diff --git a/docs/plans/release-0.3.21.md b/docs/plans/release-0.3.21.md index f29e8945f..8aa5b4c2f 100644 --- a/docs/plans/release-0.3.21.md +++ b/docs/plans/release-0.3.21.md @@ -70,3 +70,5 @@ sha256 279b1b690e854fc9, the string read back, pairing 8c728ca3 at byte 5. The d On 55768f88 in the final order, each green and pushed with the exec suite between: c631c64b → 7ee808d6, 52e96c94 → 78f0e771, f067f7c1 → be60803a, b0444f51 → 4362119b, 437f0438 → 632437a0, 2e32d5f6 → f2caa145 (the DST_ADDRESS both-add resolved), f95178a1 cherry-picked → 31407f27 (a merge would have pulled the whole genesis-forward branch; the consensus lib 120), d8bceca5 → 96161037 (the exec suite 36). Byte 5, igneum-pow 8c728ca3. On the tip: consensus-core 126, the consensus crate whole green (the lib 120 and both integration targets). The tip's set stopped twice on the boxes' inputs, not the code: the 60x file test wanted the two new switches (pool split, peer directory) in the fast-time file (added as null); the miner's pool tests read the proto-cuda packs four directories above the crate, absent on the boxes and then the 6 October stream where the node pairs 8c728ca3's amended one (the amended packs placed on both boxes). Two rules for the plan: after every merge that touches OverrideParams the 60x file's field diff is re-run; the miner's pool tests want the packs of the paired igneum-pow's stream beside the fork on every box. The set re-runs from the miner suite (miner, kaspa-pow, the three checks), then the build on build-1 and the digest pair (the previous object eada4bda, the floor-moved 4bbbe816); the binary about 17:15 BST with the node lane's two gates from that minute and the fleet's set on the warm pods. **96161037's two node gates PASS on its own binary (the node lane; f99340b3, the string read back):** the digest gate 15:59:54 to 16:01:32Z (thirteen fields a89be8a7 on both binaries with the peers right, the sixteen-field object db9a85f9 refused, the live-object and floor-file digests eada4bda and 4bbbe816 on the binary); the mixed-version gate 16:01:53 to 16:12:05Z (digest b0afb2ee on all five, 252 new and 352 old blocks accepted, 0 rejected, plain header version 2, counts equal at 316, 493 and 604 through both joins and the restart step, no panic). The node side of the first candidate is complete; the fleet's set on the same binary (the kept start with the ids gate on w-target, the wipe on c22-1, the cases on the warm set, the 12 GB line on p12-vast, N15's two live restarts after these gates) is what remains for the pin. The prover pair stands for 0.3.21 (the build-server lane: 96161037:igneum/evm-types is tree fe4145b9, the same as c4459193's and 5899f603's and the exec pin's b7fca5a0, so 71bc2438 / 263bf4ce pairs byte for byte). The shipper's 0.3.21 builds on 96161037 running since 17:01 BST (box: native, hive, the Windows cross of node and app, the app's Linux binary; Mac: the node and the DMG under the lock from 17:03); the build-server lane's hands and seed pairs under /srv/artefacts/0321-96161037/. + +**The reliability injector on 0.3.21's own binaries (the reliability lane; RunPod 3070, 15:44 to 16:13Z): nine steps PASS.** kept-datadir (MF-8): the previous release's node wrote 151 blocks and stopped in 1.0 s, the new node opened the kept datadir and read synced 4.1 s after the engine started, one start, no exit line; catch-up (MF-1/2): no worker for 120 s while the execution layer held no record, the worker up 18 s after the first blocks; card-appears (MF-3): listed 58 s, mining 69 s, removed 49 s after it left, back 72 s; own-restart: no double restart; zero-ladder (MF-2): rungs 10, 30, 120 s; no-status: restart at 93 s; node-silent: node restarted at 151 s, synced 3 s later, mining at 170 s; one-card-fails (MF-4): two cards mining, the third held 30 min with the reason, 2 exports with 2 reuses; orphan-miner (MF-7): the stray killed in 14 s; 9 FAULT lines to the intake. Binaries: the app 786c3d37 from 8ed08dcf (the engine unchanged since 4a28eb59 through every later merge), node c4459193, the 0.3.17-line igneumd as the previous release. miner-reliability-21's register table and a4d557f9 merged (no engine change). The register's 0.3.21 gate list: kept-datadir start PASS, shutdown under 1 s PASS, prover server matches the card (the app's check; the per-architecture kit owed), LG-4 not run (no Windows box).