1 # (Be in -*- mode: python; coding: utf-8 -*- mode.)
3 # ====================================================================
4 # Copyright (c) 2006-2009 CollabNet. All rights reserved.
6 # This software is licensed as described in the file COPYING, which
7 # you should have received as part of this distribution. The terms
8 # are also available at http://subversion.tigris.org/license-1.html.
9 # If newer versions of this license are posted there, you may use a
10 # newer version instead, at your option.
12 # This software consists of voluntary contributions made by many
13 # individuals. For exact contribution history, see the revision
14 # history and logs, available at http://cvs2svn.tigris.org/.
15 # ====================================================================
17 # #####################
18 # ## PLEASE READ ME! ##
19 # #####################
21 # This is a template for an options file that can be used to configure
22 # cvs2svn to convert to Mercurial rather than to Subversion. See
23 # www/cvs2git.html and www/cvs2svn.html for general information, and
24 # see the comments in this file for information about what options are
25 # available and how they can be set.
27 # "cvs2hg" is shorthand for "cvs2git in the mode where it is
28 # outputting to Mercurial instead of git". But the program that needs
29 # to be run is still called "cvs2git". Run it with the --options
30 # option, passing it this file as argument:
32 # cvs2git --options=cvs2hg-example.options
34 # Mercurial can (experimentally at this time) read git-fast-import
35 # format via its "hg fastimport" extension, with exceptions:
37 # 1. "hg fastimport" does not support blobs, so the contents of the
38 # revisions are output inline rather than in a separate blobs file.
39 # This increases the size of the output, because file contents that
40 # appear identically on multiple branches have to be output
43 # Many options do not have defaults, so it is easier to copy this file
44 # and modify what you need rather than creating a new options file
45 # from scratch. This file is in Python syntax, but you don't need to
46 # know Python to modify it. But if you *do* know Python, then you
47 # will be happy to know that you can use arbitary Python constructs to
48 # do fancy configuration tricks.
50 # But please be aware of the following:
52 # * In many places, leading whitespace is significant in Python (it is
53 # used instead of curly braces to group statements together).
54 # Therefore, if you don't know what you are doing, it is best to
55 # leave the whitespace as it is.
57 # * In normal strings, Python treats a backslash ("\") as an escape
58 # character. Therefore, if you want to specify a string that
59 # contains a backslash, you need either to escape the backslash with
60 # another backslash ("\\"), or use a "raw string", as in one if the
61 # following equivalent examples:
63 # cvs_executable = 'c:\\windows\\system32\\cvs.exe'
64 # cvs_executable = r'c:\windows\system32\cvs.exe'
66 # See http://docs.python.org/tutorial/introduction.html#strings for
69 # Two identifiers will have been defined before this file is executed,
70 # and can be used freely within this file:
72 # ctx -- a Ctx object (see cvs2svn_lib/context.py), which holds
73 # many configuration options
75 # run_options -- an instance of the GitRunOptions class (see
76 # cvs2svn_lib/git_run_options.py), which holds some variables
77 # governing how cvs2git is run
80 # Import some modules that are used in setting the options:
83 from cvs2svn_lib import config
84 from cvs2svn_lib import changeset_database
85 from cvs2svn_lib.common import CVSTextDecoder
86 from cvs2svn_lib.log import logger
87 from cvs2svn_lib.project import Project
88 from cvs2svn_lib.git_output_option import GitRevisionInlineWriter
89 from cvs2svn_lib.git_output_option import GitOutputOption
90 from cvs2svn_lib.dvcs_common import KeywordHandlingPropertySetter
91 from cvs2svn_lib.revision_manager import NullRevisionCollector
92 from cvs2svn_lib.rcs_revision_manager import RCSRevisionReader
