[sergo] Sergo Report: Two-New-Linter Tool Change + Precision Audit — 2026-06-12

[github-actions[bot]]

Executive Summary

Run R34 detected a tool change: the custom-linter registry grew from 24 → 26 analyzers — hardcodedfilepath (#38742) and timesleepnocontext (#38704) both landed via bot PRs since the last run. This run combined the proven new_linter_doc_lag + syntactic→type-identity cached patterns with a fresh precision audit of the two new linters, and surfaced a notable process finding: last run's recorded issues never actually existed.

• 🔧 Registry: 26 analyzers (was 24). Serena LSP unchanged at 23 tools.
• 📚 Doc divergence: 26 / 25 / 24 across main.go, README+spec_test, and doc.go.
• 🔍 New-linter audit: timesleepnocontext clean; hardcodedfilepath has a latent false-positive in its format-template guard.
• 🧾 3 issues filed (#38787, #38788, #38789), all single-file, all on unenforced linters (zero CI-break risk).
• ⚠️ Reconciliation: R33's claimed sg33a1/sg33a2 issues do not exist — the R33 safe-output never posted. The ctxbackground finding was therefore refiled.

Critical Findings

1. Doc-sync lag (#38787) — cmd/linters/main.go registers 26 analyzers, but pkg/linters/doc.go:3 still says "All 24" with only 24 bullets (missing both new linters), and README.md + spec_test.go documentedAnalyzers() are at 25 (missing timesleepnocontext). timesleepnocontext is therefore not covered by the spec suite despite shipping in the driver.
2. hardcodedfilepath false positive (#38788) — hasFormatVerb (hardcodedfilepath.go:88-93) is a substring check that misses flag/width/precision verbs (%05d, %-10s, %8.2f) and 9 verb letters (%x %t %c %b %o %e %g %p %U). Format-template paths like /tmp/run-%05d/out get flagged as hard-coded paths — directly contradicting ADR Add hardcodedfilepath linter to detect hard-coded file path string literals #38742, which states the linter "skips format-template strings."
3. ctxbackground precision (#38789) — isContextBackgroundCall (ctxbackground.go:86) matches ident.Name == "context" syntactically, while the same file detects the ctx parameter by type identity (types.Identical + import scan). Internal inconsistency ⇒ false positive on a locally-named context and false negative on an aliased import ctxpkg "context". Mirrors the landed sortslice precision: match sort.Slice/SliceStable via package identity (pass.TypesInfo), not the syntactic identifier name "sort [Content truncated due to length] #38029 sortslice fix.

Strategy — 50/50 Split

Cached reuse (proven, fast-landing patterns)

• new_linter_doc_lag (landed lenstringzero (22nd custom linter) is registered in the driver but undocumented — sync doc.go, README, and spec_test to 22 analy [Content truncated due to length] #37740, sortslice (23rd custom linter) is registered in the driver but undocumented — sync doc.go, README, and spec_test to 23 analyzers #38028): applied to the two newly-registered linters → #38787.
• syntactic→type-identity swap (landed sortslice precision: match sort.Slice/SliceStable via package identity (pass.TypesInfo), not the syntactic identifier name "sort [Content truncated due to length] #38029, astutil.IsFmtErrorf): refiled the never-posted R33 ctxbackground finding → #38789.
• Both are single-file changes matching precision_fix_lands (≈1-day landings).

New exploration (audit of the two brand-new linters)

• timesleepnocontext — audited and found clean: uses ObjectOf → *types.PkgName → Imported().Path()=="time" for the call, types.Identical for the ctx param, skips test files, and emits no autofix. No issue filed.
• hardcodedfilepath — found a latent FP in hasFormatVerb; strongest land-argument is that the implementation contradicts the linter's own ADR → #38788.

