Spinning a new release

Tests

  • We use two continuous integration platforms to run tests and build binaries and documentations: GitHub Actions, ReadTheDocs.

    Finished releases are published to GitHub releases, Anaconda Cloud, and the Python Package Index.

  • Make sure that the cytoflow tests pass, both locally and on GitHub:

    nose2 -c package/nose2.cfg cytoflow.tests -N 8
    
  • Make sure the cytoflowgui tests pass. You must do this locally; I’m still working on why it doesn’t run on the CI platform.

    nose2 -c package/nose2.cfg cytoflowgui.tests -N 8
    

Documentation

  • Build the API docs and check them for completeness:

    python setup.py build_sphinx
    
  • Build the online docs and check them for completeness:

    python setup.py build_sphinx -b embedded_help
    

Packaging

  • Build the conda package locally:

    conda build package/conda_recipes/cytoflow
    
  • Install the local package in a new environment:

    conda create --name cytoflow.test --use-local cytoflow
    
  • Activate the test environment, make sure you can import cytoflow, and make sure the GUI runs:

    conda activate cytoflow.test
    python -c "import cytoflow"
    cytoflow
    
  • Make sure that the pyinstaller distribution will build on your local machine (back in your development environment).

    pyinstaller package/pyinstaller.spec
    
  • Make sure that the ReadTheDocs build is working.

  • Make sure that pyinstaller built the executables on all three supported platforms. Download and test that all three start and can run a basic workflow.

Versioning and dependencies

  • We’re using versioneer to manage versions. No manual versions required.

  • If there are dependencies that don’t have packages on Anaconda, add recipes to package/conda_recipes (using conda skeleton) and upload them to the Anaconda Cloud. Unless there’s a really (really!) good reason, please make them no-arch.

  • Make sure install_requires in setup.py matches requirements.txt

  • Update the README.rst from the README.md. From the project root, say:

    pandoc --from=markdown --to=rst --output=README.rst README.md
    

Tag and upload the release

  • Push the updated docs to GitHub. Give the CI builders ~30 minutes, then check the build status on GitHub and ReadTheDocs.
  • Create a new tag on the master branch. This will re-build everything on the CI builders.
  • Download the artifacts.

Sign the Windows installer

To get rid of the “Unknown developer” warning in Windows, we sign the installer. This requires a hardware crypto token, so it must be done locally.

  • Setup: If not done already, download and install the Windows Platform SDK. I’m using 8.1 because I couldn’t get 10 to install.

  • Download the Windows installer from Github.

  • Open a terminal in C:Program FilesMicrosoft Platform SDKBin.

  • Start the signing wizard:

    signtool.exe signwizard
    
  • Select the installer binary.

  • Under “Signing options”, choose “Typical”

  • Under “Signature Certificate”, choose “Select from store…”. If the hardware key is installed and set up properly, Windows should find the correct certificate.

  • Add a description such as “Flow cytometry software”. For “Web location”, specify “http://cytoflow.readthedocs.org

  • Check the box next to “Add a timestamp to data”. Enter “http://time.certum.pl”. (Probably could use digicert or some other service.)

  • When prompted, enter the Common Profile PIN.

  • After the wizard closes, double-check that the signing process was completed by right-clicking on the executable and checking the “Digital Signatures” tab.

Upload the artifacts and update the homepage

  • Upload artifacts as appropriate to GitHub, Anaconda, and the Python Package Index. (Make sure that in the case of Anaconda, you’re uploading to the organization account, not your personal account!) The GitHub action should take care of the GitHub and Anaconda packages, but not PyPI.
  • At https://github.com/cytoflow/cytoflow.github.io, update the version in _config.yml. Push these changes to update the main download links on http://cytoflow.github.io/
  • Verify that the download links at http://cytoflow.github.io/ still work!