Google Summer of Code/2022/Ideas/Language eclass modernisation

Language eclass modernisation

Gentoo's eclasses for language support (or "bindings") have gone through several evolutions until the community largely settled on the "Python model" (now used for Lua and others).

Ruby and Java haven't yet been adapted to this newer dependency model. This often leads to counterintuitive conflicts for users and confusing errors. It's not certain that the dependencies specified are completely accurate as-is or can be expressed as necessary.

This project will lead to a significantly improved user experience and fix a whole slew of bugs in the process!

We also have a smaller-scope opportunity to work on the OCaml ecosystem.

Options:

Convert either (or both) the Java and Ruby eclasses to this modern style/model. This will make developing ebuilds with Java (or Ruby) support far easier and make dependencies more correct for users, reducing the pain of upgrades. Candidates may work on either (or both if they're feel adventurous) the Java or Ruby ecosystems. Both are in need of similar treatment:
1. In a new revision to use the modern Python-style JAVA_COMPAT and RUBY_COMPAT configurations;
2. Introduce appropriate functions for use in ebuilds (again based on the Python model and, nowadays, Lua);
3. Document these frameworks and their usage;
4. Update ebuilds to make use of them!
OCaml
1. Unify the OCaml eclasses if possible into one eclass with multiple options/variables;
2. Add a mechanism into OCaml eclasses to indicate if they support ocamlopt;
3. Audit our existing OCaml packages to properly support -ocamlopt and ocamlopt setups;
4. Document these frameworks and their usage;
5. Update ebuilds to make use of them!

For more on Ruby, see:

Contacts	Required Skills
Sam James Yuan Liao (Leo3418)	Bash (sufficient for ebuilds and eclasses) Written own Gentoo ebuilds and familiar with Gentoo eclasses Possibly some Java (or Ruby or OCaml), but mostly familiarity with Java (or Ruby or OCaml) build systems. Familiarity with Gentoo is a requirement
Expected Project Size	Expected Outcomes
175 hours or 350 hours, depending on what tasks are planned. This project could easily go as far as 350 hours depending on discussions with the candidate: it will take longer if doing both Java and Ruby, and it'll take longer if converting either a significant number of ebuilds in tree, or all ebuilds in tree. 170 hours if doing the OCaml option instead.	If Java is chosen: A set of eclasses for Java: java-any-rN (build-time only dependency), java-single-rN (single Java version), maybe java-rN (multiple implementations installed by a package if appropriate), and possibly java-utils-rN. If Ruby is chosen: A set of eclasses for Ruby: ruby-any-RN (build-time only dependency), ruby-single-rN (single Ruby version), ruby-rN (multiple Ruby implementations installed by a package), and possibly ruby-utils-rN. For the language(s) chosen, sufficient 'eclassdocs' (eclass documentation) should be written, as well as e.g. a wiki page describing typical usage. A stub version of the 'Python guide' (see stretch goal) would be a bonus. Stretch goal: A similar resource to the Python guide for Java and/or Ruby (or OCaml)
Project Difficulty
Hard (relatively). It should be totally manageable for somebody familiar with ebuilds (and eclasses if possible), but if you are new to working with ebuilds, another task may be more suitable. Ruby is likely to be easier than Java because Java has mixed build systems (e.g. Maven). The task will be easier if the candidate has familiarity with the language & ecosystem they choose. The OCaml option is easier but has less impact for our users; we have a number of OCaml packages (mostly as dependencies for useful applications) but Java and Ruby are far larger ecosystems, and hence have more impact.