Tool / Registry Changes

Surface	Count	Notes
cmd/linters/main.go	26	canonical; +hardcodedfilepath, +timesleepnocontext
README.md + spec_test.go	25	missing timesleepnocontext
doc.go	24	missing both
Serena LSP tools	23	unchanged R4–34
CI enforce flags (cgo.yml:1122)	12	unchanged; neither new linter enforced

Generated Tasks

ID	Linter	Type	Effort	Fix
#38787	doc.go / README / spec_test	doc-sync	S	bump to 26, add the two/one missing entries
#38788	hardcodedfilepath	precision FP	S	verb-aware regexp in hasFormatVerb
#38789	ctxbackground	precision	S	resolve call base via *types.PkgName

Metrics

• Findings this run: 6 (3 filed, 1 audited-clean, 2 reconciled/deferred).
• Issues created: 3 (limit 3). Duplicates skipped: only execcommandwithoutcontext precision: false positive on nil-guarded exec.Command fallback — autofix injects a nil context that pa [Content truncated due to length] #38281/execcommandwithoutcontext enforce-readiness: propagate context in connectStdioMCPServer (2 sites), add nolint support, then enfo [Content truncated due to length] #38282 open and both are DO-NOT-REFILE.
• Success score: 9 / 10 (tool change correctly detected, two new linters audited, a process gap reconciled, three distinct evidence-backed single-file findings).
• Aggregates: 34 runs · 257 findings · 68 tasks · avg score 8.75.

Historical Context

Recent runs

• R31 (06-09): 23rd linter sortslice — doc-sync + type-identity precision (sortslice (23rd custom linter) is registered in the driver but undocumented — sync doc.go, README, and spec_test to 23 analyzers #38028/sortslice precision: match sort.Slice/SliceStable via package identity (pass.TypesInfo), not the syntactic identifier name "sort [Content truncated due to length] #38029, landed).
• R32 (06-10): 24th linter execcommandwithoutcontext — nil-guard FP (execcommandwithoutcontext precision: false positive on nil-guarded exec.Command fallback — autofix injects a nil context that pa [Content truncated due to length] #38281) + enforce-readiness (execcommandwithoutcontext enforce-readiness: propagate context in connectStdioMCPServer (2 sites), add nolint support, then enfo [Content truncated due to length] #38282), still open after 2 days (complex control-flow = slow to land).
• R33 (06-11): cross-linter syntactic-match audit — recorded 2 issues that were never actually created (discovered this run).
• R34 (06-12): two-new-linter tool change + precision audit (this report).

Recommendations

1. Land #38787 first — it is pure docs/test and unblocks spec coverage of timesleepnocontext.
2. Treat the phantom-issue reconciliation as a standing check: a recorded "created N issues" in run history is not proof the issues exist — verify by both label and title each run.
3. The two execcommandwithoutcontext issues (execcommandwithoutcontext precision: false positive on nil-guarded exec.Command fallback — autofix injects a nil context that pa [Content truncated due to length] #38281/execcommandwithoutcontext enforce-readiness: propagate context in connectStdioMCPServer (2 sites), add nolint support, then enfo [Content truncated due to length] #38282) have now been open 2+ days; if they stay open ~5+ days, consider a land-rate note rather than re-engaging.

Next-Run Focus (R35)

• Reconcile whether #38787/a2/a3 landed (by label and title).
• Re-verify execcommandwithoutcontext precision: false positive on nil-guarded exec.Command fallback — autofix injects a nil context that pa [Content truncated due to length] #38281/execcommandwithoutcontext enforce-readiness: propagate context in connectStdioMCPServer (2 sites), add nolint support, then enfo [Content truncated due to length] #38282.
• Watch for a 27th linter (apply doc-lag + precision audit).
• Optional deeper hardcodedfilepath audit: unquoteStringLit ignores escape sequences (const-match edge case); isLogOrPrintCall rejects variable-receiver loggers (logger.Infof) despite its own doc example.

References:

• §27396064031

Generated by 🤖 Sergo - Serena Go Expert · 290.8 AIC · ⌖ 12.6 AIC · ⊞ 6.3K · ◷

• expires on Jun 12, 2026, 9:28 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 #39018.