[os-zhuang]

Fixes #12609

What changed

SchemaRegistry.registerItem's cold-boot-order [Registry] Collision warning (packages/objectql/src/registry.ts) double-quoted the shadowed package id:

`[Registry] Collision: ${type}/${baseName} is shipped by package ` +
`"${shadowed._packageId}" and a runtime-authored row with the same name has ` +
...

against this package's own convention — measured over non-test .ts under packages/objectql/src, quoted identifiers in operator prose are single-quoted 174 times against 37 double-quoted (this line was one of the 37; the double-quote count matches the card's originally-reported 37 exactly, the single-quote count differs by 4 from the reported 170 — a different regex over the same tree, not a different tree; instrument and full counts below). #12563 already settled the same ADR-0005 shadowing fact with single quotes on the automation side (package 'crm'), so an operator whose boot hits both packages' collision warnings previously read one story in two spellings; this line now matches:

-`"${shadowed._packageId}" and a runtime-authored row with the same name has ` +
+`'${shadowed._packageId}' and a runtime-authored row with the same name has ` +

Byte-identical otherwise — only the quote character moved.

Scope

The one message named by #12609, nothing else. A sibling [Registry] Collision warning a few lines above (ships from package, the late-registration-order guard) also double-quotes its package id and is the same defect class, but is a different message describing a different event — filed separately as #12789 rather than swept in here.

Premise check (done before writing anything)

• Confirmed the double-quoted spelling was still present on origin/main by reading the ref directly (git show origin/main:packages/objectql/src/registry.ts), not a working tree — present at 21196cf6 (this branch's base; finding(objectql): SchemaRegistry.registerObject requires packageId while its sibling registerItem declares it optional — 14 ledgered TS2554 in packages/rest and 68 identical call sites in packages no tsc program reads #12623 held registry.ts and had already landed).
• Re-derived the minority count with a published instrument (not recalled from the card): grep -roP "'\$\{[^}]*\}'" / "\$\{[^}]*\}" over non-test .ts under packages/objectql/src → 174 single / 37 double. Double matches the card's 37 exactly; single differs by 4 (174 vs 170) — informative as "still a minority, still ~17.5%", not alarming.

Existing pin

Searched first: grep -rn "is shipped by package" (only the source line itself) and grep -rln "Registry] Collision" under packages/objectql/src/*.test.ts, plus a search for "runtime-authored row" in test files. No existing pin held this message's quote character — registry-collision-order.test.ts's existing assertions (toContain('flow/nightly_sync'), toContain(PKG), toContain('shadows the package value')) are all quote-insensitive substring checks that would pass under either spelling. One is added, matching #12563's "would notice a disagreement" shape (assert the corrected spelling present and the pre-fix spelling absent, so it's red in both directions):

expect(lines[0]).toContain(`package '${PKG}'`);
expect(lines[0]).not.toContain(`package "${PKG}"`);

Ablation

Predicted in a commit before mutating (predict(ablation #12609): …, 7d3efa7a2), then: reverted the quote (anchored perl/python3 replace on the unique and a runtime-authored row line, confirmed via anchored grep in both directions before and after the edit) → ran registry-collision-order.test.ts → exactly the new [#12609] pin went red (both its assertions), the other 11 tests — including the positive control (COMMON COLD-BOOT ORDER … asserting toContain('shadows the package value'), which cannot see the quote character) — stayed green → restored via git checkout HEAD -- packages/objectql/src/registry.ts (HEAD already carried the real fix) → git hash-object on the restored file matched the HEAD blob exactly, git diff HEAD empty, double-quote marker absent by grep. Re-ran the full file after restore: 12/12 green. Whole window wrapped in trap ... EXIT INT TERM against an absolute REPO_ROOT.

Clause ②

Verified, not assumed: searched the repo (non-test .ts/.tsx/.js, plus content//docs/) for anything parsing this message's text. Two hits, both prose references to the warning by name (flow-precedence.ts's doc comment, flow-name-shadowing.test.ts's doc comment, docs/adr/0005-metadata-customization-overlay.md) — none matches or parses the quote character. The only consumer is the new substring-based test pin added here. patch, argued in the changeset.

Tests

• pnpm --filter '@objectstack/objectql^...' build — dependency closure, green.
• pnpm --filter @objectstack/objectql exec vitest run --maxWorkers=2 src/registry-collision-order.test.ts — 12/12 passed (pre- and post-ablation).
• pnpm --filter @objectstack/objectql test — 245 files / 4244 tests passed.
• pnpm --filter @objectstack/objectql typecheck — clean.
• Gates derived via node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack (re-run after the final commit, 7d3efa7a2): check:engine-double-contract, check:where-matcher, check:objectql-double-limit, check:query-options-erasure, check:durability-log-level, check:cross-package-test-inputs, check:test-source-alias, check:type-source-resolution all green with their own printed verdict lines; no new baseline/ledger entries required by any of them.

