4 Theory and requirements
5 ------ --- ------------
7 There are two main problem converting a CVS repository to SVN:
9 - CVS does not record enough information to determine what actually
10 happened to a repository. For example, CVS does not record:
12 - Which file modifications were part of the same commit
14 - The timestamp of tag and branch creations
16 - Exactly which revision was the base of a branch (there is
17 ambiguity between x.y, x.y.2.0, x.y.4.0, etc.)
19 - When the default branch was changed (for example, from a vendor
20 branch back to trunk).
22 - The timestamps in a CVS archive are not reliable. It can easily
23 happen that timestamps are not even monotonic, and large errors (for
24 example due to a failing server clock battery) are not unusual.
26 The absolutely crucial, sine qua non requirement of a conversion is
27 that the dependency relationships within a file be honored, mainly:
29 - A revision depends on its predecessor
31 - A branch creation depends on the revision from which it branched,
32 and commits on the branch depend on the branch creation
34 - A tag creation depends on the revision being tagged
36 These dependencies are reliably defined in the CVS repository, and
37 they trump all others, so they are the scaffolding of the conversion.
39 Moreover, it is highly desirable that the timestamps of the SVN
40 commits be monotonically increasing.
42 Within these constraints we also want the results of the conversion to
43 resemble the history of the CVS repository as closely as possible.
44 For example, the set of file changes grouped together in an SVN commit
45 should be the same as the files changed within the corresponding CVS
46 commit, insofar as that can be achieved in a manner that is consistent
47 with the dependency requirements. And the SVN commit timestamps
48 should recreate the time of the CVS commit as far as possible without
49 violating the monotonicity requirement.
51 The basic idea of the conversion is this: create the largest
52 conceivable changesets, then split up changesets as necessary to break
53 any cycles in the graph of changeset dependencies. When all cycles
54 have been removed, then do a topological sort of the changesets (with
55 ambiguities resolved using CVS timestamps) to determine a
56 self-consistent changeset commit order.
58 The quality of the conversion (not in terms of correctness, but in
59 terms of minimizing the number of svn commits) is mostly determined by
60 the cleverness of the heuristics used to split up cycles. And all of
61 this has to be affordable, especially in terms of conversion time and
62 RAM usage, for even the largest CVS repositories.
68 A cvs2svn run consists of a number of passes. Each pass saves the
69 data it produces to files on disk, so that a) we don't hold huge
70 amounts of state in memory, and b) the conversion process is
73 CollectRevsPass (formerly called pass1)
76 The goal of this pass is to collect from the CVS files all of the data
77 that will be required for the conversion. If the --use-internal-co
78 option was used, this pass also collects the file delta data; for
79 -use-rcs or -use-cvs, the actual file contents are read again in
82 To collect this data, we walk over the repository, collecting data
83 about the RCS files into an instance of CollectData. Each RCS file is
84 processed with rcsparse.parse(), which invokes callbacks from an
85 instance of cvs2svn's _FileDataCollector class (which is a subclass of
88 While a file is being processed, all of the data for the file (except
89 for contents and log messages) is held in memory. When the file has
90 been read completely, its data is converted into an instance of
91 CVSFileItems, and this instance is manipulated a bit then pickled and
92 stored to 'cvs-items.pck'.
94 For each RCS file, the first thing the parser encounters is the
95 administrative header, including the head revision, the principal
96 branch, symbolic names, RCS comments, etc. The main thing that
97 happens here is that _FileDataCollector.define_tag() is invoked on
98 each symbolic name and its attached revision, so all the tags and
99 branches of this file get collected.
101 Next, the parser hits the revision summary section. That's the part
102 of the RCS file that looks like this:
105 date 2002.06.12.04.54.12; author captnmark; state Exp;
111 date 2002.05.28.18.02.11; author captnmark; state Exp;
117 For each revision summary, _FileDataCollector.define_revision() is
118 invoked, recording that revision's metadata in various variables of
119 the _FileDataCollector class instance.
121 Next, the parser encounters the *real* revision data, which has the
122 log messages and file contents. For each revision, it invokes
123 _FileDataCollector.set_revision_info(), which sets some more fields in
124 _RevisionData. It also invokes RevisionRecorder.record_text(), which
125 gives the RevisionRecorder the chance to record the file text if
126 desired. record_test() is allowed to return a token, which is carried
127 along with the CVSRevision data and can be used by RevisionReader to
128 retrieve the text in OutputPass.
