[sergo] Sergo Report: Registry-Refactor Detector Migration + Autofix-Precedence Audit - 2026-07-23

[github-actions[bot]]

Executive Summary

R61 opened with a large discontinuity: my persisted memory was last updated at R60 (2026-07-08, 43 linters), but the repository has since advanced to 57 registered analyzers — and all sergo-labelled issues are closed (gh-api verified). The prior new-linter detector (grep -c Analyzer cmd/linters/main.go) now returns 0: the registry was refactored out of cmd/linters/main.go (now a bare multichecker.Main(linters.All()...)) into pkg/linters/registry.go func All(), with doc-completeness now enforced by pkg/linters/doc_sync_test.go.

Against that backdrop I ran a 50/50 strategy and filed one precise, novel, non-duplicate issue: a precedence/parenthesization gap in the stringsjoinone autofix.

Tool / Registry Updates

Item	R60	R61
Serena LSP tools	23	23 (stable)
Registered analyzers	43	57 (+14)
Registry location	cmd/linters/main.go	pkg/linters/registry.go func All()
New-linter detector	grep -c Analyzer main.go	count All() entries / doc.go All NN active analyzers
Doc-sync	manual (drift bugs)	enforced by doc_sync_test.go
CI-enforced linters	14	25 (cgo.yml LINTER_FLAGS)

The old detector is dead and has been replaced in memory for R62.

Strategy — strict 50/50 split

Cached (50%) — autofix compile-correctness / behavior-preservation vein (proven at R60, success 9; family of #45037 / #44653 / #44864 / #44865). Adapted from import-injection to a fresh sub-facet: parenthesization/precedence of extracted sub-expressions in SuggestedFix NewText.

New exploration (50%) — the new simplification-family linters added since R60 that my strategy history had never covered: stringsjoinone, appendoneelement, stringbytesroundtrip, bytesbufferstring, stringreplaceminusone. Novel target area driven by the +14 registry growth.

Run targets: ≥1 novel evidence-backed finding, 1 high-quality issue, no duplicates. Met.

Findings

FILED — stringsjoinone autofix precedence gap (#47467). The fix sets NewText: []byte(elemText), bare-replacing the entire strings.Join(...) call with the unparenthesized element text (stringsjoinone.go:88-96, :131). When the element is compound and the call is an operand of a tighter postfix operator, the fix breaks:

Failing cases

// Case A — NON-COMPILING (a, b are string):
first := strings.Join([]string{a + b}, "")[0]
// -fix -> first := a + b[0]   // b[0] is byte -> string + byte -> compile error

// Case B — silent behavior change (compiles):
tail := strings.Join([]string{a + b}, "")[1:]
// -fix -> tail := a + b[1:]   // was (a+b)[1:]

Not CI-enforced (latent, -fix-only). RWSF golden only covers bare identifiers/literals in return/assignment contexts, so RunWithSuggestedFixes passes while the fix is unsound.

AUDITED-CLEAN (no issue — contrast that sharpens the diagnosis):

Why the siblings are safe

• appendoneelement — injects elemText into append(s, elemText) (comma-arg position). Precedence-safe. Also type-safe: literal []T must be identical to s's element type for the spread to compile.
• stringreplaceminusone — injects args into pkg.ReplaceAll(s, old, new) (arg positions). Safe.
• bytesbufferstring — receiverText + ".String()"; receiver is always a selector operand (primary expr), and it intentionally excludes *bytes.Buffer to avoid the nil-panic-vs-"<nil>" divergence (:126-135). Safe.
• stringbytesroundtrip — report-only (no autofix); its "inner conversion is redundant" claim holds even for named string / byte-slice types because it targets the inner conversion, not the whole expression. Clean.

stringsjoinone is the only linter doing a bare full-expression substitution, so it is the only one exposed.

Generated Task

Precedence-safe stringsjoinone fix. Parenthesize elemText when the element is not atomic (anything other than *ast.Ident/*ast.BasicLit/*ast.ParenExpr/*ast.SelectorExpr/*ast.IndexExpr/*ast.CallExpr), or drop the SuggestedFix when the call's parent is *ast.IndexExpr/*ast.SliceExpr. Add RWSF golden cases for compound element in indexed and sliced contexts. Effort: small (~5 lines + 2 goldens).

Metrics

Metric	Value
Findings	6 (1 filed, 5 audited-clean)
Issues created	1 (#47467)
Duplicates skipped	0 (no prior stringsjoinone issue)
Self-scored success	9 / 10
Cumulative runs / findings / tasks	61 / 414 / 110

Historical Context

The autofix-correctness vein remains the single most productive Sergo theme (#45037 import-injection, #44653 writebytestring import, #44864 sprintfint import, #44865 tolowerequalfold behavior, fprintlnsprintf trailing newline). R61 adds a new sub-facet — precedence/parenthesization — orthogonal to the import and named-type-assignability facets and not previously probed.

Recommendations & Next-Run Focus (R62)

1. Sweep the remaining autofix linters for bare full-expression replacement that inlines a sub-expression without parens (the stringsjoinone shape) — this run only confirmed the immediate simplification family is otherwise arg/selector-scoped.
2. Re-verify #47467 landed and re-audit its golden coverage.
3. Watch for the 58th analyzer using the migrated detector (registry.go All() entry count / doc.go header).
4. Continue gh api reconcile first — the team lands and closes fixes within hours.

References: §29980457915

Generated by 🤖 Sergo - Serena Go Expert · age00 260.1 AIC · ⌖ 13.9 AIC · ⊞ 5.8K · ◷

• expires on Jul 23, 2026, 9:00 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 #47705.