When the container is mounted, the local contents of . obscure
/app/symposion_app in the image.
Generally speaking, this is handy for development, as it means that
local changes are detected and used immediately without needing to
restart the container.
However, it breaks in the specific case of the sass->css
generation. Prior to this change, the css is generated only after the
first time a page is hit. The generated file is placed in
static/build; however, due to the obscuration, this generated file
isn't visible to the running process.
The next time the container is built, the pre-existing static/build
directory is copied into the container as it's being built; then,
later, that version is what gets served.
This change adds the needed libraries to pre-generate the css as part
of the image build, and runs compilescss to do this, prior to the
collectstatic step.
It also adds a second collectstatic into the make_dev_container
script, so that the visible static/build directory should ahve the
same contents as the obscured static/build directory. It also expands
the .dockerignore file to make sure these files aren't copied into the
image in future.
I'm not sure if this is the right thing to do, as changes to this
directory will be ignored, which could be confusing. Perhaps never
being able to see these generated files is better?
* developers can use --target symposion_dev to get a responsive site
that reads from their homedir and reacts to changed files instantly
* without a specified target the default is to build the prod image,
which is identical except for running uwsgi instead of the django
built-in server
* Enable debug when running in a developer's test environment
* Remove the makemigrations script and dockerfile
This creates more intermediates, but moves those that are less likely
to change to the top of the file. In theory this will produce faster
builds on a developer's machine as they won't need to apt-get update
every time.
Switch Dockerfile from CMD to ENTRYPOINT so that flags can be passed
to the makemigrations command
This is required in cases where we need to use django's makemigrations
--merge to merge two migrations.
* Adds an env variable to .env to turn on dev mode
* Uses the inbuilt django auth rather than saml when in dev mode
* For laziness, re-uses the admin login form for the non-admin login
make_dev_container now comes up ready to be logged into by any
user. No non-admin users are preconfigured; but you can add one at
http://localhost:28000/admin/auth/user/add/, log out, and then hit
http://localhost:28000/ to log in as the new user.
* Adds a laptop-mode-env file which docker can read env variables from
* Adds a Dockerfile.makemigrations; mostly identical to the main
Dockerfile. Important difference: instead of the source being copied
into the docker image at build time, it's mounted from the local
machine at run time.
* Adds a makemigrations shell script which builds an imagine using the
Dockefile.makemigrations and then uses it to run makemigrations
* Because the source is mounted from the local machine, any new
migrations created are dumped in the developer's git checkout ready
for adding to git.
Rename .env -> docker/laptop-mode-env
Create a ephemeral dev container to facilitate testing of
populate_inventory. It can be created, tweeked in the admin interface,
files updated in admin's populate_inventory, then redeployed to deal
with those changes.
A branch based one will need to be created to so we can do this work
without polluting the master branch.
This builds a working docker image for the application.
This is only an initial revision. It needs testing. MySQL, PIL, etc.
Was going to use Alpine, however, python3.6 segfaulted on docker
starting up, and I couldn't find the core files...
