metadata-protocol: duplicatePackage renames a token that names both a prefixed package object and one of its fields as the object, so a copied view's column or form field names a field the copy does not have (family close-out) #23561

Filing gate ①, finding class (a), the close-out card for the duplicatePackage "names out of step" family (#23527, #23537, #23538). reach: the named producer ObjectStackProtocolImplementation.duplicatePackage (the method the packages duplicate route calls), on a real kernel (ObjectKernel + ObjectQLPlugin, memory driver), measured at main 8226755ee9 and at PR #23559's head dbd69f7592. The HTTP door was not driven.

Filed by domain:spec seat 4 (#18917) · sunhaolin · session session_0121ooBYRmS8ziFXm9ai5YHm, from #23538's dev report 6181410407 (open question and out-of-scope finding 1). ⛔ Not a claim. Lane, type and grade are triage's.

What happens (measured)

The source object dupsrc_customer has a field keyed and named dupsrc_customer. That token is both a prefixed package object's name and a field key. The free-text rewrite renames it as the object everywhere it appears as a value:

• A list view with columns: ['full_name', 'dupsrc_customer']:
  • on main it copied with column dupdst_customer, which the data plane refuses (Unknown field 'dupdst_customer' on object 'dupdst_customer');
  • with PR fix(metadata-protocol): duplicatePackage keeps a copied field's key and name, and a bare object's view item name, in step with what they name #23559 it is refused at the copy: view/dupdst_customer.board failed author-time validation: … views[0].config.columns[1] [list-view-field-unknown]. That refusal is loud and located.
• A form view with sections[].fields: ['full_name', 'dupsrc_customer'] copies silently as ['full_name', 'dupdst_customer'], a field the copy does not have.

Pull, measured: across examples/app-crm, app-showcase and app-todo (51 object names, a grep of *.object.ts), no field key equals an object name. The repro is synthetic.

What done looks like

A copied value that names a FIELD keeps naming that field, in list-view columns and form sections[].fields (the sites to enumerate). A value that names the OBJECT still follows the object's rename.

The token is ambiguous, so the fix picks a semantics. The dev's readings:

• (B) rename such a name only at object-reference sites, never in free text, extending the bare-name policy (about 15 lines plus a pin);
• (C) rename the field too, which reverses the landed rule that a fields key's field part is never read.

Triage decides whether this is a seat call or a decision card.

Noted, not measured (read from the code): a view CONTAINER row (no ., for example ticket_views) is still renamed by the prefix swap alone, so a bare object's container copies under the source's name. A container name has no declared <object>. identity, so it is not a contract breach. It is recorded here for the family, not for a fix.

Reader who acts: triage, to grade and route. The fix lands in duplicatePackage (packages/metadata-protocol/src/protocol.ts), after PR #23559 (#23538) lands.

Dedupe

MCP search_issues on objectstack-ai/objectstack, closed included, query duplicatePackage field reference renamed as object free-text rewrite field token equals object name copied view column unknown field: 1 hit, #23538 (open), the parent family card, which is not this residual.

Dedupe words: duplicatePackage field token equals object name; copied view column unknown field after duplicate; form sections field dangling after duplicate.