Skip to content

protocol

Ethereum postpones Glamsterdam Hoodi test after ChainSecurity flags EIP-8037

Core devs pulled the Oct 26 Hoodi activation and the 200M gas-limit test after a ChainSecurity note on EIP-8037/8038 contract-breakage paths. No replacement date.

by 5 min read

Ethereum core devs postponed the Glamsterdam activation on the Hoodi test network, previously scheduled for 17:42:48 UTC on October 26, 2026, after security firm ChainSecurity published an analysis of two of the fork's gas-repricing EIPs — EIP-8037 and EIP-8038 — showing new paths by which contracts that handle out-of-gas sub-calls could revert when they previously succeeded. The decision was taken at Thursday's All Core Devs — Execution meeting, per a CoinDesk report dated October 23, 2026. The planned bump of Hoodi's block gas limit from the current 60M–100M range to 200M was pulled with it, with no replacement date announced.

What happened

Glamsterdam already activated on Sepolia on October 6, 2026 and spent three months on multi-client devnets (devnet-5 and -6) with the full EIP bundle. Hoodi was the next-to-last stop before mainnet. The pulled test would have been the first non-canary network to run with the 200M gas-limit floor the fork is designed around.

The ChainSecurity note, dated October 13, modelled interactions between two of the smaller Glamsterdam items:

  • EIP-8037 — state-growth metering. Introduces a separate "state gas" reservoir for operations that grow the state tree (SSTORE on cold slots, CREATE, EXTCODECOPY on cold accounts). When a transaction's state-gas allowance is exhausted but regular gas remains, state-touching opcodes are priced at a punitive level rather than reverting the transaction. Specification at EIPs.ethereum.org/EIPS/eip-8007 and the implementation module eip8037 in reth.
  • EIP-8038 — cold-access repricing. Adjusts the gas charged for first-touch reads and writes to accounts and storage slots.

Combined, the two change the gas number a sub-call can run out at, and in some cases change which call frame runs out of gas first. Contracts that use a try/catch-style pattern around an inner call — a common shape in router aggregators and some upgradable-proxy admin paths — can now revert at the outer frame when the inner frame used to revert and let the outer frame recover. ChainSecurity gave a worked example: creating a new storage slot once a transaction's state-gas allowance is exhausted costs 110,020 gas against 22,100 gas under current rules, a ~5× jump that could push a well-tuned contract past the gas limit it was deployed with.

ChainSecurity has not publicly shown a production protocol that is exploitable through the pattern, per CoinDesk. The delay is to let client teams add a wider set of integration tests covering the specific try/catch-with-sub-call shapes.

The decision was corroborated by Crypto Briefing's write-up of the ACD call, which notes that the Hoodi activation and the 200M gas-limit test are both on hold until the compatibility review completes.

Numbers

- Hoodi activation (cancelled)  : 2026-10-26 17:42:48 UTC
- Target block gas limit        : 200,000,000 (vs ~60M–100M today)
- Sepolia activation (done)     : 2026-10-06
- ChainSecurity note            : 2026-10-13
- Worked gas-cost delta
  (new state slot after
   state-gas allowance spent)   : 110,020 gas vs 22,100 gas today (~5×)
- Replacement Hoodi date         : none announced

What breaks, and where

The affected pattern is contracts that expect certain inner sub-calls to run out of gas and handle it in Solidity (the try/catch form) or in assembly (checking the return boolean of call). Under EIP-8037, a sub-call that previously consumed, say, 70,000 gas and reverted inside the inner frame can now consume all of the parent's state-gas allowance first, which the inner revert does not reset — producing a different post-call state than the handler was written against.

The two shapes most-cited by client teams on the ACD call:

  • Router aggregators (1inch-style, 0x-style, CoW-style) that iterate over a list of candidate pools, swallow failures, and continue. Under the new rules, a failure can propagate up when the state-gas reservoir has been exhausted elsewhere in the transaction.
  • Upgradable-proxy admin paths that use low-level call to invoke the implementation and fall back on a sentinel return. A sentinel path that worked at 60M gas can revert at 200M because of per-opcode cost shifts in EIP-8038.

Most application code — fixed-flow protocols with a single call graph per transaction — is not affected. The Ethereum Foundation's own August 24, 2026 repricing-testing note said "the large majority of smart contracts are unaffected" after the gas-limit raise; ChainSecurity's new paths are a long tail that the previous test corpus did not stress.

What to watch

  1. Next ACD meeting. The All Core Devs — Execution call on October 30 is the earliest point at which a replacement Hoodi activation could be set. ethereum/pm carries the agenda and notes.
  2. ChainSecurity's full write-up. The October 13 note is the compressed version; a longer paper with the worked gas-cost tables and reference contracts is expected on the firm's blog.
  3. Client-team patches. Geth, Nethermind, Besu, Reth and Erigon on execution; Lighthouse, Prysm, Teku, Lodestar, Nimbus and Grandine on consensus. The relevant pull requests tag glamsterdam.
  4. Mainnet window. H2 2026 was the implied target when Sepolia activated October 6. A Hoodi slip of two to four weeks is survivable; a slip past December pushes mainnet into Q1 2027. The Foundation has not yet re-stated the target.

Context

Glamsterdam's gas-model rewrite is the biggest reshape of Ethereum fee mechanics since EIP-1559. Pectra (May 2025) and Fusaka (December 2025) touched validator mechanics and data availability but left opcode pricing largely untouched. EIP-8037 was always the item with the broadest cross-contract blast radius — it changes an invariant (gas remaining strictly decreases) that some live contracts rely on implicitly. The Hoodi delay is the first materially slipped date in a Glamsterdam schedule that has, until now, mostly held: devnet-5/6 shipped on time, Sepolia activated on October 6 as announced. A thorough compatibility audit before mainnet costs weeks; a mainnet revert because of a missed try/catch pattern costs quarters.

Sources:

Related stories