- simplify the process by having the riddler container build the rootfs and config
- output tarred up rootfs and config.json as otherwise file ownership not preserved
- allow easy build of a collection of container tarballs with another conversion script
This makes it easy to choose which container images you want and just convert any
set to a initrd image
```
tar cf - container1.tar container2.tar | docker run -i tartar2initrd > initrd.img
```
Next stage will use a manifest to select the ones to add for each edition.
Signed-off-by: Justin Cormack <justin.cormack@docker.com>
This is mainly for testing.
You can run with something like
```
docker run -v ~/.ssh/id_rsa.pub:/root/.ssh/authorized_keys -p 2222:22 -e TINI_SUBREAPER=1 --pid=host mobylinux/sshd:36c44542d8120e384c724a078e3e489f7a58382c
```
if you want to run in the host pid namespace. You must provide an
authorized key, you may provide a host key.
Signed-off-by: Justin Cormack <justin.cormack@docker.com>
Using docker build is slower and needs lots of Dockerfiles,
while a single image with a careful script can accept any type
of image, either with `-v` to share into `/tmp` for interactive
use (where you need the input and a tty, or by adding a tarball
for cases where there is no login such as running tests, so you
can still use a remote daemon in these cases.
Signed-off-by: Justin Cormack <justin.cormack@docker.com>
