From a5c5c0b823e32c7a7ef60d804d38e304eed769e3 Mon Sep 17 00:00:00 2001 From: tannevaled Date: Sun, 15 Nov 2026 12:27:05 +0100 Subject: [PATCH] docs: "front-end essentially complete" is true of the corpora it measured, and a second corpus has since found two gaps MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit CONFORMANCE-LIBRARIES.md closes with "Front-end (parse + compile) is essentially complete", and sends the reader on to the runtime surface. That report is dated 2026-06-27 and its measurement stands: it is about the lib/ trees of ten released gems plus Rails and Puppet, where rbgo reaches 99.82% / 100% with 0 over-permissive. What it cannot see is a construct those codebases do not write. ruby/spec does write them, and two parser defects are open against it: #669 a do...end block inside an arrow-lambda body, after a paren-less command call -- 160 ruby/spec examples, the largest single-defect figure in the tracker #905 a brace block on a name that is also a local, inside parentheses Neither contradicts the June figures. #669 is a NESTING of the paren-less-command-plus-do...end form that the report's own "What moved the needle" records as fixed, so the fix and the defect are different shapes of one construct. ⚠ And one thing stated as unchecked rather than assumed: whether any of the 6 remaining Rails gaps IS that construct has not been cross-referenced. A first draft of this change asserted that no file in Rails, Puppet or the ten gems nests it that way -- which is plausible (such a file would have shown up as a gap) and which I had not measured. It is now written as the open question it is, because "not checked" and "known absent" are not the same claim and the difference is half an hour of work. The lesson the paragraph carries: "essentially complete" was true of what was asked, and a corpus answers only the questions it contains. Refs #669, #905, #941 Co-Authored-By: Claude Opus 5 --- CONFORMANCE-LIBRARIES.md | 36 ++++++++++++++++++++++++++++++------ 1 file changed, 30 insertions(+), 6 deletions(-) diff --git a/CONFORMANCE-LIBRARIES.md b/CONFORMANCE-LIBRARIES.md index 03a378cb..440dd25e 100644 --- a/CONFORMANCE-LIBRARIES.md +++ b/CONFORMANCE-LIBRARIES.md @@ -71,9 +71,33 @@ and **RSpec DSL usage is 10/10** byte-identical to MRI. ## What is left -Front-end (parse + compile) is essentially complete. The remaining tail is a small -number of compile-time constructs (the 6 remaining Rails gaps) and, beyond the -front-end, the **runtime** surface — the rest of the stdlib, C-extension -equivalents and predefined constants — which is what stands between "parses + compiles" -and "runs the application." That is ongoing, unproven work tracked in the sibling -CONFORMANCE-*.md reports. +Front-end (parse + compile) is essentially complete **against the corpora above**. +The remaining tail is a small number of compile-time constructs (the 6 remaining +Rails gaps) and, beyond the front-end, the **runtime** surface — the rest of the +stdlib, C-extension equivalents and predefined constants — which is what stands +between "parses + compiles" and "runs the application." That is ongoing, unproven +work tracked in the sibling CONFORMANCE-*.md reports. + +⛔ **A second corpus has since found front-end gaps these could not.** Everything +above is measured against the `lib/` trees of released gems and of Rails and +Puppet. ruby/spec exercises constructs that production libraries simply do not +write, and two parser defects are open against it as of 2026-11-15: + +| issue | construct | cost | +| --- | --- | --- | +| [#669](https://github.com/go-embedded-ruby/ruby/issues/669) | a `do…end` block inside an arrow-lambda body, after a paren-less command call | **160 ruby/spec examples** — the largest single-defect figure in the tracker | +| [#905](https://github.com/go-embedded-ruby/ruby/issues/905) | a brace block on a name that is also a local, inside parentheses | open | + +Neither contradicts the measurement above. #669 is a *nesting* of the +paren-less-command-plus-`do…end` form that **What moved the needle** records as +fixed, so the fix and the defect are about different shapes of the same +construct. + +⚠ Whether any of the **6 remaining Rails gaps** is in fact that construct has +**not** been checked — it is a cross-reference nobody has made, not a thing +known to be absent, and it would be worth half an hour before anyone treats +#669 as a ruby/spec-only concern. + +The point worth carrying: *"essentially complete"* was true of what was asked, +and a corpus answers only the questions it contains. The parity plan is +[#941](https://github.com/go-embedded-ruby/ruby/issues/941).