130 When the parser is done with the file, _ProjectDataCollector takes the
131 resulting CVSFileItems object and manipulates it to handle some CVS
134 - If the file had a vendor branch, make some adjustments to the
135 file dependency graph to reflect implicit dependencies related to
136 the vendor branch. Also delete the 1.1 revision in the usual
137 case that it doesn't contain any useful information.
139 - If the file was added on a branch rather than on trunk, then
140 delete the "dead" 1.1 revision on trunk in the usual case that it
141 doesn't contain any useful information.
143 - If the file was added on a branch after it already existed on
144 trunk, then recent versions of CVS add an extra "dead" revision
145 on the branch. Remove this revision in the usual case that it
146 doesn't contain any useful information, and sever the branch from
147 trunk (since the branch version is independent of the trunk
150 - If the conversion was started with the --trunk-only option, then
152 1. graft any non-trunk default branch revisions onto trunk
153 (because they affect the history of the default branch), and
155 2. delete all branches and tags and all remaining branch
158 Finally, the RevisionRecorder.finish_file() callback is called, the
159 CVSFileItems instance is stored to a database, and statistics about
160 how symbols were used in the file are recorded.
162 That's it -- the RCS file is done.
164 When every CVS file is done, CollectRevsPass is complete, and:
166 - The basic information about each file and directory (filename,
167 path, etc) is written as a pickled CVSPath instance to
170 - Information about each symbol seen, along with statistics like
171 how often it was used as a branch or tag, is written as a pickled
172 symbol_statistics._Stat object to 'symbol-statistics.pck'. This
173 includes the following information:
175 ID -- a unique positive identifying integer
177 NAME -- the symbol name
179 TAG_CREATE_COUNT -- the number of times the symbol was used
182 BRANCH_CREATE_COUNT -- the number of times the symbol was
185 BRANCH_COMMIT_COUNT -- the number of files in which there was
186 a commit on a branch with this name.
188 BRANCH_BLOCKERS -- the set of other symbols that ever
189 sprouted from a branch with this name. (A symbol cannot
190 be excluded from the conversion unless all of its
191 blockers are also excluded.)
193 POSSIBLE_PARENTS -- a count of in how many files each other
194 branch could have served as the symbol's source.
196 These data are used to look for inconsistencies in the use of
197 symbols under CVS and to decide which symbols can be excluded or
198 forced to be branches and/or tags. The POSSIBLE_PARENTS data is
199 used to pick the "optimum" parent from which the symbol should
200 sprout in as many files as possible.
202 For a multiproject conversion, distinct symbol records (and IDs)
203 are created for symbols in separate projects, even if they have
204 the same name. This is to prevent symbols in separate projects
205 from being filled at the same time.
207 - Information about each CVS event is converted into a CVSItem
208 instance and stored to 'cvs-items.pck'. There are several types
211 CVSRevision -- A specific revision of a specific CVS file.
213 CVSBranch -- The creation of a branch tag in a specific CVS
216 CVSTag -- The creation of a non-branch tag in a specific CVS
219 The CVSItems are grouped into CVSFileItems instances, one per
220 CVSFile. But a multi-file commit will still be scattered all
223 - Selected metadata for each CVS revision, including the author and
224 log message, is written to 'metadata-index.dat' and
225 'metadata.pck'. The purpose is twofold: first, to save space by
226 not having to save this information multiple times, and second
227 because CVSRevisions that have the same metadata are candidates
228 to be combined into an SVN changeset.
230 First, an SHA digest is created for each set of metadata. The
231 digest is constructed so that CVSRevisions that can be combined
232 are all mapped to the same digest. CVSRevisions that were part
233 of a single CVS commit always have a common author and log
234 message, therefore these fields are always included in the
237 - if ctx.cross_project_commits is False, we avoid combining CVS
238 revisions from separate projects by including the project.id in
241 - if ctx.cross_branch_commits is False, we avoid combining CVS
242 revisions from different branches by including the branch name
245 During the database creation phase, the database keeps track of a
248 digest (20-byte string) -> metadata_id (int)
250 to allow the record for a set of metadata to be located
251 efficiently. As data are collected, it stores a map
253 metadata_id (int) -> (author, log_msg,) (tuple)
255 into the database for use in future passes. CVSRevision records
256 include the metadata_id.
