This is slightly patched clone of cvs2hg, adapted to work on Mercurial >= 3.1. I derived it from original repo which is available at http://hg.gerg.ca/cvs2svn (and which was created as fork of cvs2svn available at http://cvs2svn.tigris.org/cvs2svn.html)
cvs2hg converts CVS repositories into Mercurial ones and does much better job than hg convert preserving appropriate tags and branchpoints contents.
Most important case is that of manually tweaked tags. Example scenario:
cvs commit file1 file2 file3 file4 cvs tag blah_1-0-0 # covers 1.1 versions of all files cvs commit file2 file3 file1 # makes 1.2 cvs commit file1 # makes 1.3 cvs tag -r blah_1-0-0 blah_1-0-1 cvs tag -F blah_1-0-1 file1
(tag blah_1-0-0 is created as copy of blah_1-0-0, but manually moved to newer version of file1 – fairly frequent way of backporting bugfixes without CVS branch creation).
Even simpler scenario: some tag (on purpose or accidentally) omits some files.
hg convert is lost in such cases. In aggregated history there is no point of time at which file2, file3, and file4 were at 1.1, but file1 was at 1.3. So hg convert binds blah_1-0-1 at some random place (usually nearby initial revision).
cvs2hg resolves that by creating fixup commit - artificial change which brings repository content in sync with tag being converted (removes files which were omitted by cvs tag, applies partial changes to selected files only…). The net result is that:
hg up -r blah_1-0-1
gives exactly the same files, with exactly the same contents, as:
cvs up -r blah_1-0-1
The cost is that history contains those commits - in most cases tiny triangular branches like:
| o |\ | | | | | o [blah_1-0-1] <--- fixup commit |/ | | o |
The same problem and solution applies to branches which started from such a tag (there fixup commit may init longer branch).
See also longer discussion in README.cvs2hg in the source code.
Clone this repository:
hg clone https://bitbucket.org/Mekk/mercurial-cvs2hg
then install the module:
cd mercurial-cvs2hg sudo python setup.py install
(or install in virtualenv, or install locally using pip).
Usual way to run the conversion:
Grab ,v files of module being converted somehow (you need read access to the server CVS repository). To run the conversion, you need the following directory tree (say you convert libs/acme which has some files and also doc subdir):
someWorkDir | +--- CVSROOT (can be empty directory, just marks where root is) | | +--- libs | | +--- acme | +-- something.h,v | +-- something.c,v | +-- doc | +-- README.txt,v | +-- Attic | +-- BUGS.txt,v
Run the command:
cd someWorkDir cvs2hg --hgrepos=$HOME/repos/libs/acme \ --encoding utf8 --encoding iso-8859-2 \ libs/acme
(adapt encoding to your needs, that's list of encodings which could be used in commit messages)
Examine resulting Mercurial repository (with the command above it would be $HOME/repos/libs/acme). At times some history edition may make sense (like dropping unnecessary branches or tags, or even performing Mercurial to Mercurial conversion to modify usernames).