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 git 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 # The program that is run to convert from CVS to git is called
28 # cvs2git. Run it with the --options option, passing it this file
31 # cvs2git --options=cvs2git-example.options
33 # The output of cvs2git is a blob file and a dump file that can be
34 # loaded into git using the "git fast-import" command. Please read
35 # www/cvs2git.html for more information.
37 # Many options do not have defaults, so it is easier to copy this file
38 # and modify what you need rather than creating a new options file
39 # from scratch. This file is in Python syntax, but you don't need to
40 # know Python to modify it. But if you *do* know Python, then you
41 # will be happy to know that you can use arbitary Python constructs to
42 # do fancy configuration tricks.
44 # But please be aware of the following:
46 # * In many places, leading whitespace is significant in Python (it is
47 # used instead of curly braces to group statements together).
48 # Therefore, if you don't know what you are doing, it is best to
49 # leave the whitespace as it is.
51 # * In normal strings, Python treats a backslash ("\") as an escape
52 # character. Therefore, if you want to specify a string that
53 # contains a backslash, you need either to escape the backslash with
54 # another backslash ("\\"), or use a "raw string", as in one if the
55 # following equivalent examples:
57 # cvs_executable = 'c:\\windows\\system32\\cvs.exe'
58 # cvs_executable = r'c:\windows\system32\cvs.exe'
60 # See http://docs.python.org/tutorial/introduction.html#strings for
63 # Two identifiers will have been defined before this file is executed,
64 # and can be used freely within this file:
66 # ctx -- a Ctx object (see cvs2svn_lib/context.py), which holds
67 # many configuration options
69 # run_options -- an instance of the GitRunOptions class (see
70 # cvs2svn_lib/git_run_options.py), which holds some variables
71 # governing how cvs2git is run
74 # Import some modules that are used in setting the options:
77 from cvs2svn_lib import config
78 from cvs2svn_lib import changeset_database
79 from cvs2svn_lib.common import CVSTextDecoder
80 from cvs2svn_lib.log import Log
81 from cvs2svn_lib.project import Project
82 from cvs2svn_lib.git_revision_recorder import GitRevisionRecorder
83 from cvs2svn_lib.git_output_option import GitRevisionMarkWriter
84 from cvs2svn_lib.git_output_option import GitOutputOption
85 from cvs2svn_lib.revision_manager import NullRevisionRecorder
86 from cvs2svn_lib.revision_manager import NullRevisionExcluder
87 from cvs2svn_lib.fulltext_revision_recorder \
88 import SimpleFulltextRevisionRecorderAdapter
89 from cvs2svn_lib.rcs_revision_manager import RCSRevisionReader
90 from cvs2svn_lib.cvs_revision_manager import CVSRevisionReader
91 from cvs2svn_lib.checkout_internal import InternalRevisionRecorder
92 from cvs2svn_lib.checkout_internal import InternalRevisionExcluder
93 from cvs2svn_lib.checkout_internal import InternalRevisionReader
94 from cvs2svn_lib.symbol_strategy import AllBranchRule
95 from cvs2svn_lib.symbol_strategy import AllTagRule
96 from cvs2svn_lib.symbol_strategy import BranchIfCommitsRule
97 from cvs2svn_lib.symbol_strategy import ExcludeRegexpStrategyRule
98 from cvs2svn_lib.symbol_strategy import ForceBranchRegexpStrategyRule
99 from cvs2svn_lib.symbol_strategy import ForceTagRegexpStrategyRule
100 from cvs2svn_lib.symbol_strategy import ExcludeTrivialImportBranchRule
101 from cvs2svn_lib.symbol_strategy import ExcludeVendorBranchRule
102 from cvs2svn_lib.symbol_strategy import HeuristicStrategyRule
103 from cvs2svn_lib.symbol_strategy import UnambiguousUsageRule
104 from cvs2svn_lib.symbol_strategy import HeuristicPreferredParentRule
105 from cvs2svn_lib.symbol_strategy import SymbolHintsFileRule
106 from cvs2svn_lib.symbol_transform import ReplaceSubstringsSymbolTransform
