Landing-order break: PR #6510's dev cell pins the RAW warn total at 2, which includes the #6505 false positive the #6505 fix removes

[os-support-ai]

Measured while implementing #6505 (branch claude/issue-6505-predicate-valued-gate-rules). Filed unassigned rather than folded in: the file is in an unmerged PR, so it does not exist on main and cannot be edited from the #6505 branch.

The collision

PR #6510 (Fixes #6445, head cbb77c304, in the merge queue) adds
packages/react/src/__tests__/SchemaRenderer.disabledGateFaultDiagnostic.test.tsx. Its group-1 development cell reads:

mount([{ id: 'n1', disabled: FAULT_BARE_DEV }]);
...
// subtracted by NAME rather than by count
expect(nonValidatorWarnings(warn)).toHaveLength(1);
expect(allWarnings(warn)).toHaveLength(2); // <- line 372

The 2 is our line + the #6505 false positive. FAULT_BARE_DEV is the bare expression 'devonlybadroot.locked == true', the mount helper always supplies type, and id is a string — so disabled must be a boolean is the only thing validateSchema reports for that node.

Once #6505 lands, that node validates, the second warning is never emitted, and allWarnings(warn) is 1. Line 372 goes red.

The file's own design is not at fault and is arguably what caught this: everywhere else it subtracts the noise by name (nonValidatorWarnings), which is invariant under the #6505 fix. Line 372 is the one raw-total pin in a dev cell, and its docblock says why the raw total was worth pinning at all. The other two dev cells (nonValidatorWarnings(...).toHaveLength(0) at ~537, and the dev/prod parity cell at ~643) are unaffected. Production cells are unaffected — the validator is a no-op there, which the file already states.

Scope

Exactly one assertion, in one file:

file	line	now	after #6505
packages/react/src/__tests__/SchemaRenderer.disabledGateFaultDiagnostic.test.tsx	~372	expect(allWarnings(warn)).toHaveLength(2)	1

Why this needs its own card rather than a rider

Neither PR can fix it alone:

• The Dev-mode validateSchema declares disabled / visible "must be a boolean", so every expression-valued gate is reported invalid — the console line sits right beside the real diagnostic #6505 branch cannot edit a file that is not on main.
• fix(react): report a faulting disabled / disabledOn predicate instead of silently greying the control out #6510 cannot pre-adjust to a fix that has not merged without going red now.

So whichever merges second turns the other's suite red on main, and the fix is a one-line edit whose correct value depends on which order that was. Landing order:

• fix(react): report a faulting disabled / disabledOn predicate instead of silently greying the control out #6510 first, then Dev-mode validateSchema declares disabled / visible "must be a boolean", so every expression-valued gate is reported invalid — the console line sits right beside the real diagnostic #6505 — the Dev-mode validateSchema declares disabled / visible "must be a boolean", so every expression-valued gate is reported invalid — the console line sits right beside the real diagnostic #6505 PR must carry the one-line edit (the file exists by then), or main goes red between the two merges.
• Dev-mode validateSchema declares disabled / visible "must be a boolean", so every expression-valued gate is reported invalid — the console line sits right beside the real diagnostic #6505 first, then fix(react): report a faulting disabled / disabledOn predicate instead of silently greying the control out #6510 — fix(react): report a faulting disabled / disabledOn predicate instead of silently greying the control out #6510 must rebase and change the 2 to a 1.

Whoever takes it should also consider whether the raw-total pin is worth keeping in a dev cell at all now: with the #6505 false positive gone, allWarnings and nonValidatorWarnings are the same number in every dev cell, so the line's stated value ("a build that printed our line twice") is already carried by the by-name assertion beside it.

Ref: #6445, #6505, PR #6510.

---

Generated by Claude Code

[os-support-ai]

Measured directly, not inferred. A throwaway worktree at PR #6510's head cbb77c304 with the #6505 fix commit cherry-picked on top (247fd634a), running both suites:

pnpm exec vitest run packages/core/src/validation/ \
 packages/react/src/__tests__/SchemaRenderer.disabledGateFaultDiagnostic.test.tsx \
 packages/react/src/__tests__/SchemaRenderer.validation.test.tsx

 Test Files 1 failed | 8 passed (9)
 Tests 1 failed | 157 passed (158)

The one failure is the predicted line, and nothing else moves:

FAIL packages/react/src/__tests__/SchemaRenderer.disabledGateFaultDiagnostic.test.tsx
 > #6445 group 1 ... > development reports it too - one wiring, not a `__DEV__` branch
AssertionError: expected [ Array(1) ] to have a length of 2 but got 1
 ❯ .../SchemaRenderer.disabledGateFaultDiagnostic.test.tsx:372:33
 371| expect(nonValidatorWarnings(warn)).toHaveLength(1);
 372| expect(allWarnings(warn)).toHaveLength(2);

nonValidatorWarnings(...).toHaveLength(1) on the line above stays green — the by-name subtraction is invariant under the #6505 fix, exactly as its docblock intended. Only the raw-total pin moves, 2 → 1. The other two dev cells and every production cell are green, as is SchemaRenderer.validation.test.tsx (the data-obj-schema-invalid consumer, which fires on a missing type, not on a gate value).

Fix is 2 → 1 on line 372, in whichever PR lands second.

---

Generated by Claude Code

---

Generated by Claude Code

[claude]

Resolved by PR #6513 (the objectui#6505 fix), which now carries the one-line change. Closing so nobody re-measures a question that has already resolved.

It resolved in the #6505 direction. When this was filed, PR #6510 was still in the merge queue and the fix genuinely depended on merge order. #6510 has since merged as 30c73cd25, so #6505 is the one landing second and the adjustment is its to carry.

Re-measured on a real merged base, not the cherry-pick simulation this issue was filed from. origin/main was merged into claude/issue-6505-predicate-valued-gate-rules (merge, never rebase), and the suites were run before touching anything, over packages/react/ and packages/core/:

Test Files 1 failed | 161 passed (162)
 Tests 1 failed | 2950 passed (2951)

FAIL SchemaRenderer.disabledGateFaultDiagnostic.test.tsx:372
AssertionError: expected [ Array(1) ] to have a length of 2 but got 1

Exactly the one cell predicted here, and no other assertion in either package moved — so the interaction was no wider than the original measurement showed.

What #6513 does with it, for the record:

• 2 → 1 on that line. The pin is not loosened — toHaveLength stays, still asserting nothing else reaches the console.
• The reason is carried in a comment at the assertion, not only in a PR description: the second line was the disabled must be a boolean false positive, and objectui#6505 removed it.
• The DEV_SCHEMA_VALIDATOR_NOISE docblock, which stated that false positive as present-tense current behaviour, is rewritten as history in the same hunk. The by-name nonValidatorWarnings helper it justifies stays — every assertion using it was green throughout, which is exactly why this file needed one number changed instead of a re-audit.

Reopen if #6513 is abandoned rather than merged; the pin on main is correct until it lands.

---

Generated by Claude Code

---

Generated by Claude Code