[sergo] Sergo Report: Stale-Cache Reconcile + New-Linter Full Audit (43→51) - 2026-07-18

[github-actions[bot]]

Executive summary

The tool cache was stale by ~10 days: the analyzer registry jumped 43 → 51 (+8) while the strategy/stats memory files were last written at R60 (2026-07-08). Nine untracked runs (R61–R69) filed issues #44187...#46131 but never pushed their memory updates. This run reconciled that gap: audited all 8 new linters (44–51), confirmed they are uniformly clean and type-resolved, and filed one high-confidence enforce-readiness issue for the single genuine gap found — mapclearloop.

• Open sergo issues pre-run (gh api): exactly one — mapdeletecheck: autofix silently deletes comments inside the if body — no comment guard, unlike its sibling mapclearloop #46130 (mapdeletecheck autofix drops comments). Not re-filed.
• New issue filed: mapclearloop enforce-readiness (temporary_id aw_sg61a1).
• Success score: 9/10.

Tool / registry updates

LSP tools stable; linter registry surged

• Serena LSP tools: 23, unchanged (matches cache snapshot).
• Analyzer registry: grep -c 'Analyzer' cmd/linters/main.go = 51 (was 43 at R60). pkg/linters/doc.go header also reads 51 — doc-sync is current (doc-sync: pkg/linters/doc.go says "45 active analyzers" but 51 are registered — 6 newest linters omitted #46131 closed).
• 8 new linters (44–51): bytescomparestring, stringscountcontains, bytesbufferstring, mapdeletecheck, mapclearloop, ioutildeprecated, logfatallibrary, nilctxpassed.
• CI enforce set: 23 analyzers, wired at .github/workflows/cgo.yml:1208 (prod gate) and :1211 (wasm gate).

Strategy (50/50 split)

Cached component — new-linter / registry-delta audit + enforce-readiness (proven, scored 9 at R60/R59/R58). Applied the established new-linter playbook to each of the 8 additions: type-resolution vs syntactic package-match, SuggestedFix compile-correctness, RWSF golden verification, and comment/shadow guards.

New-exploration component — comment-preservation guard vein + stale-cache reconciliation. Generalized the #46130 finding (block-replacing autofix drops comments) across the SuggestedFix family, and reconciled the 43→51 registry drift that the cache had missed.

Findings (evidence)

All 8 new linters audited CLEAN (type-resolved)

• nilctxpassed — prior variadic-spread FP (nilctxpassed: spread call f(nil...) to a variadic context.Context param is a false positive — nil variadic slice misreported as [Content truncated due to length] #45903) is fixed: f(nil...) spread calls are now skipped (nilctxpassed.go:60-64). Uses TypesInfo + exact context.Context named-type match.
• ioutildeprecated — complete replacement map for all 8 exported io/ioutil identifiers; resolves package identity via TypesInfo.ObjectOf/Uses (path io/ioutil), handles both qualified and dot-imports. Diagnostic-only (no autofix, no autofix risk).
• bytesbufferstring — correctly excludes pointer receivers to avoid the nil-*bytes.Buffer panic-vs-"(nil)" semantic difference (:90-96, :124-131).
• stringscountcontains — all six comparison forms (>0/>=1/!=0/==0/<1/<=0) map to the mathematically correct Contains/!Contains; empty-substring edge is consistent.
• mapclearloop / mapdeletecheck — the only two block/statement-replacing autofixers. mapclearloop carries both a shadow guard (builtinVisibleAtPos for clear) and a comment guard (hasOverlappingComment); mapdeletecheck lacks the comment guard (that gap is already tracked in mapdeletecheck: autofix silently deletes comments inside the if body — no comment guard, unlike its sibling mapclearloop #46130).

The recurring "syntactic stdlib-package-match" defect class (FP on shadowing, FN on alias) did not reappear in any new linter — all use IsPkgSelector / Pkg().Path() / TypesInfo.

Generated task → issue

mapclearloop enforce-readiness (aw_sg61a1). The analyzer is production-ready — type-resolved detection, shadow-safe + comment-safe autofix, analysistest.RunWithSuggestedFixes golden verification — with zero production violations (repo-wide search for a range-over-map loop whose sole body is delete(m, k) matches only testdata; pkg/cli/devcontainer.go:316 is guarded by an inner if and correctly not flagged). Yet its sibling mapdeletecheck is CI-enforced while mapclearloop appears nowhere in cgo.yml. Recommendation: add -mapclearloop to both LINTER_FLAGS strings next to -mapdeletecheck.

Metrics

Metric	Value
New linters audited	8 (all clean)
Findings captured	8
Issues filed	1 (aw_sg61a1)
Duplicates skipped	#46130 (open), #45629/#45186 (enforce, different linters)
Registry delta	43 → 51 (+8)
Success score	9/10

Historical context

This run corrected a memory-persistence gap: R61–R69 filed issues but never updated sergo-strategies.jsonl / sergo-stats.json (likely failed memory pushes). Cache files were re-synced to registry=51, enforce-set=23, and trimmed to the 12 KB limit (dropped the R58 strategy entry and superseded tool_change_R59/R60 fields per the keep-3-recent rule).

Recommendations & next-run focus (R62)

• Verify aw_sg61a1 lands (add -mapclearloop to cgo.yml).
• nilctxpassed enforce-readiness — diagnostic-only, catches a real anti-pattern; verify zero nil-context production sites, then propose enforcement.
• Watch the 52nd linter via grep -c Analyzer cmd/linters/main.go.
• Track the shared hasOverlappingComment → internal/astutil extraction (already recommended in mapdeletecheck: autofix silently deletes comments inside the if body — no comment guard, unlike its sibling mapclearloop #46130).
• Always gh api reconcile first — the team closes sergo issues same-day.

References:

• §29630770411
• §29555622993

Generated by 🤖 Sergo - Serena Go Expert · 364.2 AIC · ⌖ 14 AIC · ⊞ 5.8K · ◷

• expires on Jul 18, 2026, 8:52 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 #46528.