fix(extraction): reduce A::B.new callee to the constant

[benrfairless]

What does this PR do?

Slice 2 of 5 splitting #1706. Stands alone; depends on nothing else in the series.

A Ruby Const.new call is rewritten to name the constant rather than the method, so the call links the class the way every other language's new T() does — the constructor body lives in initialize, so naming new links nothing.

That rewrite only fired for bare constant receivers. A scope-resolved receiver fell through it and kept the whole call expression as its callee: Admin::User.new. That resolves to nothing, and it cannot join the Ruby LSP's constructor rows, whose callee is the class QN.

Correction to the original description. While writing the reproduce-first test I found the claim I made in #1706 — that the callee stayed "new" — is wrong. The actual pre-fix callee is the full Admin::User.new expression. The commit message and test here state the real behaviour; the fix itself is unchanged and still needed.

Reproduce-first. Without the change, the new row prints exactly that:

ruby callees (2):
  'Admin::User.new'
  'Widget'

The test pins both spellings — scope-resolved reduces to Admin::User, bare stays Widget as a regression guard — and rejects any callee that is new or retains a .new suffix. There was previously no test anywhere for this rewrite in either form.

Where this sits in the series

#	PR	Depends on
1	#1709	—
2	this PR	—
3	feat/ruby-lsp-per-file	—
4	feat/ruby-lsp-cross-file	3
5	docs/ruby-hybrid-lsp	4

CI baseline

main at 4d7d9f1 is red for reasons unrelated to this change: #1181 gave list_projects pagination parameters, retiring the last empty-properties schema that #1359's guards used as their zero-argument example. #1704 already fixes it. Until that merges, ci-ok will be red here for that reason.

Refs #1701

Checklist

• Every commit is signed off (git commit -s)
• Tests pass locally (scripts/test.sh) — see fix(extraction): keep Ruby superclass names bare #1709 for the full-leg numbers; all failures are pre-existing on unmodified origin/main.
• Lint passes (make -f Makefile.cbm lint-ci)
• New behavior is covered by a test (reproduce-first for bug fixes)

---

Assisted by OpenCode using anthropic/claude-opus-5.

[github-actions]

Thanks for opening this — it has been seen, and it is queued.

This note is automated, but it is not a brush-off: it exists so you know where your PR stands instead of having to guess from silence.

Current review status: working through a backlog. 0.9.1-rc.1 is out, so the release freeze that held reviews is over — but it left a large queue of open pull requests behind it, and we are reading through them oldest-first. The background is in discussion #1144.

What that means for this PR, concretely:

• It will not be closed for inactivity. No stale bot touches pull requests here.
• It may still sit a while before a human reads it. That is on us, not on you.
• Older PRs are read first, so a recent one is not being skipped — it is behind a queue.

Things that will genuinely speed it up whenever review does happen:

• Keep it rebased on main — the tree is moving quickly right now, and a conflicting branch cannot be reviewed as the diff you intended.
• Get CI green, or say which failures you believe are pre-existing.
• Keep the change to one claim. Bundled features and refactors get split before they get merged, which costs you a round trip.
• Every commit needs a sign-off (git commit -s) — CI enforces DCO.

If this fixes a bug, a reproduction we can run is worth more than a description of the symptom.

Thanks for contributing, and sorry in advance for the wait.

[DeusData]

The Ruby fix is right. There is one thing in the diff that does not belong to it.

The fix itself: two lines, gated on CBM_LANG_RUBY, adding scope_resolution beside constant — the same shape as #1709, and the comment explains what resolution actually joins on (the bare leaf) rather than just what changed. The test pinning both spellings, with the bare Widget case kept as a regression guard and an explicit rejection of any callee that is new or retains a .new suffix, is the right shape. There was no test for this rewrite in either form before, so this is a coverage gain as well as a fix.

And the correction deserves saying out loud. You went back and contradicted your own #1706 description — the pre-fix callee is the whole Admin::User.new expression, not "new" — because the reproduce-first test told you so. Most people quietly ship the corrected test under the old story.

The thing that does not belong

tests/test_daemon_runtime.c removes an assertion:

-    cbm_daemon_runtime_service_active_connections(fixture.service) == 2 &&

That is not mentioned anywhere in the PR description, and it has nothing to do with Ruby constructor callees. I also checked the other four slices — none of them touches that file — so it is specific to this one rather than something shared across the series.

I am not disputing the change on its merits. I checked the precedent you cite: db91b88c is real, it is "fix(test): stop asserting a scheduling race in the forged-identity test", and it is the same fault class. Asserting active_connections == 2 in the window between a raw connect and the server's HELLO timeout is asserting a transient state, which is exactly what we do not do here. Your reasoning looks sound and your replacement argument — that the "capacity" substring proves the reason and active_clients == 1 is safe because the test still holds accepted — is the right one to make.

But it needs to be its own PR. A change that removes test coverage gets read on its own terms, by someone thinking about that test rather than about Ruby extraction. Bundled into a language fix it is very easy to wave through, which is precisely the risk. Please lift it out — the commit message you would write for it is already most of the way there in that comment.

I suspect this is a stray from local iteration rather than deliberate bundling, given how carefully the rest of this series is scoped. Worth saying, though, because your own note when closing #1706 was that bundled changes are "exactly the kind of thing an umbrella PR hides" — and this one hid in a slice.

Status

ci-ok should no longer be red for the reason in your baseline note — the #1181 × #1359 cli failures are fixed on main; I confirmed both tests pass in a full-suite run today.

Cleared and approved on the Ruby change. Split the daemon hunk out and this merges.