include some kind of gcc/makefile script

[eigood]

Your README doesn't mention how to compile noop.c. Please provide instructions on how to build this image.

Fyi, directly including the compiled version of noop is very bad. Is it 32 bit? 64 bit? What arch?

Ideally, there would be multiple phases to this repo. Phase 1, that extends some OS, installs gcc, and compiles noop. Phase 2, not using compose, would volume mount $PWD, and copy out the compiled binary. Phase 3 would then build the final resulting image with just noop.

[JanNash]

I'd greatly appreciate your help on this 👍

[eigood]

I thought you might say that. I'll see what I can do.

We've used this technique to install groovy+gradle, run fatjar, then just save the resultant output. Same for doing a git clone of nginx, adding/configuring modules, then saving the nginx binary. Keeps the containers clean, and doesn't require modifying the host environment with random $dev stuff.

[JanNash]

To avoid misunderstandings, I've of course compiled the file inside a docker-container. I'm on macOS here, so it wouldn't have worked if I compiled it locally. Also, as you mentioned, I don't want to clutter my local machine with random $dev stuff :)

I think that it could also be possible to use library/gcc to compile with a docker run --rm command and to include that command in the readme, so people can reproduce the compilation if they want to, but still to supply the compiled version (which should be compiled with that exact command, of course). I guess the byte-shaving doesn't have to go as crazy as in http://www.muppetlabs.com/~breadbox/software/tiny/teensy.html, which you mentioned in the other issue, but to get down to a few KB would definitely be awesome :)

[JanNash]

Also, there could be compiled versions for multiple different architectures included and then the image could just be tagged with the different architectures