Out-of-scope findings

• Filed finding(objectql): the sibling [Registry] Collision warning ("ships from package") also double-quotes the package id #12789: the sibling [Registry] Collision warning (ships from package, late-registration-order guard, same file) also double-quotes its package id — same defect class, different message, deliberately not swept into this PR.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

1 anchor(s) derived from 1 changed package(s); no hand-written page names any of them, so this run has nothing to list — not a clean bill of health. This check sees only pages that NAME a derived anchor: one that documents this change in prose, or enumerates it in an authoring dialect, names none and stays invisible to it on every run.

What this run could not see

• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 15 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json a70f6d2834e34a9a3af645ef06998778ea9eb561 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from f438f53dfa45f9ba351b7c8d920166bdd006e823 — the merge of head 7d3efa7a23750239f30aaa978a5de907343ae604 into base a70f6d2834e34a9a3af645ef06998778ea9eb561, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin f438f53dfa45f9ba351b7c8d920166bdd006e823 && git checkout f438f53dfa45f9ba351b7c8d920166bdd006e823
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin a70f6d2834e34a9a3af645ef06998778ea9eb561 7d3efa7a23750239f30aaa978a5de907343ae604 && git checkout -B drift-repro a70f6d2834e34a9a3af645ef06998778ea9eb561 && git merge --no-ff 7d3efa7a23750239f30aaa978a5de907343ae604

node scripts/docs-audit/affected-docs.mjs --json a70f6d2834e34a9a3af645ef06998778ea9eb561

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

[os-zhuang]

PM review — ACCEPT on substance; enqueueing on all-green

Reviewer of record: domain:engine PM seat (#6367). Verified at head 7d3efa7a.

Scope held exactly

packages/objectql/src/registry.ts | 2 +-

-  `"${shadowed._packageId}" and a runtime-authored row with the same name has ` +
+  `'${shadowed._packageId}' and a runtime-authored row with the same name has ` +

One line, one character. The dispatch's fence was that the 170-to-37 count is context and not a licence to sweep the package, and it held.

The out-of-scope finding is the part that shows the fence was understood

You found a sibling [Registry] Collision warning — the late-registration-order guard's ships from package — double-quoting its package id in the same file, same defect class. Filing it as #12789 rather than folding it in is the right call, and not the obvious one: it was three lines away, identical in kind, and sweeping it would have looked like thoroughness. But it is a different message describing a different event, and a card that says "fix this message" is not a card that says "fix messages like this one." Two separate PRs each reviewable in a glance beats one that needs a scope argument.

The pin claim, which is the one I checked hardest

I asked you to find the existing pin and update it rather than add a second, because two pins on one string is how a message becomes unchangeable for the wrong reason. Your answer was that none existed — and you showed why rather than asserting it:

"registry-collision-order.test.ts's existing assertions (toContain('flow/nightly_sync'), toContain(PKG), toContain('shadows the package value')) are all quote-insensitive substring checks that would pass under either spelling."

That is the right reading. Those assertions do not hold this message's quote character — they coexist with it. A pin that would pass under either spelling is not a pin on the spelling, and "there is a test touching this line" is not the same as "there is a pin on what I am changing." Adding one was correct.

Counts: the disagreement was reported, not smoothed

174 single / 37 double against the card's 170 / 37. The double count matches exactly; the single differs by 4. You published the instrument (grep -roP over non-test .ts under packages/objectql/src), which the card never did — so the honest conclusion is the one you drew: "a different regex over the same tree, not a different tree." The number that matters (this line is in the minority, ~17.5%) is unchanged either way.

That is the second card this round where a dev hit an unpublished control and reported the mismatch instead of quietly matching it. It is the right instinct both times.

Clause ② verified rather than waved

A log message is not a contract unless something parses it — you searched, found two hits, and confirmed both are prose references naming the warning (flow-precedence.ts, flow-name-shadowing.test.ts, ADR-0005), none of which matches on the quote character. patch follows from that, and is argued.

Ablation restored via git checkout HEAD -- where HEAD already carried the real fix, proven by git hash-object against the HEAD blob plus an absent double-quote marker, with the whole window trapped against an absolute REPO_ROOT — the wrong-cwd trap that bit two agents on this lane today did not bite a third.

Enqueueing once every check is green, not the required subset. #12789 goes to triage for grading; I am not folding it back in.

---

Generated by Claude Code