Ivan

We’re bringing ethrex to @EFDevcon. We’d like to exchange ideas with other teams, hear what you’re working on, and connect with people using ethrex or interested in trying it. DM us to coordinate a meetup.

No ethrex Community Call tomorrow. We’ll share the date for our next call soon. In the meantime, we have some updates coming your way.

@SerenitaIO was one of our early adopters, and we're grateful for that support. Great to connect with @serenita_luca and hear more about Vero. Looking forward to continuing our work together!

Great to see ethrex proving Sepolia blocks side by side with @openvm_org. Ethereum’s move toward execution proofs should preserve client diversity from day one. Thanks @ignaciohagopia, @TauLepton_, and everyone bringing EIP-8025 to life!

ethrex v30.0.1 unschedules Glamsterdam on Hoodi, after the ACDE call unscheduled it yesterday. If you run a Hoodi node on v30.0.0, or on one of its release candidates, you need this release before 26 October at 17:42:48 UTC. That is the time v30.0.0 still has the fork set for, and a node that activates it alone splits off from the rest of Hoodi. L2 operators running a sequencer or a follower should upgrade too. Mainnet and Sepolia are unchanged, so operators there who do not run an L2 or the stateless validator can stay on v30.0.0. ⚠️ The stateless-validator guests move to ERE v0.19.5 (SP1 v6.9.0, ZisK v1.3.2-alpha), and the ZisK guest's verification key changes. Anyone who verifies ethrex's stateless-validator proofs, such as zkEVM integrations, has to move to this release's keys together. Node operators who do not use the stateless validator are unaffected. The rest of the release is mostly L2 reliability. A follower whose head was past the last sealed batch used to stop importing blocks when the sequencer or the follower restarted, and only wiping its data brought it back; it now resumes from its own head. A node holding more than one connection to the sequencer could stall mid-sync with RPC and logging frozen, so blocks are now imported in order by a single importer per node. Catch-up is faster as well: the sequencer sends a reconnecting peer only the blocks it is missing and up to 16 batches at a time, and followers no longer commit trie layers to disk after every imported block. Also in v30.0.1: - Stateless validation: witnesses with trie paths that do not end at the hashed key length are rejected, and zero-amount withdrawals read their recipient, as the specification requires. The check runs while the witness tries are rebuilt, so it adds little proving cost. - Block building: withdrawals are applied before the request system calls, the order the specification and block import already use. Only chains with custom request predeploy code were affected, not mainnet or the public testnets. - Networking: node records are encoded without base64 padding, so clients with strict parsers can bootstrap from an ethrex node. If you run ethrex on Hoodi, upgrade before 26 October. No resync is needed.