What is shit?

Shit is a package manager. Or maybe not. No, actually it's not. It's just a collection of recipes which describe how to build packages. It does not include init scripts nor additional patches. Software which doesn't build with a simple configure && make install (or something comparable to that) is not allowed here. Similarly, if the build system does not respect CFLAGS or prefix or DESTDIR then it's crap. And there is no crap in shit.

Rationale

Sooner or later you are bound to compile one or the other piece of software from source. Either because the package simply doesn't exist for your platform or because the packagers are too drunk to update the package. In either case, the first thing you have to do is to find the download link. And oh boy is it fun to click through all the ads and redirects to find the actual file. Then you have to go back and find the instructions how to build it. Yes, we all hate autocrap, but it actually works (if done properly). At least it fucking respects prefix and DESTDIR. Not unlike cmake which just had to make up their own name for the prefix. Once you have the package installed you have to configure it anyway, and you probably won't use init scripts anyway but god or SMF or whatever the tool of the day is.

Recipes always describe the latest stable release of each package. If a dependent package breaks then so be it. The world is moving on, and so should you. Read that carefully: you should read the release notes of each package you're installing and decide for yourself whether you want to update or not. I will not make decisions about which package is deemed stable and which one is not. I trust upstream to do that decision.

Installing shit

Clone to a directory, then add <prefix>/bin to your $PATH and you can start using shit like this:

$ shit install llvm

The full list of commands is: help, version, build, link, unlink, list, install. The install command is just a convenient wrapper around build and link'

How it works

Shit is heavily inspired by mxcl's Homebrew, but still very different. It uses a DSL for the recipes, doesn't include any patches or additional scripts, strives to be cross-platform (mac, linux or anything that includes a C compiler). It does not manage dependencies, it's up to you what you install with the native package manager and what through shit.

The recipes are stored in <prefix>/share/shit. Just look at any of the existing recipes, copy it and modify it. Shit follows convention over configuration. A tarball with name foo-1.0.tar.gz is expected to extract the source into the directory foo-1.0. The unpack directive will unpack the tarball and set pwd to the directory with the source. The configure directive automatically sets the prefix, make parallelizes the jobs.