93 from cvs2svn_lib.cvs_revision_manager import CVSRevisionReader
94 from cvs2svn_lib.checkout_internal import InternalRevisionCollector
95 from cvs2svn_lib.checkout_internal import InternalRevisionReader
96 from cvs2svn_lib.symbol_strategy import AllBranchRule
97 from cvs2svn_lib.symbol_strategy import AllTagRule
98 from cvs2svn_lib.symbol_strategy import BranchIfCommitsRule
99 from cvs2svn_lib.symbol_strategy import ExcludeRegexpStrategyRule
100 from cvs2svn_lib.symbol_strategy import ForceBranchRegexpStrategyRule
101 from cvs2svn_lib.symbol_strategy import ForceTagRegexpStrategyRule
102 from cvs2svn_lib.symbol_strategy import ExcludeTrivialImportBranchRule
103 from cvs2svn_lib.symbol_strategy import ExcludeVendorBranchRule
104 from cvs2svn_lib.symbol_strategy import HeuristicStrategyRule
105 from cvs2svn_lib.symbol_strategy import UnambiguousUsageRule
106 from cvs2svn_lib.symbol_strategy import HeuristicPreferredParentRule
107 from cvs2svn_lib.symbol_strategy import SymbolHintsFileRule
108 from cvs2svn_lib.symbol_transform import ReplaceSubstringsSymbolTransform
109 from cvs2svn_lib.symbol_transform import RegexpSymbolTransform
110 from cvs2svn_lib.symbol_transform import IgnoreSymbolTransform
111 from cvs2svn_lib.symbol_transform import NormalizePathsSymbolTransform
112 from cvs2svn_lib.property_setters import AutoPropsPropertySetter
113 from cvs2svn_lib.property_setters import CVSBinaryFileDefaultMimeTypeSetter
114 from cvs2svn_lib.property_setters import CVSBinaryFileEOLStyleSetter
115 from cvs2svn_lib.property_setters import DefaultEOLStyleSetter
116 from cvs2svn_lib.property_setters import EOLStyleFromMimeTypeSetter
117 from cvs2svn_lib.property_setters import ExecutablePropertySetter
118 from cvs2svn_lib.property_setters import KeywordsPropertySetter
119 from cvs2svn_lib.property_setters import MimeMapper
120 from cvs2svn_lib.property_setters import SVNBinaryFileKeywordsPropertySetter
122 # To choose the level of logging output, uncomment one of the
124 #logger.log_level = logger.WARN
125 #logger.log_level = logger.QUIET
126 logger.log_level = logger.NORMAL
127 #logger.log_level = logger.VERBOSE
128 #logger.log_level = logger.DEBUG
131 # The directory to use for temporary files:
132 ctx.tmpdir = r'cvs2svn-tmp'
134 # cvs2hg does not need to keep track of what revisions will be
135 # excluded, so leave this option unchanged:
136 ctx.revision_collector = NullRevisionCollector()
138 # cvs2hg's revision reader is set via the GitOutputOption constructor,
139 # so leave this option set to None.
140 ctx.revision_reader = None
142 # Change the following line to True if the conversion should only
143 # include the trunk of the repository (i.e., all branches and tags
144 # should be omitted from the conversion):
145 ctx.trunk_only = False
147 # How to convert CVS author names, log messages, and filenames to
148 # Unicode. The first argument to CVSTextDecoder is a list of encoders
149 # that are tried in order in 'strict' mode until one of them succeeds.
150 # If none of those succeeds, then fallback_encoder (if it is
151 # specified) is used in lossy 'replace' mode. Setting a fallback
152 # encoder ensures that the encoder always succeeds, but it can cause
154 ctx.cvs_author_decoder = CVSTextDecoder(
160 #fallback_encoding='ascii'
162 ctx.cvs_log_decoder = CVSTextDecoder(
168 #fallback_encoding='ascii'
170 # You might want to be especially strict when converting filenames to
171 # Unicode (e.g., maybe not specify a fallback_encoding).
172 ctx.cvs_filename_decoder = CVSTextDecoder(
178 #fallback_encoding='ascii'
181 # Template for the commit message to be used for initial project
183 ctx.initial_project_commit_message = (
184 'Standard project directories initialized by cvs2svn.'
187 # Template for the commit message to be used for post commits, in
188 # which modifications to a vendor branch are copied back to trunk.
189 # This message can use '%(revnum)d' to include the SVN revision number
190 # of the revision that included the change to the vendor branch
191 # (admittedly rather pointless in a cvs2hg conversion).
