mirror of
https://github.com/sigp/lighthouse.git
synced 2026-04-30 11:13:34 +00:00
85b6c4d93f
PR #9178 added this Gloas-only test using ApiTester::new() which produces a phase0 chain even under FORK_NAME=gloas, so head.beacon_state.get_ptc(...) errored with IncorrectStateVariant. After switching to new_with_hard_forks() three further issues surfaced: 1. The slot clock is left at head_slot + 1 by the harness setup, so a payload attestation for head_slot fails gossip propagation as a PastSlot. Rewind the clock to head_slot in make_valid_payload_attestation_message. 2. With VALIDATOR_COUNT = 32 and 32 slots/epoch, a slot's committees often hold only a single validator. The PTC for that slot then has one distinct validator regardless of PTCSize. The original test chained JSON and SSZ sub-tests on the same harness with ptc_offset 0 and 1 and asserted both gossip-published, but the second message is a duplicate (same slot/validator) and is silently dropped as PriorPayloadAttestationMessageKnown — so the second recv() hangs forever. Split the SSZ variant into its own test with its own harness so the two don't collide in the ObservedPayloadAttesters cache. 3. Switching the JSON test to new_with_hard_forks() so the chain actually reaches Gloas under FORK_NAME=gloas (same fix as the sibling tests added in #8415 and #9100). Verified locally: full Gloas suite 197/197 passed (350s).