is “consider standardising on a Docker container for building” a reasonable thing to ask of a person publishing an Open Source project which is by definition cross-compiled?
a properly made Dockerfile creates reproducible builds, a lone Makefile and a list of un-versioned dependencies by name in a readme uh, doesn’t
a properly made Dockerfile creates reproducible builds, a lone Makefile and a list of un-versioned dependencies by name in a readme uh, doesn’t
a properly made Dockerfile creates reproducible builds, a lone Makefile and a list of un-versioned dependencies by name in a readme uh, doesn’t
@cat "properly made" is the important part here -- very, very many Dockerfiles are equivalent to a shell script that downloads and installs god-knows-what with no explicit versioning/verification constraints, and I'd argue that isn't better for reproducibility (although it is faster and less frustrating, since you can just make the machine do it instead of a human)
@cat I think the usability argument here is strong *if* people who are likely to contribute are likely to either already have Docker or be using a platform where it's reasonably lightweight to start using. I work in some projects where the user population features cranky BSD users heavily and Docker is resisted strongly there. (I used to work for Docker and currently don't have it installed anywhere, and have a lot of thoughts on this after many discussions; sorry if I'm replying too much!)
I feel like “hey, if you set up your cross compiler in a Dockerfile, it will mean that anyone can pull that down and contribute, no matter what OS and toolchains they have on their host system already” is reasonable, but?