258 During this run, each CVSFile, Symbol, CVSItem, and metadata record is
259 assigned an arbitrary unique ID that is used throughout the conversion
266 Encode the cvs revision metadata as UTF-8, ensuring that all entries
267 can be decoded using the chosen encodings. Output the results to
268 'metadata-clean-index.dat' and 'metadata-clean.pck'.
274 Use the symbol statistics collected in CollectRevsPass and any runtime
275 options to determine which symbols should be treated as branches,
276 which as tags, and which should be excluded from the conversion
279 Create 'symbols.pck', which contains a pickle of a list of TypedSymbol
280 (Branch, Tag, or ExcludedSymbol) instances indicating how each symbol
281 should be processed in the conversion. The IDs used for a TypedSymbol
282 is the same as the ID allocated to the corresponding symbol in
283 CollectRevsPass, so references in CVSItems do not have to be updated.
289 This pass works through the CVSFileItems instances stored in
290 'cvs-items.pck', processing all of the items from each file as a
291 group. (This is the last pass in which all of the CVSItems for a file
292 are in memory at once.) It does the following things:
294 - Exclude any symbols that CollateSymbolsPass determined should be
295 excluded, and any revisions on such branches. Also delete
296 references from other CVSItems to those that are being deleted.
298 - Transform any branches to tags or vice versa, also depending on
299 the results of CollateSymbolsPass, and fix up the references from
302 - Decide what line of development to use as the parent for each
303 symbol in the file, and adjust the file's dependency tree
306 - For each CVSRevision, record the list of symbols that the
307 revision opens and closes.
309 - Write a summary of each surviving CVSRevision to
310 'revs-summary.txt'. Each line of the file has the format
312 METADATA_ID TIMESTAMP CVS_REVISION
314 where TIMESTAMP is a fixed-width timestamp, and CVS_REVISION is
315 the pickled CVSRevision in a format that does not contain any
316 newlines. These summaries will be sorted in
317 SortRevisionSummaryPass then used by InitializeChangesetsPass to
318 create preliminary RevisionChangesets.
320 - Write a summary of CVSSymbols to 'symbols-summary.txt'. Each
321 line of the file has the format
325 where CVS_SYMBOL is the pickled CVSSymbol in a format that does
326 not contain any newlines. This information will be sorted by
327 SYMBOL_ID in SortSymbolSummaryPass then used to create
328 preliminary SymbolChangesets.
331 SortRevisionSummaryPass
332 =======================
334 Sort the revision summary written by FilterSymbolsPass, creating
335 'revs-summary-s.txt'. The sort groups items that might be added to
336 the same changeset together and, within a group, sorts revisions by
337 timestamp. This step makes it easy for InitializeChangesetsPass to
338 read the initial draft of RevisionChangesets straight from the file.
341 SortSymbolSummaryPass
342 =====================
344 Sort the symbol summary written by FilterSymbolsPass, creating
345 'symbols-summary-s.txt'. The sort groups together symbol items that
346 might be added to the same changeset (though not in anything
347 resembling chronological order). The output of this pass is used by
348 InitializeChangesetsPass.
351 InitializeChangesetsPass
352 ========================
354 This pass creates first-draft changesets, splitting them using
355 COMMIT_THRESHOLD and breaking up any revision changesets that have
356 internal dependencies.
358 The raw material for creating revision changesets is
359 'revs-summary-s.txt', which already has CVSRevisions sorted in such a
360 way that potential changesets are grouped together and sorted by date.
361 The contents of this file are read line by line, and the corresponding
362 CVSRevisions are accumulated into a changeset. Whenever the
363 metadata_id changes, or whenever there is a time gap of more than
364 COMMIT_THRESHOLD (currently set to 5 minutes) between CVSRevisions,
365 then a new changeset is started.
