evidence report: Db2 programs stay L1 after a Db2 refresh — the report build doesn't get the Db2 sweep's verdicts #4758

Found 2026-10-09 in the first full (Db2) refresh, bot PR #4756 (after #4740 and #4745).

The db2 job succeeded and its coverage reached the ledger: tests/equivalence/det_sweep_coverage.json has the genapp entries, and the report shows e.g. lgacus01 with paragraphs 3/3 and branches 4/8. But every Db2 program stays L1, because the report's executed-equivalence line still says:

det port not run (a Db2 case: CI's det-sweep skips Db2 cases and no local sweep was given)

So evidence_report.py --refresh --live in the refresh job isn't given the Db2 sweep results (the verdicts). Only det_coverage_ledger.py update receives both sweeps. Result: GenApp shows 0 at L2+ for its Db2 programs, and CBSA's Db2 programs are understated too, even though they're proven.

Fix:

• Pass the Db2 job's sweep.json (verdicts) to the report build alongside the non-Db2 sweep, merged with proof_sweep.py --aggregate or the report's sweep input.
• Or have the report read a case's verdict from the ledger entry the sweep wrote, if the ledger records it.
• On push runs, which skip Db2, keep the last Db2 verdict (as stale, per evidence report: keep a program's last measured level marked stale instead of dropping L3/L4 to zero on any harness change #4730) rather than "not run".

Acceptance:

• After a full refresh, Db2 programs show their measured level, e.g. GenApp lgacus01 at L2+ with its coverage.
• A push refresh doesn't drop them back to L1; they show as stale if inputs changed.
• A test covers the report build with a Db2 sweep input.

Related: #4733 / #4740, #4743 / #4745, #4730 / #4734.

🤖 Generated with Claude Code