igneum/docs/plans/release-0.3.18.md

Release 0.3.18: the app-only clock-sample cut (7 October 2026)

Main's order, 08:1xZ. The shipped engine's clock sample (app/igneum-app/src/update.rs latest_block_time, called from engine.rs every 9 s once the node reports blocks > 0 and peers > 0) POSTs eth_getBlockByNumber ["latest", false] to the node's exec port; a node before the exec RPC fix (every live node: 0.3.17's 5899f603) dies on it whenever its exec follower holds no record (rpc.rs indexes records[0] on an empty vector; the panic hook exits the process), so a fresh install's IBD and a restart whose follower loads slowly crash-loop (ledger N7; PC 1 after the project lead's reopen, 07:2xZ). This cut changes the app only: the sample asks igneum_getExecStatus first (it indexes nothing) and takes no block until executedTipHash is set (exec_has_record, exec_status_has_record, a unit test). The node pin stays at 5899f603; nothing on the fleet, the hands or the seed moves for it.

Numbers (main): this cut is 0.3.18 on the 0.3.17 app tree (branch release-0.3.18 from release-0.3.17 caacd6b9); the feature tree is 0.3.19; decimals 0.3.20. The updater has no per-machine target, so the publish reaches every home miner on its hour (wanted); the order is PC 1's update-now first, on the project lead's word.

Builds

piece	node commit	sha256 (first 8)	bytes	note
app commit		0e6c1248		app tests 163 + 28 + 8, the N7 gate test green
Igneum-Miner-0.3.18.dmg	5899f603	38322026	41,953,206	Mac node f3dcf310 (the 0.3.17 one), prover pair aboard, packaged object = the live sixteen; the engine carries igneum_getExecStatus
Windows inputs	5899f603			igneumd.exe b6b3f417 (the 0.3.17 one), workers 14b6637e / 0cbb65f0 / f7d924a7 (copied into this worktree; the first push had none and its run 37593653864 was cancelled), run 37593893376 dispatched 08:27Z on release-0.3.18 (dispatched under the branch's former name release-0.3.18-app)

Read-back after the apply: app 0.3.18; the node unchanged (commit string 5899f603, digest eada4bda); on PC 1 the node stays up past 60 s after a restart (no rpc.rs:591 line) and the three workers return at their rates.

Windows run 37593893376 green (08:34Z): Igneum-Miner-Setup-0.3.18.exe 3cd253d4, 62633175 bytes. Staged in scratch r0318-app/dlsite-stage: manifests at 0.3.18, mac 38322026, windows 3cd253d4, consensus.override = the live sixteen-field object, HiveOS alias held at 0.3.17 (the node is unchanged; no HiveOS row). Runbook r0318-app/deploy.sh (preflight passes; step_manifest, step_update_now with PC 1 first, step_readback). The live folder holds no 0.3.18 file. Waits on the project lead's word.

The publish (the project lead's word at 10:4x UK: "0.3.18 go go go")

• 08:39Z publish-manifest.sh --public --deploy from the live folder (the same inputs as the scratch copy): token and public manifests at 0.3.18, mac 38322026, windows 3cd253d4, consensus.override = the live sixteen-field object, wallet 0.1.5 unchanged, HiveOS alias held at 0.3.17. the project lead taps update-now on PC 1 first; PC 2 and the Mac take it on their hour or his tap; no fleet, hands or seed action (the node is unchanged).

(the per-machine read-backs as they land)