367 At this point a revision changeset can have internal dependencies if
368 two commits were made to the same file with the same log message
369 within COMMIT_THRESHOLD of each other. The next job of this pass is
370 to split up changesets in such a way to break such internal
371 dependencies. This is done by sorting the CVSRevisions within a
372 changeset by timestamp, then choosing the split point that breaks the
373 most internal dependencies. This procedure is continued recursively
374 until there are no more dependencies internal to a single changeset.
376 Analogously, the CVSSymbol items from 'symbols-summary-s.txt' are
377 grouped into symbol changesets. (Symbol changesets cannot have
378 internal dependencies, so there is no need to break them up at this
381 Finally, this pass writes a CVSItem database with the CVSItems written
382 in order grouped by the preliminary changeset to which they belong.
383 Even though the preliminary changesets still have to be split up to
384 form final changesets, grouping the CVSItems this way improves the
385 locality of disk accesses and thereby speeds up later passes.
387 The result of this pass is two databases:
389 - 'cvs-item-to-changeset.dat', which maps CVSItem ids to the id of
390 the changeset containing the item, and
392 - 'changesets.pck' and 'changesets-index.dat', which contain the
393 changeset objects themselves, indexed by changeset id.
395 - 'cvs-items-sorted-index.dat' and 'cvs-items-sorted.pck', which
396 contain the pickled CVSItems ordered by changeset.
399 BreakRevisionChangesetCyclesPass
400 ================================
402 There can still be cycles in the dependency graph of
403 RevisionChangesets caused by:
405 - Interleaved commits. Since CVS commits are not atomic, it can
406 happen that two commits are in progress at the same time and each
407 alters the same two files, but in different orders. These should
408 be small cycles involving only a few revision changesets. To
409 resolve these cycles, one or more of the RevisionChangesets have
410 to be split up (eventually becoming separate svn commits).
412 - Cycles involving a RevisionChangeset formed by the accidental
413 combination of unrelated items within a short period of time that
414 have the same author and log message. These should also be small
415 cycles involving only a few changesets.
417 The job of this pass is to break up such cycles (those involving only
420 This pass works by building up the graph of revision changesets and
421 their dependencies in memory, then attempting a topological sort of
422 the changesets. Whenever the topological sort stalls, that implies
423 the existence of a cycle, one of which can easily be determined. This
424 cycle is broken through the use of heuristics that try to determine an
425 "efficient" way of splitting one or more of the changesets that are
428 The new RevisionChangesets are written to
429 'cvs-item-to-changeset-revbroken.dat', 'changesets-revbroken.pck', and
430 'changesets-revbroken-index.dat', along with the unmodified
431 SymbolChangesets. These files are in the same format as the analogous
432 files produced by InitializeChangesetsPass.
435 RevisionTopologicalSortPass
436 ===========================
438 Topologically sort the RevisionChangesets, thereby picking the order
439 in which the RevisionChangesets will be committed. (Since the
440 previous pass eliminated any dependency cycles, this sort is
441 guaranteed to succeed.) Ambiguities in the topological sort are
442 resolved using the changesets' timestamps. Then simplify the
443 changeset graph into a linear chain by converting each
444 RevisionChangeset into an OrderedChangeset that stores dependency
445 links only to its commit-order predecessor and successor. This
446 simplified graph enforces the commit order that resulted from the
447 topological sort, even after the SymbolChangesets are added back into
448 the graph later. Store the OrderedChangesets into
449 'changesets-revsorted.pck' and 'changesets-revsorted-index.dat' along
450 with the unmodified SymbolChangesets.
453 BreakSymbolChangesetCyclesPass
454 ==============================
456 It is possible for there to be cycles in the graph of SymbolChangesets
459 - Split creation of branches. It is possible that branch A depends
460 on branch B in one file, but B depends on A in another file.
461 These cycles can be large, but they only involve
464 Break up such dependency loops. Output the results to
465 'cvs-item-to-changeset-symbroken.dat',
466 'changesets-symbroken-index.dat', and 'changesets-symbroken.pck'.
