Ethereum Developer Warns Glamsterdam Upgrade Could Break Up to 15,000 Contracts, Some Unpatchable


Ethereum Foundation developer Marius van der Wijden told core developers on a Monday, Nov. 16, call that two gas-pricing proposals in the Glamsterdam upgrade could break 10,000 to 15,000 smart contracts, an upper estimate that came without a confirmed inventory. Developers will take the question up again on the All Core Devs execution call Thursday, Nov. 19, with no precommitted decision on the proposed changes. The audits behind the warning tie most breakage to gas costs rather than to new mechanics, which means code that never changes can still stop working, and contracts that cannot be upgraded may have to be redeployed.

The proposals, EIP-8037 and EIP-8038, are meant to let the block gas limit rise from about 60 million to as much as 200 million, roughly 3.3 times current capacity. Several developers argue that removing them would delay Glamsterdam and the post-quantum roadmap that follows by four to six months.

Why Do Higher Gas Prices Break Contracts That Never Changed?

Gas is Ethereum’s unit of account for computation, and a transaction that runs out of gas reverts while the sender still pays for the gas consumed. Some contracts were written around assumptions about what operations cost. The Ethereum Foundation lists fixed 2,300-gas stipends, hardcoded gas values in calls, logic that branches on remaining gas and presigned transactions with fixed gas limits as the typical patterns. Developers call these assumptions invariants, and the Foundation has flagged eight smart contract invariants for builders to check. Raise the price of an operation and the assumption turns false, even though a contract can fail with no code change required.

The size of the repricing explains why. Under EIP-8037, a new storage slot would cost 97,920 units of state gas, up from 20,000, and a new account would cost 183,600, up from 25,000. The specification prices a 24-kilobyte deployment at about 37.8 million gas, up from roughly 4.9 million. EIP-8037 also splits gas into two pools: execution gas and a separate state-gas reservoir, and the opcode that reports remaining gas shows only the execution pool. ChainSecurity’s research found that this can let some existing contracts’ failure paths complete instead of reverting.

That reservoir behavior drove the first alarm. ChainSecurity disclosed 13 deployed mainnet contracts on Oct. 22, the day developers pulled Glamsterdam from the Hoodi testnet.

Audits Blame Price Hikes, Not Reservoir Mechanism, for Most of 200 Affected Contract Classes

ChainSecurity and Dedaub presented AI-assisted audit findings that identified about 200 distinct classes of affected contracts, most tied to higher gas costs rather than the new accounting mechanics. DeFi Prime describes the figure as a combined total: ChainSecurity reported 152 findings, 76 of which the Foundation’s internal search also found. Each finding is a class, so repeated deployments of one implementation do not add to the count. Most concerned invariants 4 and 7, which cover fixed gas budgets and reserves, two of the Foundation’s eight listed invariants.

The 200 classes and the 15,000 contracts measure different things. Van der Wijden did not present a confirmed inventory, and the accounts of the call reviewed here do not explain how he derived 15,000. A Foundation developer identified as Rahul reported at least 25 confirmed medium-severity issues and withheld a larger untriaged count because it included many false positives. ChainSecurity named its own blind spots: dormant execution paths, unverified code and private off-chain systems, and its scope covered Ethereum mainnet only, not layer-two networks.

How much money sits in the affected contracts is unresolved. Van der Wijden loosely put the value at risk in the low hundreds of thousands of dollars among contracts whose owners have not committed to upgrading, yet DeFi Prime reports him saying he had no reliable numbers. ChainSecurity said most findings could be mitigated fairly easily, though some require a contract upgrade.

Immutable Contracts Need Redeployment While Builders Question Composability

An upgrade is not an option for a contract deployed as immutable, because contract code cannot be updated once deployed. ChainSecurity said immutable protocols may need full redeployment. EIP-8037 itself acknowledges deployed contracts cannot be fixed when they meter work by comparing two readings of remaining gas.

That is the heart of the objection from application builders. Jango, who builds the Juicebox protocol, argued the change undermines composability, Ethereum’s promise that applications can safely build on one another, and harms developer confidence in building immutable protocols. On the call, Jango also raised the concern that callers could tune gas to trigger error-handling branches and influence whether composable behavior runs. A Uniswap developer said EIP-8037 should be reconsidered.

