On statelessnet we observed that a configuration issue caused a hard-to-debug problem with chunk misses that is caused by num_total_parts being smaller than the number of block producers. This change prevents this configuration error from happening by panicking the node if the local client's protocol version would cause such a violation of invariant. A potentially more complete fix is to add epoch_id as an argument to num_total_parts, but we don't really need that right now and this fix is simpler.
On statelessnet we observed that a configuration issue caused a hard-to-debug problem with chunk misses that is caused by
num_total_parts
being smaller than the number of block producers. This change prevents this configuration error from happening by panicking the node if the local client's protocol version would cause such a violation of invariant. A potentially more complete fix is to addepoch_id
as an argument tonum_total_parts
, but we don't really need that right now and this fix is simpler.