[sergo] Sergo Report: 34th-Linter lenstringsplit Audit + ctxbackground Scope-Boundary - 2026-06-24

[github-actions[bot]]

Executive Summary

Run R46 detected a registry delta (33 → 34 analyzers) and audited the new 34th linter lenstringsplit, plus a second never-deep-audited linter ctxbackground. Both yielded distinct, single-file precision bugs with confirming test gaps. Filed 2 issues; cached reconciliation surfaced no new syntactic holdout (the 5 known ones remain open and unfixed).

Metric	Value
Serena LSP tools	23 (no change, stable R4–46)
Registered analyzers	34 (Δ +1: lenstringsplit)
doc.go header	still "30 active" → 30-vs-34 drift (already #40436)
Issues filed	2 (#41163, #41164)
Prior-run landing	R45 sg45a1 → #40947

Tool & Registry Updates

• Serena tools: 23, unchanged. No additions/removals vs cached snapshot.
• Registry delta 33→34: grep -c Analyzer cmd/linters/main.go = 34. The new linter was identified via the doc-omitted-minus-known detector: doc.go omits 5 registered linters; four (hardcodedfilepath, sprintferrdot, sprintferrorsnew, wgdonenotdeferred) are known prior additions, leaving lenstringsplit as the genuinely new, never-referenced one.

Strategy (50/50 Split)

New exploration (50%) — fresh audits of unaudited linters. Target the 34th linter lenstringsplit and the never-deep-audited ctxbackground. Success criterion: a demonstrable correctness/precision bug backed by a concrete input and a test-coverage gap.

Cached reuse (50%) — syntactic_stdlib_match reconcile + landing recheck. Re-grep all ident.Name ==/!= "pkg" sites and recheck prior landings. Success criterion: confirm holdout status (open ≠ fixed) without filing duplicates.

Findings

#41163 — lenstringsplit: empty raw-string separator escapes the non-empty guard (FP + wrong autofix)

hasProvablyNonEmptySeparator (pkg/linters/lenstringsplit/lenstringsplit.go:104-116) has a *ast.BasicLit fast-path that compares the raw source token text lit.Value != "\"\"" instead of the decoded value. An empty raw string separator `` has lit.Value of two backticks, which is `!= """"`, so it is treated as "provably non-empty".

• Impact: len(strings.Split(s, ``)) is flagged (the exact empty-separator case the linter intends to exclude) and the suggested rewrite strings.Count(s, ``)+1 is not semantically equivalent (Split("") → runes, len = N; Count("")+1 = N+2).
• Irony: the general constant branch directly below would decode `` to "" and reject it — but it is unreachable.
• Test gap: testdata countRunes exercises only the double-quoted "" form.
• Fix: drop the BasicLit special-case; rely on constant.StringVal (handles all string-constant forms uniformly).

#41164 — ctxbackground: enclosing-scope walk only inspects *ast.FuncDecl, skipping FuncLit (FN for closures with own ctx)

The linter resolves the enclosing function's ctx parameter with cur.Enclosing((*ast.FuncDecl)(nil)) (pkg/linters/ctxbackground/ctxbackground.go:43), which steps over any intervening *ast.FuncLit.

• False negative (primary): context.Background() inside a closure that receives its own ctx context.Context is missed — the walk jumps to the outer FuncDecl, finds no ctx, and breaks. The nearest function scope is exactly the linter's target.
• Secondary (suspect autofix): for a detached go func(){ context.Background() }() inside a ctx-receiving function, the call is flagged and rewritten to the outer captured ctx — often semantically wrong for goroutines that intentionally outlive the request.
• Test gap: testdata has no closure/FuncLit cases.
• Fix: widen to Enclosing(FuncDecl, FuncLit) and resolve the nearest function scope via *ast.FuncType.
• Not a dup: closed ctxbackground precision: isContextBackgroundCall matches identifier name "context" syntactically while the same file detects the [Content truncated due to length] #38789 was the package-name syntactic match (since fixed, now type-resolved); this is the distinct scope-boundary root cause — same family as wgdonenotdeferred precision: non-deferred wg.Done() inside a goroutine launched in a loop escapes detection (FuncLit scope bound [Content truncated due to length] #40947 (wgdonenotdeferred).

Cached Reconciliation (no new issue)

Holdout & landing status

• 5 syntactic ident.Name holdouts STILL open + unfixed in code: Linter precision: migrate the 3 remaining CI-enforced linters that match stdlib packages by identifier name to astutil.IsPkgSele [Content truncated due to length] #40243 (osexitinlibrary:59, fprintlnsprintf:144, errstringmatch:100), Linter precision: two newest linters (fileclosenotdeferred, contextcancelnotdeferred) match stdlib packages by identifier name — [Content truncated due to length] #40435 (contextcancelnotdeferred:138, fileclosenotdeferred:172). Open ≠ fixed — confirmed via re-grep.
• tolowerequalfold now fully type-resolved (IsPkgSelector at :259/:274 + collectCaseConvAliases) → tolowerequalfold precision: strings.ToLower/ToUpper matched by identifier name "strings" — shadowing FP + alias-import FN in a C [Content truncated due to length] #40580 landed in code (issue still open, team rarely auto-closes).
• No new holdout surfaced this run.

Generated Tasks

1. #41163 — Remove the BasicLit text-compare fast-path in lenstringsplit; rely on decoded constant value. Effort: small.
2. #41164 — Make ctxbackground consider FuncLit scope boundaries (Enclosing(FuncDecl, FuncLit) + *ast.FuncType helper). Effort: small–medium.

Metrics & Historical Context

• Total runs: 46 · cumulative findings: 329 · tasks: 92 · avg success: 8.8/10.
• scope_boundary_funclit pattern now confirmed twice (wgdonenotdeferred wgdonenotdeferred precision: non-deferred wg.Done() inside a goroutine launched in a loop escapes detection (FuncLit scope bound [Content truncated due to length] #40947 stack-walk-no-break; ctxbackground Enclosing(FuncDecl)-only) — a recurring, high-yield probe class.

Recommendations & Next-Run Focus (R47)

• Recheck landings: wgdonenotdeferred precision: non-deferred wg.Done() inside a goroutine launched in a loop escapes detection (FuncLit scope bound [Content truncated due to length] #40947, #41163, #41164, tolowerequalfold precision: strings.ToLower/ToUpper matched by identifier name "strings" — shadowing FP + alias-import FN in a C [Content truncated due to length] #40580 (may close).
• Watch the 35th analyzer via grep -c Analyzer + doc-omitted-minus-known detector.
• Keep probing scope_boundary_funclit: any linter resolving enclosing-function context that filters only *ast.FuncDecl or walks past FuncLit.
• Fresh-probe remaining: sortslice, execcommandwithoutcontext, hardcodedfilepath, manualmutexunlock, panicinlibrarycode.
• Suppression parity: ctxbackground has test-skip but no nolint — evaluate vs the Threshold linters: excessivefuncparams lints test files (every sibling FuncDecl linter skips them); largefunc & excessivefuncpar [Content truncated due to length] #40734 theme.

References: §28076275497

Generated by 🤖 Sergo - Serena Go Expert · 302.6 AIC · ⌖ 11.3 AIC · ⊞ 5.9K · ◷

• expires on Jun 24, 2026, 9:12 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 #41378.