Build static D binaries with docker

This repo provides dockerfiles to create docker images used for creating fully-static D program binaries. It serves as an example of how to compile static Dlang programs using alpine linux, musl-c, and the ldc compiler. It also serves as an example of how to configure github actions to build a docker image, push it to Dockerhub, and create a release draft that has the static binary attached from the uploaded docker image. The created binary should work on any x86_64 linux distribution.

The dlang/ldc.dockerfile provides a base environment that can be used to create static dlang binaries. An example D program can be found here: tests/test-app.

Shared Libraries

If your program requires other shared libraries you will have to manipulate the dub.json file further. The dub.json excerpt below example is taken from another D program, fade:

    "configurations":[
		{
			"name":"shared",
			"targetType": "executable"
		},
		{
			"name":"static-alpine",
			"targetType": "executable",
			"dflags-ldc": [
				"-link-defaultlib-shared=false",
				"-static",
				"--linker=gold",
				"-L-lz",
				"-L-lbz2",
				"-L-ldeflate",
				"-L-llzma",
				"-L-lcurl", 
				"-L-lssl", 
				"-L-lssh2", 
				"-L-lcrypto"
			],
			"sourceFiles": ["/usr/local/lib/mimalloc-2.0/mimalloc.o"]
		}
	]

The dflags must be appended to include "-L-lcurl" lines for neccessary libraries, and these libraries will need to be present in the docker image in both shared and static library forms. For example, many libraries in the htslib/Dockerfile image had to be compiled from source:

# compiling libdeflate
RUN wget https://github.com/ebiggers/libdeflate/archive/refs/tags/v$DEFLATE_VER.tar.gz \
    && tar -xf v$DEFLATE_VER.tar.gz \
    && cd libdeflate-$DEFLATE_VER \
    && make install -j 8 \
    && cd /home/ \
    && rm -r libdeflate-$DEFLATE_VER

Though if the alpine repositories have it in both static and shared forms you may be able to avoid compiling them:

RUN apk add --update --no-cache \
    autoconf automake make cmake \
    zlib zlib-dev zlib-static \
    xz xz-dev \
    bzip2 bzip2-dev bzip2-static 

These libraries also must be installed into locations on the LD_LIBRARY_PATH and LIBRARY_PATH or those environment variables must be altered. These are defined in our images as such:

ENV LD_LIBRARY_PATH="/lib:/usr/local/lib:/usr/local/lib64:/usr/lib:$LD_LIBRARY_PATH"
ENV LIBRARY_PATH="$LD_LIBRARY_PATH"

Mimalloc

Since many of our programs that rely on htslib are multi-threaded, we have also included this line:

"sourceFiles": ["/usr/local/lib/mimalloc-2.0/mimalloc.o"]

musl-c's allocation performance is poor compared to mimalloc in multi-threaded situations: https://www.linkedin.com/pulse/testing-alternative-c-memory-allocators-pt-2-musl-mystery-gomes. Our htslib image includes a compiled version of mimalloc, including the prior line in our dub.json ensures that mimalloc is used for all allocation across our binary, even for libraries that weren't compiled with mimalloc.

Docker and Github Actions

If your docker image is uploaded to Dockerhub, you can use the github actions .github/workflows/build_and_release.yml in this repo to have your docker image built and uploaded to Dockerhub via github actions. You will have to set up a dockerhub API key. The github actions script will also pull the test app docker image from Dockerhub, extract the static binary from it, and create a release draft with the binary attached, just look at our v0.0.1 release.

The docker images built by our github actions are availiable through dockerhub:

docker pull blachlylab/dlang-ldc-static
docker pull blachlylab/dlang-htslib-static

or use with the FROM docker syntax.