Researcher Ansgar Dietrichs pushed the other way, and cautioned against formalizing permanent invariants too early. His caution points to the structural problem: when gas prices are effectively part of an immutable contract’s interface and the protocol reserves the right to reprice, every future increase in the gas limit raises the same compatibility question. A hard fork is, by definition, a protocol change that is not backward compatible and that every participant must adopt. Ethereum has argued over how far such changes should reach into applications before, most famously in 2016, when a hard fork moved The DAO’s funds to a recovery address and the chain that declined it became Ethereum Classic.

Roughly 3.3x Capacity Could Add 387 GiB of State per Year, EIP-8037 Estimates

The repricing exists to keep the network’s stored data in check. EIP-8037 says deployment costs about 200 gas per new byte while new storage slots cost about 313 gas per byte. After the gas limit doubled from 30 million to 60 million, average daily new state rose from about 105 mebibytes to about 326, roughly 116 gibibytes a year. The proposal projects about 387 gibibytes of yearly growth at a 200 million limit, which it says would pass a 650-gibibyte threshold within a year, against a Geth state database of about 390 gibibytes in January 2026. The figure is a projection rather than a measurement, so it depends on how recent growth carries forward.

The Foundation says repricing is a prerequisite for raising the gas limit further, with state-access costs last adjusted in 2021. Its replay of historical mainnet transactions found most were unaffected and most flagged cases could be fixed by raising a gas limit.

Developers Weigh Shipping, Amending or Removing Repricing as Hoodi Stays Without Date

Nothing is precommitted for Thursday. The Foundation says specification revisions are being prototyped, with their benefits and any newly broken assumptions to be presented, and van der Wijden said the alternatives break other assumptions and need fresh investigation. Several developers want to ship as planned. On Nov. 10, Ethlabs recommended waiting for audit clarity before any decision on specification changes, and said none of the reported issues was exploitable at that point.

The calendar leaves little slack. Glamsterdam activated on Sepolia on Oct. 6. Developers pulled the Hoodi activation, set for Oct. 26, on Oct. 22, and Sepolia may need a second fork if the specification changes. For the previous upgrade, Fusaka, the Foundation expected a mainnet date 30 days from the All Core Devs call that follows a successful Hoodi fork, and its schedule listed mainnet as a minimum of 30 days after Hoodi. Hoodi has no date now, so mainnet has none. Nixo said the delay “more than likely” pushes Glamsterdam on mainnet into 2027.

Van der Wijden said the team wanted significant public outreach but timelines slipped amid reorganizations and budget cuts at the foundation, and that he shares a lot of the blame as the last remaining scaling lead there. The Foundation named him co-lead of Layer 1 scaling in August 2025 alongside Dietrichs and Tim Beiko. Dietrichs is also a co-author of EIP-8007, the meta-proposal that catalogs Glamsterdam’s gas repricings.

What Should Contract Owners Do Before Thursday?

Maintainers of layer-one contracts can paste a contract address into the Foundation’s affected-contracts search and test fixes on the Platåberget testnet, also called glam-devnet-8. Wallet, RPC and node-tool builders need to update gas estimation, because cached constants will underestimate costs and cause failed transactions. Teams that find a break can report it privately to glamsterdam@ethereum.org, according to DeFi Prime, with the affected contracts, the broken assumption and a reproducible test if available. Where possible, contracts should propagate call failures so the whole operation reverts, and where continuing is necessary they should keep a recovery path rather than deciding from remaining gas alone.

Ordinary users need no action, according to the Foundation, which says updated wallets and infrastructure will handle the changes. Whether users of specific applications are exposed depends on which contracts break, and Thursday’s call is the next point at which that picture could change.


Frequently Asked Questions

Does Glamsterdam put ETH holders’ funds at risk?

The Foundation says regular users need no action and that updated wallets and infrastructure will handle the changes. The exposure sits with specific applications whose contracts embed fixed gas assumptions, and van der Wijden said he had no reliable numbers on the value inside them.

Could future gas increases break contracts again?

They could, unless the protocol settles which gas assumptions it will honor. Dietrichs argued against permanent invariants too early, while Jango argued the change harms confidence in immutable protocols, and no decision was made on the call.

When could Glamsterdam reach mainnet?

No date exists. Hoodi stays delayed until the issues are resolved, and Nixo expects the delay to push mainnet into 2027.

Is 15,000 contracts a large share of Ethereum?

By count it is small: 15,000 against roughly 90 million deployed contracts is about 0.017% by simple arithmetic, a share van der Wijden characterized as a fringe subset. A raw count says nothing about how much value or how many users depend on those contracts, and the audits count classes of contracts rather than every deployment.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *