[sergo] Sergo Report: GoLSP-Outage-Fallback-Manual-Audit - 2026-08-22

[github-actions[bot]]

Executive summary

Run §32550272009 picked up ~45 days after the last recorded run (R60, 2026-07-08). The Serena Go language server was unavailable for the entire run ("go: Go is not installed", despite activate_project reporting go as an active language server), so all symbol/reference/diagnostic tools fell back to manual Grep/Read auditing. Two confirmed, evidence-backed issues were filed. This was reported as a missing_tool gap.

Tool updates

• Serena tool count: 23 → 24 (added find_declaration, find_implementations, get_diagnostics_for_file, open_dashboard, rename_memory, replace_in_files, safe_delete_symbol; removed check_onboarding_performed, find_file, list_dir, think_about_collected_information, think_about_task_adherence, think_about_whether_you_are_done).
• Go LSP broken — all Go-backed semantic tools errored out; worked around with Grep/Read/Bash for the full run. Reported via missing_tool (first attempt used an invalid field name and was rejected with a schema error; corrected and resubmitted successfully).
• Linter registry relocated: cmd/linters/main.go no longer exists. The registry now lives at pkg/linters/registry.go (67 import lines), cross-checked against pkg/linters/doc.go's header ("All 67 active analyzers") — confirmed in sync.

Strategy

Given the LSP outage, the planned 50/50 cached-reuse/new-exploration split was adapted mid-run: cached-reuse effort went into re-checking previously-flagged bug classes (hardcoded diagnostic messages, exact-string-type assertions) against the newly-grown linter set, while new-exploration covered linters not yet individually audited.

Findings

2 confirmed issues filed (click to expand)

1. lenstringzero: diagnostic message hardcodes literal s/len(s) instead of interpolating the real matched expression — every reported violation shows misleading placeholder text (e.g. reports len(s) == 0 for len(username) == 0). This is the third occurrence of this exact bug class in the codebase (previously fixed in bytescomparestring and stringbytesroundtrip). Not yet CI-enforced, so it went unnoticed.
2. stringbytesroundtrip: isExactString is a no-op alias of isStringType — it currently gives the right answer only because of which type snapshot each caller happens to pass in (raw vs. .Underlying()-resolved). This is the exact same gap whose absence caused the writebytestring bug fixed as sg60a1 last run (a shipped, non-compiling autofix). Currently latent since no SuggestedFix uses this branch yet, but flagged before a future contributor adds one.

Reviewed clean, no issue filed

mapdeletecheck, seenmapbool, httprespbodyclose, contextcancelnotdeferred (suppression wiring confirmed correct via the shared resourcetracker abstraction — an initial suspected gap turned out to be a false alarm), ssljson (legitimately filesystem-based, not a per-file AST linter), sprintferrdot, plus a hardcoded-message grep sweep across manualpathconcat, stringsconcatloop, httpstatuscode, packagelevelmutableslicemap, generatedyamlheredoc, errorfwrapv, hardcodedfilepath.

Tasks generated

1. Fix lenstringzero's message to interpolate the real expression via astutil.NodeText, matching the pattern already used correctly by sibling linters.
2. Give stringbytesroundtrip.isExactString its own logic (distinguishing *types.Basic from *types.Named) independent of isStringType, with a named-string-type regression test.

Metrics

• Findings surfaced: 10 (2 filed, 8 reviewed clean)
• Issues created: 2 / 3 max
• Tasks created: 2
• Success score: 8/10 (docked for the LSP outage forcing a slower manual-only audit and for not individually reconciling the ~40+ sergo issues closed during the 45-day gap)

Historical context

• ~45 days since the last recorded run; roughly 40+ sergo-labeled issues closed in that window (not itemized this run — too large to reconcile token-efficiently, targeted gh api search was used per candidate instead).
• Linter registry grew from 43 → 67 analyzers.
• CI enforcement grew from 14 → ~53 enforced linters; 14 remain unenforced: errorfwrapv, excessivefuncparams, generatedyamlheredoc, hardcodedfilepath, httprespbodyclose, httpstatuscode, largefunc, lenstringzero, manualpathconcat, packagelevelmutableslicemap, seenmapbool, sprintferrdot, ssljson, stringsconcatloop.

Recommendations

• Re-verify the Go LSP status at the start of the next run before falling back to manual auditing — if restored, prefer Serena's semantic tools per the mission's instructions.
• Continue working through the not-yet-enforced linter list for either genuine bugs or enforce-readiness sign-off.

Next-run focus

1. Re-verify both filed issues landed.
2. Re-check Go LSP availability early.
3. Audit the remaining untouched not-yet-enforced linters: excessivefuncparams, largefunc, generatedyamlheredoc, hardcodedfilepath, manualpathconcat, packagelevelmutableslicemap, stringsconcatloop.

References:

• §32550272009

Generated by 🤖 Sergo - Serena Go Expert · agent · 275.9 AIC · ⌖ 9.07 AIC · ⊞ 6.8K · ◷

• expires on Aug 22, 2026, 8:16 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-08-23T04:16:04.295Z.

Closed by Workflow