192 ctx.post_commit_message = (
193 'This commit was generated by cvs2svn to track changes on a CVS '
197 # Template for the commit message to be used for commits in which
198 # symbols are created. This message can use '%(symbol_type)s' to
199 # include the type of the symbol ('branch' or 'tag') or
200 # '%(symbol_name)s' to include the name of the symbol.
201 ctx.symbol_commit_message = (
202 "This commit was manufactured by cvs2svn to create %(symbol_type)s "
206 # Some CVS clients for MacOS store resource fork data into CVS along
207 # with the file contents itself by wrapping it all up in a container
208 # format called "AppleSingle". Subversion currently does not support
209 # MacOS resource forks. Nevertheless, sometimes the resource fork
210 # information is not necessary and can be discarded. Set the
211 # following option to True if you would like cvs2svn to identify files
212 # whose contents are encoded in AppleSingle format, and discard all
213 # but the data fork for such files before committing them to
214 # Subversion. (Please note that AppleSingle contents are identified
215 # by the AppleSingle magic number as the first four bytes of the file.
216 # This check is not failproof, so only set this option if you think
218 ctx.decode_apple_single = False
220 # This option can be set to the name of a filename to which are stored
221 # statistics and conversion decisions about the CVS symbols.
222 ctx.symbol_info_filename = None
223 #ctx.symbol_info_filename = 'symbol-info.txt'
225 # cvs2svn uses "symbol strategy rules" to help decide how to handle
226 # CVS symbols. The rules in a project's symbol_strategy_rules are
227 # applied in order, and each rule is allowed to modify the symbol.
228 # The result (after each of the rules has been applied) is used for
231 # 1. A CVS symbol might be used as a tag in one file and as a branch
232 # in another file. cvs2svn has to decide whether to convert such a
233 # symbol as a tag or as a branch. cvs2svn uses a series of
234 # heuristic rules to decide how to convert a symbol. The user can
235 # override the default rules for specific symbols or symbols
236 # matching regular expressions.
238 # 2. cvs2svn is also capable of excluding symbols from the conversion
239 # (provided no other symbols depend on them.
241 # 3. CVS does not record unambiguously the line of development from
242 # which a symbol sprouted. cvs2svn uses a heuristic to choose a
243 # symbol's "preferred parents".
245 # The standard branch/tag/exclude StrategyRules do not change a symbol
246 # that has already been processed by an earlier rule, so in effect the
247 # first matching rule is the one that is used.
249 global_symbol_strategy_rules = [
250 # It is possible to specify manually exactly how symbols should be
251 # converted and what line of development should be used as the
252 # preferred parent. To do so, create a file containing the symbol
253 # hints and enable the following option.
255 # The format of the hints file is described in the documentation
256 # for the --symbol-hints command-line option. The file output by
257 # the --write-symbol-info (i.e., ctx.symbol_info_filename) option
258 # is in the same format. The simplest way to use this option is
259 # to run the conversion through CollateSymbolsPass with
260 # --write-symbol-info option, copy the symbol info and edit it to
261 # create a hints file, then re-start the conversion at
262 # CollateSymbolsPass with this option enabled.
