multica-ai/multica · error

scan dry-run summary: %w

Error message

scan dry-run summary: %w

What it means

Returned when rows.Scan fails while reading one dry-run summary row. The query returns nine columns (workspace_id, date_utc, rows, input_before, input_after, overcount, clamped_rows, min/max created), and a scan mismatch — wrong column count/type expectation or a NULL in a non-nullable scan target — produces this. In practice this indicates the binary and the SQL schema are out of sync, not bad user input.

Source

Thrown at server/cmd/backfill_codex_usage_cache/main.go:221

	}
	defer rows.Close()

	var summaries []summaryRow
	var total totals
	for rows.Next() {
		var row summaryRow
		if err := rows.Scan(
			&row.WorkspaceID,
			&row.DateUTC,
			&row.Rows,
			&row.InputBefore,
			&row.InputAfter,
			&row.Overcount,
			&row.ClampedRows,
			&row.MinCreatedUTC,
			&row.MaxCreatedUTC,
		); err != nil {
			return nil, totals{}, fmt.Errorf("scan dry-run summary: %w", err)
		}
		summaries = append(summaries, row)
		total.Rows += row.Rows
		total.InputBefore += row.InputBefore
		total.InputAfter += row.InputAfter
		total.Overcount += row.Overcount
		total.ClampedRows += row.ClampedRows
	}
	if err := rows.Err(); err != nil {
		return nil, totals{}, fmt.Errorf("iterate dry-run summary: %w", err)
	}
	return summaries, total, nil
}

func logSummary(cfg config, rows []summaryRow, total totals) {
	slog.Info("Codex usage cache backfill candidate total",
		"execute", cfg.execute,
		"cutoff", cfg.cutoff.Format(time.RFC3339),

View on GitHub (pinned to 2c0912b6ec)

Solutions

1. Rebuild/redeploy the backfill binary from the same commit as the database schema (make build within server/).
2. Run migrations so schema and binary match, then re-run the dry-run.
3. Inspect the wrapped error — pgx names the offending column/type when it is a conversion failure.

Example fix

# before
./old-backfill_codex_usage_cache --cutoff ...  # scan dry-run summary: ... (columns moved by migration)

# after
git checkout <schema-matching-commit> && make -C server build
./server/bin/backfill_codex_usage_cache --cutoff 2026-06-01T12:00:00Z

# Ensure binary and schema come from the same commit
git -C server log -1 --format=%H > /tmp/backfill.build
git checkout $(cat /tmp/backfill.build) && make -C server build

When it happens

Trigger: Running a backfill binary built against a schema where the summary query returns a different column set than the compiled Scan expects; NULLs appearing in columns scanned into non-pointer types.

Common situations: Schema drifted after the binary was compiled (a migration added/renamed summary columns); using an old binary against an upgraded database or vice versa.

Related errors

• connect to database: %w
• ping database: %w
• acquire advisory-lock connection: %w
• acquire advisory lock %d: %w
• rebuild hourly rollup for update window %s..%s: %w

AI-assisted analysis of multica-ai/multica@2c0912b6ec (2026-08-15). Data as JSON: /api/errors/e1972fe481a1d0ca. Report an issue: GitHub.