[sergo] Sergo Report: Tool-Change Detect (22nd linter lenstringzero) — Doc/Spec Sync + Precision FN — 2026-06-08

[github-actions[bot]]

Executive Summary

Run R30 detected a tool change: a 22nd custom analyzer, lenstringzero, landed in the live linter driver (ADR-37618). Serena's own LSP toolset is unchanged (23 tools, stable since R4). Both R29 precision issues landed and were closed by the team (#37492 var-alias, #37493 nolint comma-list), confirming the precision_fix_lands pattern (single-file fixes land in ~1 day). The new linter shipped in the driver ahead of its documentation/spec surface, producing one clean self-consistency gap and one documented precision false-negative.

• 🔧 Tool change: linter registry 21 → 22 (lenstringzero); Serena LSP 23 (unchanged)
• 🟢 Landed & closed: sg29a1 (tolowerequalfold linter: false negative when ToLower/ToUpper result is stored in a local variable before comparison #37492), sg29a2 (internal/nolint: (nolint/redacted) directive only suppresses when the linter is first in a comma-separated list #37493)
• 🆕 New findings: 2 issues created (sg30a1 docs/spec, sg30a2 precision FN)
• ⚖️ Enforcement: 12 enforced (unchanged); lenstringzero not enforced, lacks internal/nolint
• Success score: 9 / 10

Tool / Linter Updates

Surface	Before (R29)	After (R30)	Note
Serena LSP tools	23	23	stable R4–R30
Custom linters (driver main.go)	21	22	lenstringzero added
Documented (doc.go/README/spec_test)	21	21 ❌	not synced (sg30a1)
CI-enforced (cgo.yml:1078)	12	12	lenstringzero not enforced
internal/nolint coverage	14	14	lenstringzero lacks it

Strategy Split (50 / 50)

Cached reuse (50%) — reverify open issues

Applied the proven reverify-landed strategy against the R29/R26 open set:

• sg29a1 (tolowerequalfold var-alias FN) → LANDED: collectCaseConvAliases now builds a map[types.Object]ast.Expr of ToLower/ToUpper result aliases with mutation-invalidation. Issue tolowerequalfold linter: false negative when ToLower/ToUpper result is stored in a local variable before comparison #37492 closed.
• sg29a2 (internal/nolint first-in-comma-list) → LANDED: BuildLineIndex now uses strings.SplitSeq(payload, ",") + membership (nolint.go:55–61). Issue internal/nolint: (nolint/redacted) directive only suppresses when the linter is first in a comma-separated list #37493 closed.
• sg26a1 (seenmapbool escape-blind) → issue Sergo sg26a1: seenmapbool (20th linter) flags function-escaping maps — add escape check before enforcing #36406 closed while inspectBody (lines 64–175) remains escape-blind ⇒ treated as dismissed / won't-fix. Recorded as DO-NOT-REFILE to avoid churn.

New exploration (50%) — registry-diff + new-analyzer audit

Diffed the driver analyzer list (cmd/linters/main.go) against doc.go, README.md, and spec_test.go. Surfaced the 22nd linter lenstringzero and audited it end-to-end: registration ✅, documentation ❌, spec-test ❌, enforcement ❌, nolint ❌, and a precision false-negative documented in its own ADR. New new_linter_doc_lag pattern recorded for future runs.

Run targets: ≥2 evidence-backed findings, ≥2 distinct/non-overlapping issues, all grounded in file:line evidence + an in-repo fix template. All met.

Findings

🆕 sg30a1 — lenstringzero registered but undocumented (issue created)

The driver runs 22 analyzers but doc.go ("All 21 active analyzers"), README.md (bullet list + table + import list), and spec_test.go documentedAnalyzers() (21 entries) all still say 21. Consequence: the newest analyzer has zero spec-test coverage — TestSpec_PublicAPI_SubpackageAnalyzers never validates lenstringzero.Analyzer's Name/Run/name-uniqueness.

• Evidence: cmd/linters/main.go:29,64 vs doc.go:3, README (~28/57/122), spec_test.go:62–86
• Impact: silent regression risk + stale docs; breaks one-package-per-linter self-consistency invariant
• Fix: bump all three artifacts to 22 (Low effort)

🆕 sg30a2 — lenstringzero intermediate-variable false-negative (issue created)

lenstringzero flags only direct len(s) == 0; n := len(s); n == 0 slips through. This is documented as a Negative consequence in ADR-37618 and is the same var-alias class just fixed in tolowerequalfold (#37492). Recommended fix mirrors collectCaseConvAliases (alias pre-pass + mutation invalidation).

• Evidence: pkg/linters/lenstringzero/lenstringzero.go:48–56; ADR Consequences › Negative
• Impact: missed idiom remediation on intermediate-length comparisons
• Fix: alias-tracking pre-pass (Medium effort, proven template)

Carried / roadmap (not refiled this run)

• lenstringzero enforce roadmap: not enforced + lacks internal/nolint. Path to enforcement = add nolint precondition → remediate the ADR's ~20+ direct string sites → append -lenstringzero to cgo.yml:1078. Tracked in cache; not yet filed (pre-remediation).
• sg28a2 nolint parity: coverage 14; 8 linters now lack internal/nolint (the prior 7 + lenstringzero).

Generated Tasks

1. Sync docs/spec to 22 analyzers (sg30a1) — doc.go count+bullet, README bullet/table/import, spec_test table entry. Low effort; self_consistency (proven-to-land).
2. Add alias-tracking to lenstringzero (sg30a2) — mirror collectCaseConvAliases; add positive + negative analysistest fixtures. Medium effort.

Metrics

Metric	R30	Cumulative
Findings	6	233
Issues created	2	—
Tasks generated	2	59
Total runs	—	30
Avg success score	9	8.71

Historical Context

Linter count trajectory: 20 (seenmapbool, R26) → 21 (tolowerequalfold, R27) → 22 (lenstringzero, R30). Enforcement: 11 → 12 (tolowerequalfold landed R29) and stable this run. The team continues to convert Sergo issues into PRs and close them within ~1–2 days; the open-sergo set is currently empty, which is why R30 pivoted to tool-change detection and new-analyzer auditing rather than reverification alone.

Recommendations / Next-Run Focus (R31)

• Verify sg30a1 doc-sync and sg30a2 alias-FN have landed (expect fast).
• Advance the lenstringzero enforce roadmap once the ~20+ direct string sites are remediable: nolint integration → CI flag.
• Continue per-run main.go ↔ doc.go ↔ spec_test.go registry-diff (new new_linter_doc_lag guard); watch for a 23rd linter.

References:

• §27117446388
• ADR-37618 docs/adr/37618-add-lenstringzero-linter.md
• Landed: tolowerequalfold linter: false negative when ToLower/ToUpper result is stored in a local variable before comparison #37492 (var-alias FN), internal/nolint: (nolint/redacted) directive only suppresses when the linter is first in a comma-separated list #37493 (nolint comma-list)

Generated by 🤖 Sergo - Serena Go Expert · 236 AIC · ⌖ 15.2 AIC · ⊞ 6.5K · ◷

• expires on Jun 8, 2026, 9:25 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Sergo - Serena Go Expert.

A newer discussion is available at Discussion #38030.