[ASDAlexander77]

A batch of -mm=own gap fixes (plus one rc fix), from a new fuzz campaign and the open leak list. There are five commits, each with its own tests.

1. rc: a conditional between record literals keeps the objects its branch made (d82cfbea)
c ? { start: this.pos, dir: new V(1) } : {...} caused heap corruption under rc. The hand-written fuzz seed review7__pM1 still failed on main. Each branch gave back its new V before the receiver after the ts.If retained the literal's fields. consumeConditionalResults now makes such a branch retain the literal it yields, as it already does for a field read out of a branch temporary.

• Test: tests/00conditional_record_literal_temporary.ts (gc, plus rc and none through the corpus).

2. own: three borrowing literals that compiled and then crashed (edeb47a9)
A 2000-program mm_fuzz round (seeds 73000-74999) found 3 own crashes among the 1735 programs own accepted. Each is now a compile error:

• f([HG.v]) where f does p[0] = v: an array literal that borrows its elements no longer borrows when its callee may overwrite an element or destroy anything.
• f({ v: n.items[0] }) where f does p.v = ...: a borrowing record given to a callee whose overwrites include a field is treated as a write of that field.
• walkRecord treated ts.Invoke (a call inside try) as a branch and skipped it. A borrowing record given to a call in a try was therefore never checked: f({ v: n.items[0] }, (n = mk(), n)) freed the element first.
• Tests: own_err_literal_element_overwritten, own_err_record_field_overwritten, own_err_record_borrow_in_try. Each compiled without error before this change.

3. own: a getter-filled interface field read put in a let is freed at the returns (e5f304ba)
let x = i.item; read across a loop leaked on the getter's path, because #751 only frees reads that stay in their block. For 2M calls, peak working set went from 280.5 MB to 4.3 MB. The read is released at each return when all of these hold:

• the let is only read, and nothing read from it leaves the function;
• the read reaches every return;
• the object it reads through is a parameter's.

A read that would still leave its block is now an error (commit 4).

• Tests: own_interface_getter_let (own, no-counting, rc, gc, none) and test-own-interface-getter-let-released, which reads the IR (fails before the fix).

4. own: a getter-filled interface field read that would leave its block is an error (a588196d, your ruling)
return i.item is now a compile error when a class in the module fills item with a getter. Before, it leaked one value per call on the getter's path. A field that no class fills with accessors is a borrow on every path and may still be returned. A method called on the read (i.item.get()) is now followed, so that read is freed rather than rejected. The DefaultLib's src/lib.ts compiles under own and rc with no errors.

• Test: own_err_getter_read_returned.

5. own: a callback may give a captured object a new value (9804fff0, item 4)
bag.forEach(b => { last = new Box(b.n) }) was rejected. Assigning a captured variable in a closure counted as "may destroy anything", so a method that walks a field's array and calls the callback could not borrow its elements. Spec 52 already handled numbers and strings precisely; this extends it to objects and arrays:

• The assignment is now its own drop place, OWN_CAPTURED_CELL_POSITION. It drops only borrows read through a captured variable's cell.
• To keep that sound, a call that may run such a closure and is also given something read through the assigned variable is an error at the call (bag.forEach(b => { bag = new Bag() })).
• Tests: own_callback_assigns_capture (own, no-counting, rc, gc, none; rejected before), own_err_callback_assigns_receiver, own_err_callback_assigns_borrowed.

The same commit also fixes an older own crash found by fuzz seed 77055, which predates this batch: f(n, (n = mk(3), ...)), where n is given as an interface and reassigned in a later argument. checkArgumentsAssigned looked through casts but not through the interface view, so it never found n. It now looks through views.

• Test: own_err_interface_arg_assigned_in_call.

Verification

• Release ctest: 6518/6518.
• Fuzz: seeds 72000-72399 had 0 failures before these fixes. Seeds 75000-75599 run after them: 0 failures in 1111 checked programs; seeds 76000-76399 after commit 4, and 77000-77599 and 78000-78599 after commit 5, also had 0 failures.
• Seed replay: 2 of 30 still fail, review6r1__v1/v1b. Those are rc crashes from a function cast through as any to another interface's signature; they are not addressed here.

🤖 Generated with Claude Code