[sergo] Sergo Report: 32nd-Linter Reconcile + Fresh-Linter Audit - 2026-06-22

[github-actions[bot]]

Executive Summary

Run R44 (44th cumulative). Strategy: strict 50/50 cached-reconcile + new-exploration. The cached half (syntactic_stdlib_match) re-grepped all 32 analyzers and confirmed the five known stdlib-identity holdouts are still syntactic = unfixed (no new holdout surfaced). The new half audited the fresh, never-audited threshold/heuristic linters and found two real defects. 2 issues filed, both single-theme, non-duplicate.

• [seenmapbool: duplicate diagnostics for set-maps declared inside function literals (double AST traversal) #40733] seenmapbool emits duplicate diagnostics for set-maps declared inside closures (double AST traversal).
• [Threshold linters: excessivefuncparams lints test files (every sibling FuncDecl linter skips them); largefunc & excessivefuncpar [Content truncated due to length] #40734] Threshold linters excessivefuncparams/largefunc lack //nolint suppression; excessivefuncparams also lints test files (the lone FuncDecl linter that does).

Tool / Registry Updates

Signal	R43	R44
Serena LSP tools	23	23 (identical)
Registered analyzers (grep -c 'Analyzer')	31	32 (delta +1)
doc.go header	"29 active"	"30 active" (lists 29 bullets, omits 3, registry 32)

The doc.go drift is the already-open #40436 — skipped, not refiled. The exact 32nd linter could not be identified under the shallow clone (single squashed commit, no per-file dates).

Strategy Split

Cached half — syntactic_stdlib_match reconcile (proven, high-recency-stale)

Re-grepped \.Name (==|!=) \"<pkg>\" across pkg/linters using both operators (linters use != for early-return; an ==-only grep misses holdouts). Result — all five known sites unchanged in code:

Site	Pkg	Issue
osexitinlibrary.go:59	os	#40243
fprintlnsprintf.go:144	fmt	#40243
errstringmatch.go:100	strings	#40243
fileclosenotdeferred.go:172	os	#40435
contextcancelnotdeferred.go:138	context	#40435

All still match the stdlib package by identifier name (no astutil.IsPkgSelector) → both issues remain unfixed and open. Separately, tolowerequalfold has been refactored to type-resolved alias collection (collectCaseConvAliases), with NodeText now used only for suggested-fix text — so #40580 is likely being addressed (flagged to recheck next run). No new holdout surfaced, so the cached half produced zero new issues — pure reconcile.

New half — fresh-linter precision audit (underused targets)

Audited the heuristic/threshold and set-map linters never examined before. Findings:

• seenmapbool double-report (filed): node filter registers {FuncDecl, FuncLit} and inspectBody recurses with ast.Inspect into nested closures — so a set-map declared in a closure is collected/reported once by the enclosing function's pass and again by the closure's own Preorder visit (2× normally, 3× if nested two deep). go/analysis does not de-dup by position. largefunc uses the same node filter but measures only each node's own span (no recursion) and is therefore correct — a clean contrast.
• Threshold-linter suppression gaps (filed): 20/32 analyzers wire internal/nolint; largefunc and excessivefuncparams do not, so a legitimately-large function/wide signature cannot be suppressed at the call site (blocks CI enforcement). excessivefuncparams additionally omits the filecheck.IsTestFile guard that every sibling FuncDecl linter applies.

Audited clean (no issue): errorfwrapv (type-aware IsFmtErrorf + types.Implements(errorIface) + full format-verb/flag/width/explicit-index parsing; %w correctly skipped), ossetenvlibrary (fn.Pkg().Path()=="os" — type-resolved, the right way to do package identity), largefunc, deferinloop (cursor Enclosing with correct FuncLit scope boundary).

Generated Tasks (2)

1. Fix seenmapbool double traversal — scope each map declaration to exactly one function scope (stop inspectBody recursion at nested FuncLit boundaries while keeping the write-disqualification pass able to see closure writes, or de-dup reports by declaration position). Add a closure testdata case (current testdata has none). Effort: small.
2. Add //nolint + test-skip parity to threshold linters — wire internal/nolint into largefunc and excessivefuncparams; add filecheck.IsTestFile to excessivefuncparams. Prerequisite for promoting either to CI enforcement. Effort: small, follows 20 existing analyzers.

Metrics

Metric	Value
Cumulative runs	44
Cumulative findings	317
Cumulative tasks	89
Avg success score	8.81
This run	6 findings, 2 issues, score 9

Historical Context

The stdlib-identity family (syntactic_stdlib_match) has been the durable thread since R33: landed fixes (sortslice #38029, rawloginlib #39981, regexp #39733) plus open holdouts (#40243, #40435). Each new linter is probed for re-introducing ident.Name==pkg; R44 found the newest fresh linters split — ossetenvlibrary did package identity correctly (type-resolved), while fileclose/contextcancel (R42) regressed it. R44 adds a new reusable pattern: double_traversal_dup — any linter combining a {FuncDecl, FuncLit} node filter with an inner ast.Inspect recursion risks duplicate diagnostics.

Recommendations & Next-Run Focus (R45)

• Recheck landings: tolowerequalfold precision: strings.ToLower/ToUpper matched by identifier name "strings" — shadowing FP + alias-import FN in a C [Content truncated due to length] #40580 (refactored — may now be closeable), 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/Linter precision: two newest linters (fileclosenotdeferred, contextcancelnotdeferred) match stdlib packages by identifier name — [Content truncated due to length] #40435 (still syntactic at R44), errstringmatch coverage gap: only strings.Contains(err.Error(), ...) is flagged — HasPrefix/HasSuffix/EqualFold/Index on err.Err [Content truncated due to length] #40244/jsonmarshalignoredeerror FN: bare-call json.Unmarshal(data, &v) (ExprStmt) discards the error but is never inspected — only As [Content truncated due to length] #39982/sprintferrdot precision: verb handling {s,v} is wrong in both directions — %#v false positive, %q/%x/%X false negative #40434/lenstringzero coverage gap: len(s) != 0 is flagged but the identical idiom len(s) > 0 (and len(s) < 1 / <= 0) escapes — relation [Content truncated due to length] #40581/seenmapbool: duplicate diagnostics for set-maps declared inside function literals (double AST traversal) #40733-2.
• Probe double_traversal_dup on errormessage and any other {FuncDecl, FuncLit} + recursive-walk linter.
• Suppression parity audit across the 12 non-enforced analyzers.
• Identify the 32nd linter; watch for a 33rd registry delta.

References:

• §27931512989

Generated by 🤖 Sergo - Serena Go Expert · 357.9 AIC · ⌖ 10.9 AIC · ⊞ 5.8K · ◷

• expires on Jun 22, 2026, 9:42 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-23T05:42:44.664Z.

Closed by Workflow