107 from cvs2svn_lib.symbol_transform import RegexpSymbolTransform
108 from cvs2svn_lib.symbol_transform import IgnoreSymbolTransform
109 from cvs2svn_lib.symbol_transform import NormalizePathsSymbolTransform
110 from cvs2svn_lib.property_setters import AutoPropsPropertySetter
111 from cvs2svn_lib.property_setters import CVSBinaryFileDefaultMimeTypeSetter
112 from cvs2svn_lib.property_setters import CVSBinaryFileEOLStyleSetter
113 from cvs2svn_lib.property_setters import DefaultEOLStyleSetter
114 from cvs2svn_lib.property_setters import EOLStyleFromMimeTypeSetter
115 from cvs2svn_lib.property_setters import ExecutablePropertySetter
116 from cvs2svn_lib.property_setters import KeywordsPropertySetter
117 from cvs2svn_lib.property_setters import MimeMapper
118 from cvs2svn_lib.property_setters import SVNBinaryFileKeywordsPropertySetter
120 # To choose the level of logging output, uncomment one of the
122 #Log().log_level = Log.WARN
123 #Log().log_level = Log.QUIET
124 Log().log_level = Log.NORMAL
125 #Log().log_level = Log.VERBOSE
126 #Log().log_level = Log.DEBUG
129 # During CollectRevsPass, cvs2git records the contents of file
130 # revisions into a "blob" file in git-fast-import format. This option
131 # configures that process:
132 ctx.revision_recorder = SimpleFulltextRevisionRecorderAdapter(
133 # The following option specifies how the revision contents of the RCS
134 # files should be read.
136 # RCSRevisionReader uses RCS's "co" program to extract the revision
137 # contents of the RCS files during CollectRevsPass. The constructor
138 # argument specifies how to invoke the "co" executable.
140 # CVSRevisionReader uses the "cvs" program to extract the revision
141 # contents out of the RCS files during OutputPass. This option is
142 # considerably slower than RCSRevisionReader because "cvs" is
143 # considerably slower than "co". However, it works in some situations
144 # where RCSRevisionReader fails; see the HTML documentation of the
145 # "--use-cvs" option for details. The constructor argument specifies
146 # how to invoke the "co" executable.
148 # Uncomment one of the two following lines:
149 #RCSRevisionReader(co_executable=r'co'),
150 CVSRevisionReader(cvs_executable=r'cvs'),
152 # The file in which to write the git-fast-import stream that
153 # contains the file revision contents:
154 GitRevisionRecorder('cvs2svn-tmp/git-blob.dat'),
157 # cvs2git does not need to keep track of what revisions will be
158 # excluded, so leave this option unchanged:
159 ctx.revision_excluder = NullRevisionExcluder()
161 # cvs2git doesn't need a revision reader because OutputPass only
162 # refers to blobs that were output during CollectRevsPass, so leave
163 # this option set to None.
164 ctx.revision_reader = None
166 # Change the following line to True if the conversion should only
167 # include the trunk of the repository (i.e., all branches and tags
168 # should be omitted from the conversion):
169 ctx.trunk_only = False
171 # How to convert CVS author names, log messages, and filenames to
172 # Unicode. The first argument to CVSTextDecoder is a list of encoders
173 # that are tried in order in 'strict' mode until one of them succeeds.
174 # If none of those succeeds, then fallback_encoder (if it is
175 # specified) is used in lossy 'replace' mode. Setting a fallback
176 # encoder ensures that the encoder always succeeds, but it can cause
178 ctx.cvs_author_decoder = CVSTextDecoder(
184 #fallback_encoding='ascii'
186 ctx.cvs_log_decoder = CVSTextDecoder(
192 #fallback_encoding='ascii'
194 # You might want to be especially strict when converting filenames to
195 # Unicode (e.g., maybe not specify a fallback_encoding).
196 ctx.cvs_filename_decoder = CVSTextDecoder(
202 #fallback_encoding='ascii'
205 # Template for the commit message to be used for initial project
207 ctx.initial_project_commit_message = (
208 'Standard project directories initialized by cvs2svn.'
