Tai branch

[jmurty]

@mrmachine Can you please summarise the changes on the tai branch?

[mrmachine]

• Update settings for compatibility with old Django versions and consistency.
  • Use tuple instead of list for template dirs, context processors, loaders.
  • Define only current version settings (e.g. TEMPLATES, MIDDLEWARE) and then add compatibility for earlier Django versions at the bottom of ixc_django_docker/settings/__init__.py.
  • Fix debug and request context processor paths for Django <1.8.

• setup.sh and setup-tests.sh no longer automatically run any npm script, even if SETUP_NPM_RUN or SETUP_TESTS_NPM_RUN is defined.

  Instead, export SETUP_COMMAND and SETUP_TESTS_COMMAND, which can be any shell command instead of the name of an npm script.

  This allows us to execute any combination of npm scripts in sequence (via run-s) or parallel (via run-p), without cluttering up package.json with near identical wrapper scripts.

• Store nginx and supervisor pid files in a dedicated run directory, optionally namespaced by $CONTEXT. This allows multiple instances of nginx and supervisor (e.g. gunicorn, twisted) to run with a shared var volume.
• Use .$(uname) as a filename suffix to avoid conflicts between native (macOS) and Docker (Linux) environments.
• Apply migrations last via setup.sh. Because the database is shared state and there might be backwards incompatible changes. This should reduce the duration of any potential downtime before recreating old containers, caused by a slow SETUP_COMMAND.
• Use python -m pip (best practice) to be more sure that we are installing into the right environment.
• Fix XViewMiddleware name in Django <1.8.
• Update datadog env vars. DATADOG_ENV -> DD_ENV, per current ddtrace convention.
• Delete duplicate/vestigial Django >=1.8 only MIDDLEWARE and TEMPLATES settings on earlier versions.
• Fix npm-install.sh when run interactively inside containers as root.
• Use a unique name for the django-compressor and ixc-whitenoise manifest files, so we can run new and old versions of a codebase side by side with a shared volume.