lint: the general outer-mutation-without-local fence has no open tracker — lint.c:1844 points at #396 and #404, both closed

[InauguralPhysicist]

Finding

src/lint.c:1836-1845 explains why W015 is scoped to function-name clobbering only, and hands the general case off:

The general local-discipline lint belongs to the scope-aware name-resolution pass (#404), which can do the dataflow to tell benign reuse from a real clobber; #396 tracks that.

Both referenced issues are closed (#404 "Lint/LSP: scope-aware name-resolution pass", #396 "Lint: fence the two load-bearing traps"). Nothing open tracks the general fence, so the comment reads as "covered elsewhere" while pointing at nothing.

The gap that remains open

llms.txt:63-64 calls missing local "the single most common scope bug". Neither shape is currently flagged at v0.38.0 / 078e759:

total is 100
define helper as:
    total is 0            # no `local` -> clobbers the module binding
    for i in range of n:
        total is total + i
    return total
print of (str of (helper of 4))
print of ("caller total is now " + (str of total))

6
caller total is now 6

define f(flag) as:
    if flag == 1:
        local t is 1
    else:
        t is 2           # sibling branch, no `local` -> writes outward
    return t
print of (str of (f of 0))

2

Both: no issues found.

The second is the more interesting one, because it is the trap llms.txt:66-68 documents explicitly ("each sibling if/elif/else branch needs its own local for a first assignment") and it has a statically decidable signature: a name is local-declared on one branch of an if/elif/else and bare-assigned on a sibling. That does not need full dataflow, and it does not have W015's noise problem — the presence of the local on the sibling branch is direct evidence the author intended a local.

Not arguing with W015's scoping

The rationale at lint.c:1836-1845 is evidence-backed (155 benign firings for the broad reading) and I think it is right — a blanket outer-mutation warning would be noise. This is only about the tracker being dangling and about the narrower, higher-precision subcase above being available without the full pass.

Suggested scope

1. Reopen or supersede the tracking issue so lint.c:1844's reference resolves.
2. Land the sibling-branch check as its own W-code — high precision, no dataflow required.
3. Leave the general case where it is until the scope-aware pass exists.

Related: #396, #404, #373, W015/W016 in src/lint.c.