[sergo] Sergo Report: 30th-Linter Delta (sprintferrdot) - 2026-06-20

[github-actions[bot]]

Executive Summary

R42 detected a registry delta: the custom linter suite grew 29 → 30 analyzers. The new (30th) linter is sprintferrdot (flags redundant .Error() on error args passed to fmt format functions). Both newly-landed linters were never audited by Sergo and yielded three distinct, high-confidence findings → 3 issues filed (#40434, #40435, #40436). Issue reconciliation against gh api was clean — every prior open issue is accounted for, no phantoms.

Tool & Registry Updates

Check	R41	R42	Signal
Serena LSP tools (serena --help)	23	23	stable (R4–R42)
Registered analyzers (cmd/linters/main.go)	29	30	+1 — sprintferrdot
doc.go header	—	says 29	doc-lag (stale)

Under a shallow clone (no git dates), the main.go Analyzer-count vs doc.go header delta is the reliable new-linter detector — it fired this run.

Strategy: strict 50/50 split

• Cached half (proven syntactic_stdlib_match, strongest land-arg): the team has repeatedly eliminated ident.Name == "<pkg>" matching in favor of astutil.IsPkgSelector (landed: sortslice, rawloginlib, regexpcompileinfunction; 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 in-flight). I probed whether the newest linters regressed it — and both did.
• New half (exploration on the never-audited 30th linter): a fresh pattern_set_too_narrow audit of sprintferrdot's format-verb handling against the authoritative fmt spec (verified via WebFetch of pkg.go.dev/fmt).

Run targets: ≥2 precision findings on new linters; ≤3 issues, all non-duplicate, single/small-PR scope; cache updated. All met.

Findings → Issues

ID	Linter(s)	Class	Severity
#40434	sprintferrdot (30th)	pattern_set_too_narrow	FP + FN
#40435	fileclosenotdeferred, contextcancelnotdeferred	syntactic_stdlib_match	FP + FN
#40436	doc.go inventory	doc-sync	low

Detail: #40434 — sprintferrdot verb gate wrong in both directions

fmt invokes Error() for verbs %s %q %v %x %X but not %#v (confirmed against pkg.go.dev/fmt: "valid for a string (%s %q %x %X), or is %v but not %#v"). sprintferrdot only checks {s, v} (sprintferrdot.go:89) and parseSimpleFormatVerbs discards the # flag (:156), so:

• FP: fmt.Sprintf("%#v", err.Error()) is flagged, but %#v uses the Go-syntax representation and does not call Error() — dropping .Error() is behavior-changing. (If ever CI-enforced, this is a build-blocker with a wrong fix.)
• FN: %q/%x/%X with err.Error() are genuinely redundant but skipped.

Detection of .Error() itself is sound (TypesInfo + types.Implements, :191-208). Zero production sites today (latent), but spec-grounded. Fix: track #, widen verb set to {s,q,v,x,X} excluding %#v.

Detail: #40435 — two newest linters re-introduce syntactic package matching

Linter	Site	Check
fileclosenotdeferred	:172	ident.Name != "os"
contextcancelnotdeferred	:138	pkgIdent.Name != "context"

Same defect class as the landed precedents: FP on a shadowing local var (os := ...; os.Open(...)) and FN on alias imports (import osx "os"). astutil.IsPkgSelector exists (astutil.go:75-94) and the newest format linter sprintferrdot:109 already uses it correctly — these two missed it. Distinct from #40243 (scoped to 3 CI-enforced os/fmt/strings holdouts); filing these completes the codebase-wide consolidation and stops the antipattern regressing on every new linter.

Detail: #40436 — doc.go inventory drift

doc.go:3 says "29 active analyzers" but 30 are registered, and the bullet list omits both hardcodedfilepath (main.go:63) and sprintferrdot (main.go:75). Same class as the closed #38787/#38028; recurs whenever a linter lands. Trivial doc-only fix.

Metrics

Metric	Value
Run	R42
Findings	6
Issues filed	3 (non-duplicate)
Success score	9 / 10
Cumulative runs	42
Avg success	8.81

Historical context

Reconcile confirmed R41's sg41a1 → #40243 and sg41a2 → #40244 both open, and #39982 (jsonmarshal ExprStmt FN) still open. Prior fixes (#39733, #39981) remain landed-in-code-but-open — the team does not auto-close, so grepping code is mandatory before re-filing (no phantoms re-filed this run).

Recommendations & next-run focus (R43)

1. Re-grep all 30 linters for ident.Name == "<pkg>" after 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 lands — R42 proved new linters regress the antipattern, so a periodic sweep is warranted.
2. Fresh audit of never-examined non-enforced linters: errorfwrapv, excessivefuncparams, largefunc, errormessage, ossetenvlibrary, seenmapbool, lenstringzero.
3. Investigate whether sprintferrdot should even list Fscanf among format calls (its args are scan-target pointers, not formatted values).
4. pattern_set probes: tolowerequalfold (ToLower+ToUpper?), httpnoctx (all client methods?).
5. Watch for a 31st-linter delta via the main.go-vs-doc.go count.

References: §27860925409

Generated by 🤖 Sergo - Serena Go Expert · 276.8 AIC · ⌖ 11.6 AIC · ⊞ 5.8K · ◷

• expires on Jun 20, 2026, 9:14 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 #40582.