469 BreakAllChangesetCyclesPass
470 ===========================
472 The complete changeset graph (including both RevisionChangesets and
473 BranchChangesets) can still have dependency cycles cause by:
475 - Split creation of branches. The same branch tag can be added to
476 different files at completely different times. It is possible
477 that the revision that was branched later depends on a
478 RevisionChangeset that involves a file on the branch that was
479 created earlier. These cycles can be large, but they always
480 involve a SymbolChangeset. To resolve these cycles, the
481 SymbolChangeset is split up into two changesets.
483 In fact, tag changesets do not have to be considered--CVSTags cannot
484 participate in dependency cycles because no other CVSItem can depend
487 Since the input of this pass has been through
488 RevisionTopologicalSortPass, all revision cycles have already been
489 broken up and the order that the RevisionChangesets will be committed
490 has been determined. In this pass, the complete changeset graph is
491 created in memory, including the linear list of OrderedChangesets from
492 RevisionTopologicalSortPass plus all of the symbol changesets.
493 Because this pass doesn't break up any OrderedChangesets, it is
494 constrained to finding places within the revision changeset sequence
495 in which the symbol changeset commits can be inserted.
497 The new changesets are written to
498 'cvs-item-to-changeset-allbroken.dat', 'changesets-allbroken.pck', and
499 'changesets-allbroken-index.dat', which are in the same format as the
500 analogous files produced by InitializeChangesetsPass.
506 Now that the earlier passes have broken up any dependency cycles among
507 the changesets, it is possible to order all of the changesets in such
508 a way that all of a changeset's dependencies are committed before the
509 changeset itself. This pass does so by again building up the graph of
510 changesets in memory, then at each step picking a changeset that has
511 no remaining dependencies and removing it from the graph. Whenever
512 more than one dependency-free changeset is available, symbol
513 changesets are chosen before revision changesets. As changesets are
514 processed, the timestamp sequence is ensured to be monotonic by the
515 simple expedient of adjusting retrograde timestamps to be later than
516 their predecessor. Timestamps that lie in the future, on the other
517 hand, are assumed to be bogus and are adjusted backwards, also to be
518 just later than their predecessor.
520 This pass writes a line to 'changesets-s.txt' for each
521 RevisionChangeset, in the order that the changesets should be
522 committed. Each lines contains
524 CHANGESET_ID TIMESTAMP
526 where CHANGESET_ID is the id of the changeset in the
527 'changesets-allbroken' databases and TIMESTAMP is the timstamp that
528 should be assigned to it when it is committed. Both values are
529 written in hexadecimal.
532 CreateRevsPass (formerly called pass5)
535 This pass generates SVNCommits from Changesets and records symbol
536 openings and closings. (One Changeset can result in multiple
537 SVNCommits, for example if it causes symbols to be filled or copies to
540 This pass does the following:
542 1. Creates a database file to map Subversion revision numbers to
543 SVNCommit instances ('svn-commits-index.dat' and
544 'svn-commits.pck'). Creates another database file to map CVS
545 Revisions to their Subversion Revision numbers
546 ('cvs-revs-to-svn-revnums.db').
548 2. When a file is copied to a symbolic name in cvs2svn, it is copied
549 from a specific source: either a CVSRevision, or a copy created by
550 a previous CVSBranch of the file. The copy has to be made from an
551 SVN revision that is during the lifetime of the source. The SVN
552 revision when the source was created is called the symbol's
553 "opening", and the SVN revision when it was deleted or overwritten
554 is called the symbol's "closing". In this pass, the
555 SymbolingsLogger class writes out a line to 'symbolic-names.txt'
556 for each symbol opening or closing. Note that some openings do not
557 have closings, namely if the corresponding source is still present
558 at the HEAD revision.
560 The format of each line is:
562 SYMBOL_ID SVN_REVNUM TYPE CVS_SYMBOL_ID
571 Here is what the columns mean:
573 SYMBOL_ID -- The id of the branch or tag that has an opening in
574 this SVN_REVNUM, in hexadecimal.
576 SVN_REVNUM -- The Subversion revision number in which the opening
577 or closing occurred. (There can be multiple openings and
578 closings per SVN_REVNUM).
580 TYPE -- "O" for openings and "C" for closings.
582 CVS_SYMBOL_ID -- The id of the CVSSymbol instance whose opening or
583 closing is being described, in hexadecimal.
585 Each CVSSymbol that tags a non-dead file has exactly one opening
586 and either zero or one closing. The closing, if it exists, always
587 occurs in a later SVN revision than the opening.
