bench log: Devnet 2 zero program-id class closed (paidSegments 4 at 22:28Z)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
parent
6b53ed261f
commit
fd1aedaad4
1 changed files with 13 additions and 0 deletions
|
|
@ -1975,3 +1975,16 @@ Consequence per tier: the rule costs an honest node nothing on the hot path (a p
|
|||
cache hit); a cold proof costs a 2019 core 0.4 to 0.5 s and a current one 0.2 s, in parallel per record; a forger
|
||||
loses its block and its peer. Until 0.3.17.1's finality-horizon skip, a fresh joiner fetches every carried proof in
|
||||
its IBD window from its syncer (1.27 MB a record), so a syncer must still hold them.
|
||||
|
||||
### 6 October 2026, 22:10Z to 22:28Z, Devnet 2: the zero program-id class and its close (the fleet's reads)
|
||||
|
||||
Devnet 2 (igneum-devnet-2, five nodes on 4c6b129d, 1 block/s, every switch at DAA 0): the nodes ran without a verifier
|
||||
host or IGNEUM_PROOF_PROGRAM_IDS, so their expected segment statements carried zero program ids at hex offset 472 and
|
||||
every segment record from the shipped host was refused while 8 of 8 shards passed (seg 2868: chain proof 1.27 MB in
|
||||
281 s on a 4090). Fix on 4c6b129d: IGNEUM_PROOF_PROGRAM_IDS=0x2b1a81cb...,0x474678f3... on each node process, restart
|
||||
22:10Z to 22:16Z, same genesis. Result: paidSegments 2 on every node at 22:22:28Z and 4 at 22:28:10Z (two provers,
|
||||
8-block segments), 0 PoW rejected, one exec state root on all five at height 3,792 (0xab19a0be...), one version
|
||||
(2.1.0-1279a1d6, digest 4a0b8726). On the 0.3.17 node (fork a344fea3) the override object's proving_shard_program_id
|
||||
and proving_aggregator_id are the statement's ids when set, so the env is not needed. Consequence per tier: a node
|
||||
without a host (a pool hand, a seed, a home node that does not prove) must carry the ids in its object or its env, or
|
||||
it refuses every segment record and pays no aggregator; the 0.3.17 object carries them.
|
||||
|
|
|
|||
Loading…
Reference in a new issue