releasing.rst

How to release

• Download and unpack the marsbar example data to some directory, e.g. ~/data/marbar_example_data-0.3.

• Set the environment variable with something like:

  export MARSBAR_EG_DATAPATH=$HOME/data/marsbar_example_data-0.3
  

• Change directory to the marsbar git repository root directory (containing test.m file);

• Start matlab with SPM on the path (marsbar gets appended automatically);

• Run tests with:

  >> test
  >> data_test
  

  The second of these will ask for the location of the data directory above if you have not set it with the environment variable MARSBAR_EG_DATAPATH. data_test runs the preprocessing batch scripts and checks the results estimated in SPM against the results estimated in marsbar;

• Do this for as many versions of MATLAB and SPM as possible;

• Review the documentation;

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

  git log v0.42.. | grep '^Author' | cut -d' ' -f 2- | sort | uniq
  

  where v0.42 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.

• Bump the version number in marsbar.m and doc/source/conf.py; commit;

• Run the pre release script in <marsbar>/marsbar/release/pre_release.m:

  mkdir dist
  cd dist
  matlab
  >> addpath ../marsbar
  >> addpath ../marsbar/release
  >> pre_release
  

  This creates the marsbar code archives (.tar.gz, .zip) in the current working directory (by default);

• Check you have a copy of the m2html scripts on your default matlab path;

• Run the doc scripts to build the documentation (these depend on m2html):

  cd doc
  make clean
  make doc-release
  

• Do the full monty with:

  make clean
  make public-web
  

  This should upload the web documentation;

• Upload source code to the file release directories;

• Tag the release commit;

• Announce.