589 See SymbolingsLogger for more details.
592 SortSymbolsPass (formerly called pass6)
595 This pass sorts 'symbolic-names.txt' into 'symbolic-names-s.txt'.
596 This orders the file first by symbol ID, and second by Subversion
597 revision number, thus grouping all openings and closings for each
598 symbolic name together.
601 IndexSymbolsPass (formerly called pass7)
604 This pass iterates through all the lines in 'symbolic-names-s.txt',
605 writing out a pickle file ('symbol-offsets.pck') mapping SYMBOL_ID to
606 the file offset in 'symbolic-names-s.txt' where SYMBOL_ID is first
607 encountered. This will allow us to seek to the various offsets in the
608 file and sequentially read only the openings and closings that we
612 OutputPass (formerly called pass8)
615 This pass opens the svn-commits database and sequentially plays out
616 all the commits to either a Subversion repository or to a dumpfile.
617 It also decides what sources to use to fill symbols.
619 In --dumpfile mode, the result of this pass is a Subversion repository
620 dumpfile (suitable for input to 'svnadmin load'). The dumpfile is the
621 data's last static stage: last chance to check over the data, run it
622 through svndumpfilter, move the dumpfile to another machine, etc.
624 When not in --dumpfile mode, no full dumpfile is created. Instead,
625 miniature dumpfiles representing a single revisions are created,
626 loaded into the repository, and then removed.
628 In both modes, the dumpfile revisions are created by walking through
631 The database 'mirror-nodes.db' holds a skeletal mirror of the
632 repository structure at each SVN revision. This mirror keeps track of
633 which files existed on each LOD, but does not record any file
634 contents. cvs2svn requires this information to decide which paths to
635 copy when filling branches and tags.
637 When .cvsignore files are modified, cvs2svn computes the corresponding
638 svn:ignore properties and applies the properties to the parent
639 directory. The .cvsignore files themselves are not included in the
640 output unless the --keep-cvsignore option was specified. But in
641 either case, the .cvsignore files are recorded within the repository
642 mirror as if they were being written to disk, to ensure that the
643 containing directory is not pruned if the directory in CVS still
644 contained a .cvsignore file.
647 ===============================
648 Branches and Tags Plan.
649 ===============================
651 This pass is also where tag and branch creation is done. Since
652 subversion does tags and branches by copying from existing revisions
653 (then maybe editing the copy, making subcopies underneath, etc), the
654 big question for cvs2svn is how to achieve the minimum number of
655 operations per creation. For example, if it's possible to get the
656 right tag by just copying revision 53, then it's better to do that
657 than, say, copying revision 51 and then sub-copying in bits of
660 Tags are created as soon as cvs2svn encounters the last CVS Revision
661 that is a source for that tag. The whole tag is created in one
664 Branches are created as soon as all of their prerequisites are in
665 place. If a branch creation had to be broken up due to dependency
666 cycles, then non-final parts are also created as soon as their
667 prerequisites are ready. In such a case, the SymbolChangeset
668 specifies how much of the branch can be created in each step.
670 How just-in-time branch creation works:
672 In order to make the "best" set of copies/deletes when creating a
673 branch, cvs2svn keeps track of two sets of trees while it's making
676 1. A skeleton mirror of the subversion repository, that is, a
677 record of which file existed on which LOD for each SVN revision.
679 2. A tree for each CVS symbolic name, and the svn file/directory
680 revisions from which various parts of that tree could be copied.
682 Each LOD is recorded as a tree using the following schema: unique keys
683 map to marshal.dumps() representations of dictionaries, which in turn
684 map path component names to other unique keys:
686 root_key ==> { entryname1 : entrykey1, entryname2 : entrykey2, ... }
687 entrykey1 ==> { entrynameX : entrykeyX, ... }
688 entrykey2 ==> { entrynameY : entrykeyY, ... }
689 entrykeyX ==> { etc, etc ...}
690 entrykeyY ==> { etc, etc ...}
692 (The leaf nodes -- files -- are represented by None.)
694 The repository mirror allows cvs2svn to remember what paths exist in
697 For details on how branches and tags are created, please see the
698 docstring the SymbolingsLogger class (and its methods).
700 -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*-
701 - -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -
702 -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*- -*-
704 Some older notes and ideas about cvs2svn. Not deleted, because they
705 may contain suggestions for future improvements in design.