263 #SymbolHintsFileRule('symbol-hints.txt'),
265 # To force all symbols matching a regular expression to be
266 # converted as branches, add rules like the following:
267 #ForceBranchRegexpStrategyRule(r'branch.*'),
269 # To force all symbols matching a regular expression to be
270 # converted as tags, add rules like the following:
271 #ForceTagRegexpStrategyRule(r'tag.*'),
273 # To force all symbols matching a regular expression to be
274 # excluded from the conversion, add rules like the following:
275 #ExcludeRegexpStrategyRule(r'unknown-.*'),
277 # Sometimes people use "cvs import" to get their own source code
278 # into CVS. This practice creates a vendor branch 1.1.1 and
279 # imports the code onto the vendor branch as 1.1.1.1, then copies
280 # the same content to the trunk as version 1.1. Normally, such
281 # vendor branches are useless and they complicate the SVN history
282 # unnecessarily. The following rule excludes any branches that
283 # only existed as a vendor branch with a single import (leaving
284 # only the 1.1 revision). If you want to retain such branches,
285 # comment out the following line. (Please note that this rule
286 # does not exclude vendor *tags*, as they are not so easy to
288 ExcludeTrivialImportBranchRule(),
290 # To exclude all vendor branches (branches that had "cvs import"s
291 # on them but no other kinds of commits), uncomment the following
293 #ExcludeVendorBranchRule(),
295 # Usually you want this rule, to convert unambiguous symbols
296 # (symbols that were only ever used as tags or only ever used as
297 # branches in CVS) the same way they were used in CVS:
298 UnambiguousUsageRule(),
300 # If there was ever a commit on a symbol, then it cannot be
301 # converted as a tag. This rule causes all such symbols to be
302 # converted as branches. If you would like to resolve such
303 # ambiguities manually, comment out the following line:
304 BranchIfCommitsRule(),
306 # Last in the list can be a catch-all rule that is used for
307 # symbols that were not matched by any of the more specific rules
308 # above. (Assuming that BranchIfCommitsRule() was included above,
309 # then the symbols that are still indeterminate at this point can
310 # sensibly be converted as branches or tags.) Include at most one
311 # of these lines. If none of these catch-all rules are included,
312 # then the presence of any ambiguous symbols (that haven't been
313 # disambiguated above) is an error:
315 # Convert ambiguous symbols based on whether they were used more
316 # often as branches or as tags:
317 HeuristicStrategyRule(),
318 # Convert all ambiguous symbols as branches:
320 # Convert all ambiguous symbols as tags:
323 # The last rule is here to choose the preferred parent of branches
324 # and tags, that is, the line of development from which the symbol
326 HeuristicPreferredParentRule(),
329 # Specify a username to be used for commits for which CVS doesn't
330 # record the original author (for example, the creation of a branch).
331 # This should be a simple (unix-style) username, but it can be
332 # translated into a hg-style name by the author_transforms map.
333 ctx.username = 'cvs2svn'
335 # ctx.file_property_setters and ctx.revision_property_setters contain
336 # rules used to set the svn properties on files in the converted
337 # archive. For each file, the rules are tried one by one. Any rule
338 # can add or suppress one or more svn properties. Typically the rules
339 # will not overwrite properties set by a previous rule (though they
340 # are free to do so). ctx.file_property_setters should be used for
341 # properties that remain the same for the life of the file; these
342 # should implement FilePropertySetter. ctx.revision_property_setters
343 # should be used for properties that are allowed to vary from revision
344 # to revision; these should implement RevisionPropertySetter.
346 # Obviously, SVN properties per se are not interesting for a cvs2hg
347 # conversion, but some of these properties have side-effects that do
348 # affect the hg output. FIXME: Document this in more detail.
349 ctx.file_property_setters.extend([
350 # To read auto-props rules from a file, uncomment the following line
351 # and specify a filename. The boolean argument specifies whether
352 # case should be ignored when matching filenames to the filename
353 # patterns found in the auto-props file:
354 #AutoPropsPropertySetter(
355 # r'/home/username/.subversion/config',
359 # To read mime types from a file and use them to set svn:mime-type
360 # based on the filename extensions, uncomment the following line
361 # and specify a filename (see
362 # http://en.wikipedia.org/wiki/Mime.types for information about
364 #MimeMapper(r'/etc/mime.types', ignore_case=False),
366 # Omit the svn:eol-style property from any files that are listed
367 # as binary (i.e., mode '-kb') in CVS:
368 CVSBinaryFileEOLStyleSetter(),
370 # If the file is binary and its svn:mime-type property is not yet
371 # set, set svn:mime-type to 'application/octet-stream'.
