The new version of Raku++, v5.0.0, has been released and it passes 100.00% of Roast.
Roast is a test suite for Raku conformance: if an implementation passes it, it can call itself a Raku implementation. So Raku++ is formally an official Raku implementation now.
I added .00% to the number on purpose, as a rounded 100% mark could be achieved without the last couple of dozen tests passing.
It took some time to close the last gap. The first sharp rise came with implementing Unicode 17 support, which unlocked about 92k suite tests. Then, for a couple of months, Roast coverage stalled at about 90-92%. Reaching the last 8% was a much slower process, but here we are.
The raw numbers now are:
220054 passed of 220055 declared tests.
218421 passed of 218422, which are listed in the so-called spectest list. It includes the majority of Roast with some exceptions.
The only failing test is where the test itself needs updating, as it does not list the most recent macOS version.
Interestingly, when I started, I didn’t think that reaching a raw hundred per cent was possible. There’s even a—now historical—document in the repository saying so.
Future plans? The previous v4.0.0 version was tagged after Raku++ could test and run 1000 modules from the Raku ecosystem (without modifying them). Now, with v5.0.0, full Roast conformance is unlocked ✅. The future plans include further performance improvements and wider ecosystem support. Ideas for what the next version should cover are collected in the V6-PLAN document. An attentive reader may notice the magic version number here.
This is great news!
Please note that all RakuAST specific tests that currently live in the Rakudo repository, will be moved to roast when 6.e is released.
But I have no doubt Raku++ will be able to consume those as well, as all of the RakuAST internals are written in NQP and Raku.
Note that the tests fudging for MacOS have been generalized, and should now pass on even the newest version of MacOS!
Indeed, it works now!
BRILLIANT!
Exciting news! Can’t wait to try this out for my Raku projects.
Quick question. What is this project’s policy for LLM contributions? Asking because the README seems AI-generated, and I couldn’t find a policy for this in the repo. Apologies if this is already answered elsewhere and I missed it. Thanks!
There’s no explicit policy published. We have a number of issues typed with AI and submitted by a human. So far there’s nothing to worry about. I doubt that any pull requests to the engine itself will be accepted though.