diff --git a/README.md b/README.md index f79a232..954079d 100644 --- a/README.md +++ b/README.md @@ -199,9 +199,11 @@ The first build of any client compiles the prover and takes several minutes. ```bash git clone --recurse-submodules https://github.com/randprotocol/clients.git cd clients -git checkout v0.8.5 && git submodule update --init # a release, rather than main +git checkout v0.8.3 && git submodule update --init # the newest clients release, rather than main ``` +`v0.8.3` is the newest clients release tag. `v0.8.4` and `v0.8.5` are fullnode tags, not clients tags: `main` vendors fullnode v0.8.5, and there is no clients release built on it yet. + or unpack `rand-wallet--source.tar.gz` from a release, which has the submodule in it already. The submodule is the full node (`core/vendor/fullnode`, at `v0.8.5` `d8efd47`, v0.8.4 plus the audit v12 node fixes): the wallet's cryptography is the node's own code. diff --git a/docs/releases/v0.8.0.md b/docs/releases/v0.8.0.md index af3b399..5be6dd5 100644 --- a/docs/releases/v0.8.0.md +++ b/docs/releases/v0.8.0.md @@ -50,6 +50,8 @@ A device that can make a proof (the desktop app, about 8 GiB) makes it. One that **Status, added after release: a preview, not yet shown to work end to end.** No proposal signed this way has yet been proved and submitted: the multisig authorisation proof that `rand prove` makes for it (tier 18, a large machine) had not been completed on any chain when 0.8.0 shipped, so these apps make shares that verify, but that a spend from a private multisig account goes through is untested. Two faults in this release: in the extension, the desktop app and the web wallet a real signer profile could not be opened again after the next lock, and no app checked the proposal's fee against the payments. A later release fixes both in those three apps; the fee check is not yet in iOS and Android. Do not rely on private multisig for anything that matters until a release says it has been proved. **Correction, 4 November 2026:** the sentence "the fee check is not yet in iOS and Android" above is wrong. The iOS and Android signers got the same fee check on 27 October 2026 (commit 65aab20), and it ships in 0.8.2 together with the fix in the other three apps (see the 0.8.2 notes). The rest of the status stands: private multisig authorisation is still a preview, no multisig authorisation proof has yet been completed and submitted, and it is not yet shown to work end to end. + + **Note, 8 November 2026:** on 6 November 2026 a Production multisig authorisation (msauth) proof was completed (M=3, tier 18): 3.7 MB, 69 minutes to prove, peak RSS 92.9 GiB, so proving it needs a host with at least 128 GB of memory, not the 76 GB stated earlier. The logs are kept in the project's private records. That proof has not yet been submitted on chain 21 by a wallet flow, so private multisig is still a preview and is still not shown to work end to end. - **Fullnode v0.8.0's RPC reads.** The wallet's RPC method table and reply validators cover the seven reads v0.8.0 added (perps, multisig and the RAND price among them), and tolerate the new `rand_getLimits` sections on chain 21 or their absence. - **The wallet against the real node.** A new end-to-end suite runs the real wasm core and engine, with real `fetch`, against a `rand-node` from the fullnode v0.8.0 release on a Test-profile chain it cuts itself: chain identity, the whole client-side RPC contract, faucet → scan → balance → activity → rescan, a proved spend through a real `rand-prover`, failover and submit rules, and a node restart. It is the `real-node` CI job on every push. - Coverage: wallet-ffi and wallet-wasm fully covered, the core's refusals, the ui's pure-logic suites (core 93.2 % → 95.7 %).