211 # Template for the commit message to be used for post commits, in
212 # which modifications to a vendor branch are copied back to trunk.
213 # This message can use '%(revnum)d' to include the SVN revision number
214 # of the revision that included the change to the vendor branch
215 # (admittedly rather pointless in a cvs2git conversion).
216 ctx.post_commit_message = (
217 'This commit was generated by cvs2svn to track changes on a CVS '
221 # Template for the commit message to be used for commits in which
222 # symbols are created. This message can use '%(symbol_type)s' to
223 # include the type of the symbol ('branch' or 'tag') or
224 # '%(symbol_name)s' to include the name of the symbol.
225 ctx.symbol_commit_message = (
226 "This commit was manufactured by cvs2svn to create %(symbol_type)s "
230 # Template for the commit message to be used for commits in which
231 # tags are pseudo-merged back to their source branch. This message can
232 # use '%(symbol_name)s' to include the name of the symbol.
233 # (Not used by default unless you enable tie_tag_fixup_branches on
235 ctx.tie_tag_ancestry_message = (
236 "This commit was manufactured by cvs2svn to tie ancestry for "
237 "tag '%(symbol_name)s' back to the source branch."
240 # Some CVS clients for MacOS store resource fork data into CVS along
241 # with the file contents itself by wrapping it all up in a container
242 # format called "AppleSingle". Subversion currently does not support
243 # MacOS resource forks. Nevertheless, sometimes the resource fork
244 # information is not necessary and can be discarded. Set the
245 # following option to True if you would like cvs2svn to identify files
246 # whose contents are encoded in AppleSingle format, and discard all
247 # but the data fork for such files before committing them to
248 # Subversion. (Please note that AppleSingle contents are identified
249 # by the AppleSingle magic number as the first four bytes of the file.
250 # This check is not failproof, so only set this option if you think
252 ctx.decode_apple_single = False
254 # This option can be set to the name of a filename to which are stored
255 # statistics and conversion decisions about the CVS symbols.
256 ctx.symbol_info_filename = None
257 #ctx.symbol_info_filename = 'symbol-info.txt'
259 # cvs2svn uses "symbol strategy rules" to help decide how to handle
260 # CVS symbols. The rules in a project's symbol_strategy_rules are
261 # applied in order, and each rule is allowed to modify the symbol.
262 # The result (after each of the rules has been applied) is used for
265 # 1. A CVS symbol might be used as a tag in one file and as a branch
266 # in another file. cvs2svn has to decide whether to convert such a
267 # symbol as a tag or as a branch. cvs2svn uses a series of
268 # heuristic rules to decide how to convert a symbol. The user can
269 # override the default rules for specific symbols or symbols
270 # matching regular expressions.
272 # 2. cvs2svn is also capable of excluding symbols from the conversion
273 # (provided no other symbols depend on them.
275 # 3. CVS does not record unambiguously the line of development from
276 # which a symbol sprouted. cvs2svn uses a heuristic to choose a
277 # symbol's "preferred parents".
279 # The standard branch/tag/exclude StrategyRules do not change a symbol
280 # that has already been processed by an earlier rule, so in effect the
281 # first matching rule is the one that is used.
283 global_symbol_strategy_rules = [
284 # It is possible to specify manually exactly how symbols should be
285 # converted and what line of development should be used as the
286 # preferred parent. To do so, create a file containing the symbol
287 # hints and enable the following option.
289 # The format of the hints file is described in the documentation
290 # for the --symbol-hints command-line option. The file output by
291 # the --write-symbol-info (i.e., ctx.symbol_info_filename) option
292 # is in the same format. The simplest way to use this option is
293 # to run the conversion through CollateSymbolsPass with
294 # --write-symbol-info option, copy the symbol info and edit it to
295 # create a hints file, then re-start the conversion at
296 # CollateSymbolsPass with this option enabled.