372 CVSBinaryFileDefaultMimeTypeSetter(),
374 # To try to determine the eol-style from the mime type, uncomment
375 # the following line:
376 #EOLStyleFromMimeTypeSetter(),
378 # Choose one of the following lines to set the default
379 # svn:eol-style if none of the above rules applied. The argument
380 # is the svn:eol-style that should be applied, or None if no
381 # svn:eol-style should be set (i.e., the file should be treated as
384 # The default is to treat all files as binary unless one of the
385 # previous rules has determined otherwise, because this is the
386 # safest approach. However, if you have been diligent about
387 # marking binary files with -kb in CVS and/or you have used the
388 # above rules to definitely mark binary files as binary, then you
389 # might prefer to use 'native' as the default, as it is usually
390 # the most convenient setting for text files. Other possible
391 # options: 'CRLF', 'CR', 'LF'.
392 DefaultEOLStyleSetter(None),
393 #DefaultEOLStyleSetter('native'),
395 # Prevent svn:keywords from being set on files that have
396 # svn:eol-style unset.
397 SVNBinaryFileKeywordsPropertySetter(),
399 # If svn:keywords has not been set yet, set it based on the file's
401 KeywordsPropertySetter(config.SVN_KEYWORDS_VALUE),
403 # Set the svn:executable flag on any files that are marked in CVS as
405 ExecutablePropertySetter(),
407 # The following causes keywords to be collapsed in all text to be
409 KeywordHandlingPropertySetter('collapsed'),
412 ctx.revision_property_setters.extend([
415 # To skip the cleanup of temporary files, uncomment the following
417 #ctx.skip_cleanup = True
420 # In CVS, it is perfectly possible to make a single commit that
421 # affects more than one project or more than one branch of a single
422 # project. Subversion also allows such commits. Therefore, by
423 # default, when cvs2svn sees what looks like a cross-project or
424 # cross-branch CVS commit, it converts it into a
425 # cross-project/cross-branch Subversion commit.
427 # However, other tools and SCMs have trouble representing
428 # cross-project or cross-branch commits. (For example, Trac's Revtree
429 # plugin, http://www.trac-hacks.org/wiki/RevtreePlugin is confused by
430 # such commits.) Therefore, we provide the following two options to
431 # allow cross-project/cross-branch commits to be suppressed.
433 # cvs2hg only supports single-project conversions (multiple-project
434 # conversions wouldn't really make sense for hg anyway). So this
435 # option must be set to False:
436 ctx.cross_project_commits = False
438 # Mercurial itself doesn't allow commits that affect more than one
439 # branch, so this option must be set to False:
440 ctx.cross_branch_commits = False
442 # cvs2hg does not yet handle translating .cvsignore files into
443 # .hgignore content, so by default, the .cvsignore files are included
444 # inthe conversion output. If you would like to omit the .cvsignore
445 # files from the output, set this option to False:
446 ctx.keep_cvsignore = True
448 # By default, it is a fatal error for a CVS ",v" file to appear both
449 # inside and outside of an "Attic" subdirectory (this should never
450 # happen, but frequently occurs due to botched repository
451 # administration). If you would like to retain both versions of such
452 # files, change the following option to True, and the attic version of
453 # the file will be written to a subdirectory called "Attic" in the
455 ctx.retain_conflicting_attic_files = False
457 # CVS uses unix login names as author names whereas "hg fastimport"
458 # format requires author names to be of the form "foo <bar>". The
459 # default is to set the author to "cvsauthor <cvsauthor>".
460 # author_transforms can be used to map cvsauthor names (e.g.,
461 # "jrandom") to a true name and email address (e.g., "J. Random
462 # <jrandom@example.com>" for the example shown). All strings should
463 # be either Unicode strings (i.e., with "u" as a prefix) or 8-bit
464 # strings in the utf-8 encoding. The values can either be strings in
465 # the form "name <email>" or tuples (name, email). Please substitute
466 # your own project's usernames here to use with the author_transforms
467 # option of GitOutputOption below.
469 'jrandom' : ('J. Random', 'jrandom@example.com'),
470 'mhagger' : 'Michael Haggerty <mhagger@alum.mit.edu>',
471 'brane' : (u'Branko Čibej', 'brane@xbc.nu'),
472 'ringstrom' : 'Tobias Ringström <tobias@ringstrom.mine.nu>',
473 'dionisos' : (u'Erik Hülsmann', 'e.huelsmann@gmx.net'),
475 # This one will be used for commits for which CVS doesn't record
476 # the original author, as explained above.
