The problem: i make heavy use of s2sh in CGI scripts on my websites, but building the binaries directly on my web hoster risks using too much CPU time and causing undesired Grief with the administrators. Building purely static binaries on Linux is not as straightforward as it really could be because glibc (the most widely-used standard C library on Linux systems) cannot "really" statically link certain networking libraries.

One solution is to build, on my local machine, purely static s2sh binaries in a Docker environment using a build chain which does not have that quirk/limitation. It's still important that the build architecture is the same as the hoster's, but it's not important that we build with the same libc. So...

(Disclaimer: Docker is still largely alien to me, so the details described here might not reflect optimal, or even necessarily correct, use of Docker.)

The following sequence of shell commands, run from this directory (or a directory containing this one's Dockerfile), ought to build a static binary of s2sh:

docker build -f Dockerfile-cwal-static -t s2sh_static  . ;
docker create --name s2sh s2sh_static ;
docker cp s2sh:/cwal-src/s2/s2sh s2sh ;
strip s2sh ;
# OPTIONALLY copy the s2sh wrapper script:
#   docker cp s2sh:/cwal-src/s2/s2sh.sh s2sh.sh ;
docker container rm s2sh ;

The result is a fully static s2sh binary which includes various modules which are commonly required in CGI scripts ("commonly" meaning "on my systems").

About the optional s2sh wrapper script: s2sh benefits from having several environment variables configured before it starts, so the common practice ("common" meaning "on my systems") is to move the compiled s2sh binary to ~/s2/s2sh.bin, move s2sh.sh to ~/bin/s2sh, make it executable, and edit it to suit (noting that ~/bin is in my $PATH and ~/s2 (a.k.a. $S2_HOME) is where my numerous s2 bits are stored). See ../s2/manual/s2-home.md for more details.

Cleanup:

docker image rm s2sh_static

Aaaaarrrgggg: sometimes that removes only the latest image in the stack and sometimes it removes everything. What's up with that?

For various reasons beyond my meager docker ken, it's common to see leftover images which may or may not be useful later on. To remove all of the intermediary images, use docker image ls -a to get a list of all images and remove them by using their IDs:

docker image rm ID_1... ID_N

Certainly there's an easier way without resorting to awk? How about:

docker system prune --force

Optionally, remove the base OS image:

docker image rm alpine:edge

Because the s2sh_static image has to install so many OS packages, removing the small core image (alpine) doesn't save all that much time unless the internet connection is slow.