297 #SymbolHintsFileRule('symbol-hints.txt'),
299 # To force all symbols matching a regular expression to be
300 # converted as branches, add rules like the following:
301 #ForceBranchRegexpStrategyRule(r'branch.*'),
303 # To force all symbols matching a regular expression to be
304 # converted as tags, add rules like the following:
305 #ForceTagRegexpStrategyRule(r'tag.*'),
307 # To force all symbols matching a regular expression to be
308 # excluded from the conversion, add rules like the following:
309 #ExcludeRegexpStrategyRule(r'unknown-.*'),
311 # Sometimes people use "cvs import" to get their own source code
312 # into CVS. This practice creates a vendor branch 1.1.1 and
313 # imports the code onto the vendor branch as 1.1.1.1, then copies
314 # the same content to the trunk as version 1.1. Normally, such
315 # vendor branches are useless and they complicate the SVN history
316 # unnecessarily. The following rule excludes any branches that
317 # only existed as a vendor branch with a single import (leaving
318 # only the 1.1 revision). If you want to retain such branches,
319 # comment out the following line. (Please note that this rule
320 # does not exclude vendor *tags*, as they are not so easy to
322 ExcludeTrivialImportBranchRule(),
324 # To exclude all vendor branches (branches that had "cvs import"s
325 # on them bug no other kinds of commits), uncomment the following
327 #ExcludeVendorBranchRule(),
329 # Usually you want this rule, to convert unambiguous symbols
330 # (symbols that were only ever used as tags or only ever used as
331 # branches in CVS) the same way they were used in CVS:
332 UnambiguousUsageRule(),
334 # If there was ever a commit on a symbol, then it cannot be
335 # converted as a tag. This rule causes all such symbols to be
336 # converted as branches. If you would like to resolve such
337 # ambiguities manually, comment out the following line:
338 BranchIfCommitsRule(),
340 # Last in the list can be a catch-all rule that is used for
341 # symbols that were not matched by any of the more specific rules
342 # above. (Assuming that BranchIfCommitsRule() was included above,
343 # then the symbols that are still indeterminate at this point can
344 # sensibly be converted as branches or tags.) Include at most one
345 # of these lines. If none of these catch-all rules are included,
346 # then the presence of any ambiguous symbols (that haven't been
347 # disambiguated above) is an error:
349 # Convert ambiguous symbols based on whether they were used more
350 # often as branches or as tags:
351 HeuristicStrategyRule(),
352 # Convert all ambiguous symbols as branches:
354 # Convert all ambiguous symbols as tags:
357 # The last rule is here to choose the preferred parent of branches
358 # and tags, that is, the line of development from which the symbol
360 HeuristicPreferredParentRule(),
363 # Specify a username to be used for commits for which CVS doesn't
364 # record the original author (for example, the creation of a branch).
365 # This should be a simple (unix-style) username, but it can be
366 # translated into a git-style name by the author_transforms map.
367 ctx.username = 'cvs2svn'
369 # ctx.svn_property_setters contains a list of rules used to set the
370 # svn properties on files in the converted archive. For each file,
371 # the rules are tried one by one. Any rule can add or suppress one or
372 # more svn properties. Typically the rules will not overwrite
373 # properties set by a previous rule (though they are free to do so).
375 # Obviously, SVN properties per se are not interesting for a cvs2git
376 # conversion, but some of these properties have side-effects that do
377 # affect the git output. FIXME: Document this in more detail.
378 ctx.svn_property_setters.extend([
379 # To read auto-props rules from a file, uncomment the following line
380 # and specify a filename. The boolean argument specifies whether
381 # case should be ignored when matching filenames to the filename
382 # patterns found in the auto-props file:
383 #AutoPropsPropertySetter(
384 # r'/home/username/.subversion/config',
388 # To read mime types from a file, uncomment the following line and
389 # specify a filename:
390 #MimeMapper(r'/etc/mime.types'),
392 # Omit the svn:eol-style property from any files that are listed
393 # as binary (i.e., mode '-kb') in CVS:
394 CVSBinaryFileEOLStyleSetter(),
396 # If the file is binary and its svn:mime-type property is not yet
397 # set, set svn:mime-type to 'application/octet-stream'.
