How to write a Dockerfile for gnu-poke

image credits

GNU poke is an interactive, extensible editor for binary data. Not limited to editing basic entities such as bits and bytes, it provides a full-fledged procedural, interactive programming language designed to describe data structures and to operate on them. – gnu-poke

gnu-poke 1.0 was released, but I didn’t find a corresponding package for any Linux distribution. So I build my own version and wrote a Dockerfile for it.


Bevor I write a Dockerfile I start a docker image, e.g. docker run -it debian:bullseye /bin/bash. Then I run all commands inside of the docker image and document each individual step in the Dockerfile.

For gnu-poke I use Debian bullseye as base distribution, because the library gettext should be >= version 0.21.

FROM debian:bullseye

The gnu-plot project describes the necessary libraries in two files. In the DEPENDENCIES file the description is detailed in the HACKING file with some kind of prose📙. But unfortunately, only some of the necessary libraries are described.

I collected all necessary libraries and added them in the Dockerfile:

RUN apt-get update
RUN apt-get install -y build-essential \
  libgc-dev libreadline-dev libjson-c-dev tcl-dev tk-dev libnbd-dev \
  gettext pkg-config \
  automake dejagnu flex bison \
  libtool texinfo autopoint help2man gawk \

I found the appropriate libraries by trial and error and with the help of the Debian package search (e.g. gettext).

The released tar file does not work. Therefore, I need git and also for the bootstrap call. This call is described in the HACKING file:

RUN git clone /poke


RUN ./bootstrap --skip-po

Now I can run the usual configure and make:

RUN mkdir build/ && cd build && ../configure && make && make check

To save some space I remove a few libraries. Maybe I can remove more?, but the developers might know it better ;)

RUN apt-get remove -y build-essential \
  pkg-config \
  automake dejagnu flex bison \
  libtool texinfo autopoint help2man \
RUN rm -rf /var/lib/apt/lists/*

Finally, I define the entrypoint:

ENTRYPOINT [ "/poke/build/run", "poke" ]

This call, which is also described in the HACKING file, starts poke. Since it is defined as entrypoint, I can also specify parameters with the call.


First, I build the image:

docker build -t gnupoke .


… and this is how I run the image:

docker run -it gnupoke

But I can also change the entrypoint to call a bash shell. In this example I also mount the current host directory into the /data directory of the image. This allows me to access data from the host.

docker run -it -v $PWD:/data --entrypoint /bin/bash gnupoke

I this case I need to start poke inside of the image with: /poke/build/run poke


As mentioned above I wonder why not all necessary libraries are described in the documentation. Of course, this depends on the distribution and version and can be complex. But why is not a Dockerfile created? On one hand, the Dockerfile documents the necessary steps and at the same time it setup a defined environment. Nobody needs to say: but it runs on my machine.

And if you want to try gnu-poke by your own you can take a look at the manual.