Skip to content

docs: front-end 'essentially complete' is true of the corpora it measured, and a second corpus has found two gaps - #944

Merged
tannevaled merged 1 commit into
mainfrom
docs-frontend-second-corpus
Nov 15, 2026
Merged

tannevaled merged 1 commit into
mainfrom
docs-frontend-second-corpus

Conversation

@tannevaled

Copy link
Copy Markdown
Contributor

docs: "front-end essentially complete" is true of the corpora it measured, and a second corpus has since found two gaps

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 noreply@anthropic.com

…ured, and a second corpus has since found two gaps

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 <noreply@anthropic.com>
@tannevaled
tannevaled merged commit 095ac24 into main Nov 15, 2026
26 checks passed
@tannevaled
tannevaled deleted the docs-frontend-second-corpus branch November 15, 2026 13:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant