slicemakezerolength (69th linter): make()-only scope misses the more idiomatic var s []T nil-slice pattern

[github-actions]

Problem

slicemakezerolength (added 2026-09-11 via PR #60310, the 69th registered analyzer) flags make([]T, 0) calls that precede an unconditional one-element-per-iteration range+append loop, suggesting capacity len(range-expr). Its entire detection surface requires the slice variable to originate from a two-arg make([]T, 0) call: zeroLengthSliceAssignment (pkg/linters/slicemakezerolength/slicemakezerolength.go:67-122) only matches an AssignStmt or DeclStmt whose right-hand value is a *ast.CallExpr resolving to the builtin make, and the DeclStmt branch additionally requires len(spec.Values) == 1, i.e. an explicit initializer.

Evidence

pkg/parser/content_extractor.go:169-180 has exactly the pattern this linter exists to catch:

var normalizedValue []any
switch v := fieldValue.(type) {
case []string:
    for _, s := range v {
        normalizedValue = append(normalizedValue, s)
    }
}

v is a []string (hasKnownRangeSize accepts string/slice/array/map underlying types) and the loop body is a single unconditional append (appendsOneElement accepts exactly this shape). The only reason this real production site passes undetected is that normalizedValue is declared with var normalizedValue []any, a zero-value nil slice with no initializer, instead of make([]any, 0). Because ValueSpec.Values is empty in that case, zeroLengthSliceAssignment returns false at the len(spec.Values) != 1 check before it ever inspects a make call.

Impact

var x []T is the idiomatic, more common Go spelling for exactly this initialize-then-append-in-a-known-size-loop shape (common Go style guidance prefers the nil-slice form over make([]T, 0) for a var that will only ever grow via append). A linter whose stated goal is finding capacity-preallocation opportunities but whose match rule only recognizes the less-idiomatic make([]T, 0) spelling will have a systematically low real-world hit rate: developers who already reach for the preferred spelling are entirely invisible to it.

Recommendation

Extend zeroLengthSliceAssignment to also accept a DeclStmt ValueSpec with zero Values (a bare var s []T where the type is an *ast.ArrayType with Len nil) as a starting state equivalent to make([]T, 0) — both begin as an empty, zero-capacity slice. Add a testdata case mirroring content_extractor.go (var s []T; for _, x := range known-size-expr { s = append(s, x) }) alongside the existing badZeroLengthNoCapacity case.

Validation checklist

• New testdata case for the bare var declaration produces a diagnostic.
• Existing make([]T, 0) and make([]T, 0x0) cases keep passing unchanged.
• A bare var s []T with no follow-on loop (mirroring goodWithoutGrowth) still produces no diagnostic.

Effort: small — one additional branch in zeroLengthSliceAssignment plus one testdata case.

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• api.anthropic.com

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "api.anthropic.com"

See Network Configuration for more information.

Generated by 🤖 Sergo - Serena Go Expert · claude · agent · 202 AIC · ⌖ 7.68 AIC · ⊞ 6.8K · ◷

• expires on Sep 18, 2026, 7:56 PM UTC-08:00

[github-actions]

This issue was automatically closed because it expired on 2026-09-19T03:56:16.130Z.

Closed by Workflow