477 'cvs2svn' : 'cvs2svn <admin@example.com>',
480 # This is the main option that causes cvs2svn to output to an "hg
481 # fastimport"-format dumpfile rather than to Subversion:
482 ctx.output_option = GitOutputOption(
483 # The file in which to write the "hg fastimport" stream that
484 # contains the changesets and branch/tag information:
485 os.path.join(ctx.tmpdir, 'hg-dump.dat'),
487 # Write the file contents inline in the "hg fastimport" stream,
488 # rather than using a separate blobs file (which "hg fastimport"
490 revision_writer=GitRevisionInlineWriter(
491 # cvs2hg uses either RCS's "co" command or CVS's "cvs co -p" to
492 # extract the content of file revisions. Here you can choose
493 # whether to use RCS (faster, but fails in some rare
494 # circumstances) or CVS (much slower, but more reliable).
495 #RCSRevisionReader(co_executable=r'co')
496 CVSRevisionReader(cvs_executable=r'cvs')
499 # Optional map from CVS author names to hg author names:
500 author_transforms=author_transforms,
503 # Change this option to True to turn on profiling of cvs2svn (for
504 # debugging purposes):
505 run_options.profiling = False
508 # Should CVSItem -> Changeset database files be memory mapped? In
509 # some tests, using memory mapping speeded up the overall conversion
510 # by about 5%. But this option can cause the conversion to fail with
511 # an out of memory error if the conversion computer runs out of
512 # virtual address space (e.g., when running a very large conversion on
513 # a 32-bit operating system). Therefore it is disabled by default.
514 # Uncomment the following line to allow these database files to be
516 #changeset_database.use_mmap_for_cvs_item_to_changeset_table = True
518 # Now set the project to be converted to hg. cvs2hg only supports
519 # single-project conversions, so this method must only be called once:
520 run_options.set_project(
521 # The filesystem path to the part of the CVS repository (*not* a
522 # CVS working copy) that should be converted. This may be a
523 # subdirectory (i.e., a module) within a larger CVS repository.
524 r'test-data/main-cvsrepos',
526 # A list of symbol transformations that can be used to rename
527 # symbols in this project.
529 # Use IgnoreSymbolTransforms like the following to completely
530 # ignore symbols matching a regular expression when parsing
531 # the CVS repository, for example to avoid warnings about
532 # branches with two names and to choose the preferred name.
533 # It is *not* recommended to use this instead of
534 # ExcludeRegexpStrategyRule; though more efficient,
535 # IgnoreSymbolTransforms are less flexible and don't exclude
536 # branches correctly. The argument is a Python-style regular
537 # expression that has to match the *whole* CVS symbol name:
538 #IgnoreSymbolTransform(r'nightly-build-tag-.*')
540 # RegexpSymbolTransforms transform symbols textually using a
541 # regular expression. The first argument is a Python regular
542 # expression pattern and the second is a replacement pattern.
543 # The pattern is matched against each symbol name. If it
544 # matches the whole symbol name, then the symbol name is
545 # replaced with the corresponding replacement text. The
546 # replacement can include substitution patterns (e.g., r'\1'
547 # or r'\g<name>'). Typically you will want to use raw strings
548 # (strings with a preceding 'r', like shown in the examples)
549 # for the regexp and its replacement to avoid backslash
550 # substitution within those strings.
551 #RegexpSymbolTransform(r'release-(\d+)_(\d+)',
553 #RegexpSymbolTransform(r'release-(\d+)_(\d+)_(\d+)',
554 # r'release-\1.\2.\3'),
556 # Simple 1:1 character replacements can also be done. The
557 # following transform, which converts backslashes into forward
558 # slashes, should usually be included:
559 ReplaceSubstringsSymbolTransform('\\','/'),
561 # This last rule eliminates leading, trailing, and repeated
562 # slashes within the output symbol names:
563 NormalizePathsSymbolTransform(),
566 # See the definition of global_symbol_strategy_rules above for a
567 # description of this option:
568 symbol_strategy_rules=global_symbol_strategy_rules,