NiBabel

Access a cacophony of neuro-imaging file formats

Table Of Contents

Previous topic

Design of data packages for the nipy suite

Next topic

DICOM concepts and implementations

Reggie -- the one

A guide to making a nibabel release

A guide for developers who are doing a nibabel release

  • Edit info.py and bump the version number

Release tools

In the tools directory, among other files, you will find the following utilities:

tools/
|- build_release
|- release
|- compile.py
|- make_tarball.py
|- toollib.py

There are also some release utilities in nisext/testers.py, with makefile targets for their use. The relevant targets are:

make check-version-info
make sdist-tests

The first installs the code from a git archive, from the repository, and for in-place use, and runs the get_info() function to confirm that installation is working and information parameters are set correctly.

The second (sdist-tests) makes an sdist source distribution archive, installs it to a temporary directory, and runs the tests of that install.

Release checklist

  • Review the open list of issues . Check whether there are outstanding issues that can be closed, and whether there are any issues that should delay the release. Label them !

  • Review and update the release notes. Review and update the Changelog file. Get a partial list of contributors with something like:

    git log 0.9.0.. | grep '^Author' | cut -d' ' -f 2- | sort | uniq

    where 0.9.0 was the last release tag name.

    Then manually go over the git log to make sure the release notes are as complete as possible and that every contributor was recognized.

  • Make sure all tests pass:

    nosetests --with-doctest nibabel
  • Make sure all tests pass from sdist:

    make sdist-tests
  • First pass run build_release from the tools directory:

    cd tools
    ./build_release
  • The release should now be ready.

  • Edit nibabel/info.py to set _version_extra to ''; commit

  • Once everything looks good, run release from the tools directory.

  • Tag the release with tag of form 1.0.0:

    git tag -am 'First public release' 1.0.0
  • Now the version number is OK, push the docs to sourceforge with:

    make upload-htmldoc-mysfusername

    where mysfusername is obviously your own sourceforge username.

  • Set up maintenance / development branches

    If this is this is a full release you need to set up two branches, one for further substantial development (often called ‘trunk’) and another for maintenance releases.

    • Branch to maintainance:

      git co -b maint/1.0.0

      Set _version_extra back to .dev and bump _version_micro by 1. Thus the maintenance series will have version numbers like - say - ‘1.0.1.dev’ until the next maintenance release - say ‘1.0.1’. Commit.

    • Start next development series:

      git co main-master

      then restore .dev to _version_extra, and bump _version_minor by 1. Thus the development series (‘trunk’) will have a version number here of ‘1.1.0.dev’ and the next full release will be ‘1.1.0’.

    If this is just a maintenance release from maint/1.0.0 or similar, just tag and set the version number to - say - 1.0.2.dev.

  • Make next development release tag

    After each release the master branch should be tagged with an annotated (or/and signed) tag, naming the intended next version, plus an ‘upstream/’ prefix and ‘dev’ suffix. For example ‘upstream/1.0.0.dev’ means “development start for upcoming version 1.0.0.

    This tag is used in the Makefile rules to create development snapshot releases to create proper versions for those. The version derives its name from the last available annotated tag, the number of commits since that, and an abbrevated SHA1. See the docs of git describe for more info.

    Please take a look at the Makefile rules devel-src, devel-dsc and orig-src.

  • Announce to the mailing lists.