398 CVSBinaryFileDefaultMimeTypeSetter(),
400 # To try to determine the eol-style from the mime type, uncomment
401 # the following line:
402 #EOLStyleFromMimeTypeSetter(),
404 # Choose one of the following lines to set the default
405 # svn:eol-style if none of the above rules applied. The argument
406 # is the svn:eol-style that should be applied, or None if no
407 # svn:eol-style should be set (i.e., the file should be treated as
410 # The default is to treat all files as binary unless one of the
411 # previous rules has determined otherwise, because this is the
412 # safest approach. However, if you have been diligent about
413 # marking binary files with -kb in CVS and/or you have used the
414 # above rules to definitely mark binary files as binary, then you
415 # might prefer to use 'native' as the default, as it is usually
416 # the most convenient setting for text files. Other possible
417 # options: 'CRLF', 'CR', 'LF'.
418 DefaultEOLStyleSetter(None),
419 #DefaultEOLStyleSetter('native'),
421 # Prevent svn:keywords from being set on files that have
422 # svn:eol-style unset.
423 SVNBinaryFileKeywordsPropertySetter(),
425 # If svn:keywords has not been set yet, set it based on the file's
427 KeywordsPropertySetter(config.SVN_KEYWORDS_VALUE),
429 # Set the svn:executable flag on any files that are marked in CVS as
431 ExecutablePropertySetter(),
435 # The directory to use for temporary files:
436 ctx.tmpdir = r'cvs2svn-tmp'
438 # To skip the cleanup of temporary files, uncomment the following
440 #ctx.skip_cleanup = True
443 # In CVS, it is perfectly possible to make a single commit that
444 # affects more than one project or more than one branch of a single
445 # project. Subversion also allows such commits. Therefore, by
446 # default, when cvs2svn sees what looks like a cross-project or
447 # cross-branch CVS commit, it converts it into a
448 # cross-project/cross-branch Subversion commit.
450 # However, other tools and SCMs have trouble representing
451 # cross-project or cross-branch commits. (For example, Trac's Revtree
452 # plugin, http://www.trac-hacks.org/wiki/RevtreePlugin is confused by
453 # such commits.) Therefore, we provide the following two options to
454 # allow cross-project/cross-branch commits to be suppressed.
456 # cvs2git only supports single-project conversions (multiple-project
457 # conversions wouldn't really make sense for git anyway). So this
458 # option must be set to False:
459 ctx.cross_project_commits = False
461 # git itself doesn't allow commits that affect more than one branch,
462 # so this option must be set to False:
463 ctx.cross_branch_commits = False
465 # cvs2git does not yet handle translating .cvsignore files into
466 # .gitignore files, so by default, the .cvsignore files are included
467 # in the conversion output. If you would like to omit the .cvsignore
468 # files from the output, set this option to False:
469 ctx.keep_cvsignore = True
471 # By default, it is a fatal error for a CVS ",v" file to appear both
472 # inside and outside of an "Attic" subdirectory (this should never
473 # happen, but frequently occurs due to botched repository
474 # administration). If you would like to retain both versions of such
475 # files, change the following option to True, and the attic version of
476 # the file will be written to a subdirectory called "Attic" in the
478 ctx.retain_conflicting_attic_files = False
480 # CVS uses unix login names as author names whereas git requires
481 # author names to be of the form "foo <bar>". The default is to set
482 # the git author to "cvsauthor <cvsauthor>". author_transforms can be
483 # used to map cvsauthor names (e.g., "jrandom") to a true name and
484 # email address (e.g., "J. Random <jrandom@example.com>" for the
485 # example shown). All values should be either Unicode strings (i.e.,
486 # with "u" as a prefix) or 8-bit strings in the utf-8 encoding.
487 # Please substitute your own project's usernames here to use with the
488 # author_transforms option of GitOutputOption below.