707 -----------------------------------------------------------------------
709 An email from John Gardiner Myers <jgmyers@speakeasy.net> about some
710 considerations for the tool.
713 From: John Gardiner Myers <jgmyers@speakeasy.net>
714 Subject: Thoughts on CVS to SVN conversion
716 Date: Sun, 15 Apr 2001 17:47:10 -0700
718 Some things you may want to consider for a CVS to SVN conversion utility:
720 If converting a CVS repository to SVN takes days, it would be good for
721 the conversion utility to keep its progress state on disk. If the
722 conversion fails halfway through due to a network outage or power
723 failure, that would allow the conversion to be resumed where it left off
724 instead of having to start over from an empty SVN repository.
726 It is a short step from there to allowing periodic updates of a
727 read-only SVN repository from a read/write CVS repository. This allows
728 the more relaxed conversion procedure:
730 1) Create SVN repository writable only by the conversion tool.
731 2) Update SVN repository from CVS repository.
732 3) Announce the time of CVS to SVN cutover.
733 4) Repeat step (2) as needed.
734 5) Disable commits to CVS repository, making it read-only.
736 7) Enable commits to SVN repository.
737 8) Wait for developers to move their workspaces to SVN.
738 9) Decomission the CVS repository.
740 You may forward this message or parts of it as you seem fit.
743 -----------------------------------------------------------------------
745 Further design thoughts from Greg Stein <gstein@lyra.org>
747 * timestamp the beginning of the process. ignore any commits that
748 occur after that timestamp; otherwise, you could miss portions of a
749 commit (e.g. scan A; commit occurs to A and B; scan B; create SVN
750 revision for items in B; we missed A)
752 * the above timestamp can also be used for John's "grab any updates
753 that were missed in the previous pass."
755 * for each file processed, watch out for simultaneous commits. this
756 may cause a problem during the reading/scanning/parsing of the file,
757 or the parse succeeds but the results are garbaged. this could be
758 fixed with a CVS lock, but I'd prefer read-only access.
760 algorithm: get the mtime before opening the file. if an error occurs
761 during reading, and the mtime has changed, then restart the file. if
762 the read is successful, but the mtime changed, then restart the
765 * use a separate log to track unique branches and non-branched forks
766 of revision history (Q: is it possible to create, say, 1.4.1.3
767 without a "real" branch?). this log can then be used to create a
768 /branches/ directory in the SVN repository.
770 Note: we want to determine some way to coalesce branches across
771 files. It can't be based on name, though, since the same branch name
772 could be used in multiple places, yet they are semantically
773 different branches. Given files R, S, and T with branch B, we can
774 tie those files' branch B into a "semantic group" whenever we see
775 commit groups on a branch touching multiple files. Files that are
776 have a (named) branch but no commits on it are simply ignored. For
777 each "semantic group" of a branch, we'd create a branch based on
778 their common ancestor, then make the changes on the children as
779 necessary. For single-file commits to a branch, we could use
780 heuristics (pathname analysis) to add these to a group (and log what
781 we did), or we could put them in a "reject" kind of file for a human
782 to tell us what to do (the human would edit a config file of some
783 kind to instruct the converter).
785 * if we have access to the CVSROOT/history, then we could process tags
786 properly. otherwise, we can only use heuristics or configuration
787 info to group up tags (branches can use commits; there are no
788 commits associated with tags)
790 * ideally, we store every bit of data from the ,v files to enable a
791 complete restoration of the CVS repository. this could be done by
792 storing properties with CVS revision numbers and stuff (i.e. all
793 metadata not already embodied by SVN would go into properties)
795 * how do we track the "states"? I presume "dead" is simply deleting
796 the entry from SVN. what are the other legal states, and do we need
797 to do anything with them?
799 * where do we put the "description"? how about locks, access list,
802 * note that using something like the SourceForge repository will be an
803 ideal test case. people *move* their repositories there, which means
804 that all kinds of stuff can be found in those repositories, from
805 wherever people used to run them, and under whatever development
806 policies may have been used.
808 For example: I found one of the projects with a "permissions 644;"
809 line in the "gnuplot" repository. Most RCS releases issue warnings
810 about that (although they properly handle/skip the lines), and CVS
811 ignores RCS newphrases altogether.