containing --isolated')
2410 group.add_option(
2411 '--ignore_broken_items', action='store_true',
2412 default=bool(os.environ.get('ISOLATE_IGNORE_BROKEN_ITEMS')),
2413 help='Indicates that invalid entries in the isolated file to be '
2414 'only be logged and not stop processing. Defaults to True if '
2415 'env var ISOLATE_IGNORE_BROKEN_ITEMS is set')
2416 self.add_option_group(group)
2417
2418 def parse_args(self, *args, **kwargs):
2419 """Makes sure the paths make sense.
2420
2421 On Windows, / and \ are often mixed together in a path.
2422 """
2423 options, args = tools.OptionParserWithLogging.parse_args(
2424 self, *args, **kwargs)
2425 if not self.allow_interspersed_args and args:
2426 self.error('Unsupported argument: %s' % args)
2427
2428 cwd = file_path.get_native_path_case(unicode(os.getcwd()))
2429 parse_isolated_option(self, options, cwd, self.require_isolated)
2430 parse_variable_option(options)
2431
2432 if options.isolate:
2433 # TODO(maruel): Work with non-ASCII.
2434 # The path must be in native path case for tracing purposes.
2435 options.isolate = unicode(options.isolate).replace('/', os.path.sep)
2436 options.isolate = os.path.normpath(os.path.join(cwd, options.isolate))
2437 options.isolate = file_path.get_native_path_case(options.isolate)
2438
2439 if options.outdir and not is_url(options.outdir):
2440 options.outdir = unicode(options.outdir).replace('/', os.path.sep)
2441 # outdir doesn't need native path case since tracing is never done from
2442 # there.
2443 options.outdir = os.path.normpath(os.path.join(cwd, options.outdir))
2444
2445 return options, args
2446
2447
2448 def main(argv):
2449 dispatcher = subcommand.CommandDispatcher(__name__)
2450 try:
2451 return dispatcher.execute(OptionParserIsolate(version=__version__), argv)
2452 except Exception as e:
2453 tools.report_error(e)
2454 return 1
2455
2456
2457 if __name__ == '__main__':
2458 fix_encoding.fix_encoding()
2459 tools.disable_buffering()
2460 colorama.init()
2461 sys.exit(main(sys.argv[1:]))
OLDNEW
« no previous file with comments | « swarm_client/googletest/trace_test_cases.py ('k') | swarm_client/isolate_merge.py » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698