[sergo] Sergo Report: STABLE-74 Reconcile-Gap-Catch and Vein-Closure - 2026-10-01

[github-actions[bot]]

Overview

Run 83. The custom-linter registry held steady at 74 analyzers (no new linter since closeerrorunchecked at run 82). The headline result this run is not a new bug class but a catch in the reconcile process itself: two previously-filed issues had quietly flipped to closed not_planned without any prior run individually re-verifying them, because recent reconciles only cross-checked the running open-label list rather than every historically-filed issue number. Both were re-verified as still-present bugs and refiled. New-exploration time went into closing out three open questions carried over from recent runs, all of which resolved clean with no new finding.

Key issues this run

• internal/nolint space-prefixed nolint-colon directive still silently ignored (5th generation of the same bug, first filed as issue 59122)
• blankassigncomma still re-flags 14+ deliberate idiomatic error-discards with no allow-list (3rd generation, first filed as issue 61024)

Both are reverse-phantoms: GitHub auto-expired the prior tracking issues as not_planned, but the underlying code is byte-for-byte unchanged from when each was originally reported.

Tool scan and registry status

Serena: 24 tools, stable, activate_project OK for go/typescript/bash with no approval prompt. Registry: grep count of active Analyzer entries in pkg/linters/registry.go = 74, unchanged since run 82s addition of closeerrorunchecked. No new linter to audit this run.

Strategy split (50/50 cached vs new)

Cached (reconcile): Standard practice queries open sergo-labeled issues and diffs against the prior runs predicted list. That query returned exactly 15 open issues this run, matching the count runs 80-82 predicted, but a direct per-number query on two specific historically-filed issues showed both had actually closed as not_planned on 2026-09-30, a day they were never individually re-checked. This is a blind spot in the open-label-list-only approach: an issue that flips can simply vanish from the list without anyone noticing it was ever there to begin with.

Both were re-read directly in code. pkg/linters/internal/nolint/nolint.go lines 58-60 still strip only the leading two-character comment-slash prefix before checking for the nolint-colon prefix, with no whitespace trim in between, so the common spaced directive spelling is dropped from the suppression index. pkg/linters/blankassigncomma/blankassigncomma.go lines 44-57 still flag any all-blank assignment with no allow-list or comment-aware exception, still re-flagging the io.WriteString discard pattern at pkg/workflow/strings.go lines 172-175 that the codebase already uses, with an explanatory comment, as a workaround for errcheck being disabled in .golangci.yml (line 37, broken exclude-functions in golangci-lint v2).

New-explore: Resolved three carried-over questions, all clean. First, closeerrorunchecked's ident-only fallback path (flagged at run 82 as an unconfirmed promoted-method blind spot) is confirmed dead code - go/types always populates its Selections map for a real method call through a value, including promoted and embedded methods, so the fallback only ever triggers for package-qualified calls where it correctly returns false. Second, swept nolint and filecheck wiring across all 74 linters for gaps: full parity (70 direct call sites plus the shared resourcetracker helper covering 3 linters plus ssljson correctly exempt as a non-AST JSON validator equals 74). Third, swept the comment-overlap autofix-safety guard across all 20 SuggestedFix-producing linters: full parity, including two string-matching linters which get it indirectly via a shared helper function.

Metrics

• Findings: 4 (2 refiled reverse-phantoms plus 2 resolved-clean investigations)
• Tasks generated: 2
• Issues created: 2 (sergo label, both code-verified with concrete file and line evidence)
• Duplicates skipped: 0 (both filings are refiles of auto-expired issues, not duplicates of currently-open ones)

Historical context

This is the 12th and 13th confirmed reverse-phantom overall across all bug lineages tracked by this workflow - GitHub repeatedly auto-expires old issues as not_planned on a schedule, independent of whether the code was actually touched. 83 runs, 511 cumulative findings, average success score 8.75.

Next-run focus

• Re-verify both issues filed this run landed as filed.
• Query every historically-filed sergo issue number directly at reconcile time, not just the current open-label list, to avoid repeating this runs methodology gap.
• If the registry grows to a 75th linter, give it a first-pass audit.
• All established bug-class veins (paren-unwrap gaps, branch-state merges, position-blind snapshots, selector-key collapses, node-filter narrowness, suppression wiring, comment-overlap guards) are now exhaustively mapped or closed - future new-exploration will likely need either new registry growth or a genuinely untried hypothesis.

References:

• Run 36812700179

Generated by 🤖 Sergo - Serena Go Expert · claude · agent · 271.2 AIC · ⌖ 4.26 AIC · ⊞ 5.3K · ◷

• expires on Oct 1, 2026, 8:06 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 #64906.