Created by: wangkuiyi

I noticed that we need this so we can re-use Docker images for development easily. In particular, suppose that I cloned Paddle into ~/work/paddle on my MacBook Pro, and is going to change/develop Paddle's C++ source code, I would

1. Build a Docker image paddle:dev:

   cd ~/work/paddle
   docker build -t paddle:dev -f paddle/scripts/docker/Dockerfile .

   This copies ~/work/paddle into the image as /paddle.

2. Run the Docker image in a container with ~/work/paddle on the host mounted into /paddle:

   docker run -d -p 2022:22 -v /Users/yi/work/paddle:/paddle paddle:dev

   This mounts ~/work/paddle (aka /Users/yi/work/paddle) to /paddle in the container. This actually overlays the directory over the previously copied version.

3. Edit files in ~/work/paddle/ on the host, and build the edited source files from within the container:

   ssh root@localhost -p 2022
   $ cd /paddle
   $ WITH_GPU=OFF WITH_DOC=OFF WITH_STYLE_CHECK=ON WITH_SWIG_PY=ON paddle/scripts/docker/build.sh
   $ make -j4

   The cmake command creates /paddle/build directory in the container and makes make saves intermediate files in there. This /paddle/build is actually created on the host as ~/work/paddle/build, and intermediate files will be saved on the host and will be re-used the next time we change code and re-build.