490 'jrandom' : ('J. Random', 'jrandom@example.com'),
491 'mhagger' : ('Michael Haggerty', 'mhagger@alum.mit.edu'),
492 'brane' : (u'Branko Čibej', 'brane@xbc.nu'),
493 'ringstrom' : ('Tobias Ringström', 'tobias@ringstrom.mine.nu'),
494 'dionisos' : (u'Erik Hülsmann', 'e.huelsmann@gmx.net'),
496 # This one will be used for commits for which CVS doesn't record
497 # the original author, as explained above.
498 'cvs2svn' : ('cvs2svn', 'admin@example.com'),
501 # This is the main option that causes cvs2svn to output to a
502 # "fastimport"-format dumpfile rather than to Subversion:
503 ctx.output_option = GitOutputOption(
504 # The file in which to write the git-fast-import stream that
505 # contains the changesets and branch/tag information:
506 'cvs2svn-tmp/git-dump.dat',
508 # The blobs will be written via the revision recorder, so in
509 # OutputPass we only have to emit references to the blob marks:
510 GitRevisionMarkWriter(),
512 # Optional map from CVS author names to git author names:
513 author_transforms=author_transforms,
516 # Change this option to True to turn on profiling of cvs2svn (for
517 # debugging purposes):
518 run_options.profiling = False
521 # Should CVSItem -> Changeset database files be memory mapped? In
522 # some tests, using memory mapping speeded up the overall conversion
523 # by about 5%. But this option can cause the conversion to fail with
524 # an out of memory error if the conversion computer runs out of
525 # virtual address space (e.g., when running a very large conversion on
526 # a 32-bit operating system). Therefore it is disabled by default.
527 # Uncomment the following line to allow these database files to be
529 #changeset_database.use_mmap_for_cvs_item_to_changeset_table = True
531 # Now set the project to be converted to git. cvs2git only supports
532 # single-project conversions, so this method must only be called
534 run_options.set_project(
535 # The filesystem path to the part of the CVS repository (*not* a
536 # CVS working copy) that should be converted. This may be a
537 # subdirectory (i.e., a module) within a larger CVS repository.
538 r'test-data/main-cvsrepos',
540 # A list of symbol transformations that can be used to rename
541 # symbols in this project.
543 # Use IgnoreSymbolTransforms like the following to completely
544 # ignore symbols matching a regular expression when parsing
545 # the CVS repository, for example to avoid warnings about
546 # branches with two names and to choose the preferred name.
547 # It is *not* recommended to use this instead of
548 # ExcludeRegexpStrategyRule; though more efficient,
549 # IgnoreSymbolTransforms are less flexible and don't exclude
550 # branches correctly. The argument is a Python-style regular
551 # expression that has to match the *whole* CVS symbol name:
552 #IgnoreSymbolTransform(r'nightly-build-tag-.*')
554 # RegexpSymbolTransforms transform symbols textually using a
555 # regular expression. The first argument is a Python regular
556 # expression pattern and the second is a replacement pattern.
557 # The pattern is matched against each symbol name. If it
558 # matches the whole symbol name, then the symbol name is
559 # replaced with the corresponding replacement text. The
560 # replacement can include substitution patterns (e.g., r'\1'
561 # or r'\g<name>'). Typically you will want to use raw strings
562 # (strings with a preceding 'r', like shown in the examples)
563 # for the regexp and its replacement to avoid backslash
564 # substitution within those strings.
565 #RegexpSymbolTransform(r'release-(\d+)_(\d+)',
567 #RegexpSymbolTransform(r'release-(\d+)_(\d+)_(\d+)',
568 # r'release-\1.\2.\3'),
570 # Simple 1:1 character replacements can also be done. The
571 # following transform, which converts backslashes into forward
572 # slashes, should usually be included:
573 ReplaceSubstringsSymbolTransform('\\','/'),
575 # This last rule eliminates leading, trailing, and repeated
576 # slashes within the output symbol names:
577 NormalizePathsSymbolTransform(),
580 # See the definition of global_symbol_strategy_rules above for a
581 # description of this option:
582 symbol_strategy_rules=global_symbol_strategy_rules,