Update EVM Version to Shanghai

[smol-ninja]

I recommend that we should update the evm_version to shanghai to take advantage of the PUSH0 opcode. This evm version is supported by all chains we deploy to.

Context

Currently, we rely on paris version of evm for deployment. Foundry uses the value of evm_version to generate bytecode during on-chain deployments.

Since paris, there have been two mainnet hardforks - shanghai and dencun. While dencun did not introduce anything relevant to us, shanghai introduced PUSH0 - an opcode to place "0" on the stack.

Benefits of PUSH0:

1. PUSH0 consumes 1 less gas compared to PUSH1 0 making deployed bytecode smaller and reducing deployment costs.
2. Cheaper transactions for functions that require placing "0" on the stack.

Support for PUSH0 Opcode Across Chains

I have deployed the following bytecode on every chain supported by Sablier (except Lightlink, I didn't deploy on it). All the deployments were successful which means all of them support PUSH0.

0x6080604052348015600e575f80fd5b50600436106026575f3560e01c80632ee921d714602a575b5f80fd5b5f60405190815260200160405180910390f3fea2646970667358221220721ae7de3b3c2a6157c6cb0c3bd4239cf5cd317e94590094fcee9bde551aa99464736f6c63430008170033

Important

Since zksync has its own variation of foundry, the deployed bytecode on zksync is also different. Therefore, updating the evm version to shanghai will have no impact on zksync deployment.

RFC @sablier-labs/solidity.

[PaulRBerg]

Thanks for the detailed investigation!

Happy to switch to Shanghai. The reason we went with Paris is that when we first deployed V2 in 2023, it wasn't supported by most rollups.

except Lightlink, I didn't deploy on it

Sorry to be a pain, but could you do that, please?

LightLink is a paid customer, and we've promised to host their protocol for at least a year.

[smol-ninja]

Lightlink supports PUSH0 as well. https://phoenix.lightlink.io/tx/0x3f7bda3036770d75db38e54eaf0a640e94100681b3d687bb824bf33268debe9e

[PaulRBerg]

Fantastic, in this case let's proceed with switching to Shanghai.

Note: you might also have to do this in PRBMath and all upstream dependencies.

[smol-ninja]

It's not required for upstream dependencies since this change only relates to the build process. When we compile v2-core, Foundry handles the compilation of all dependencies. Regardless, I will update it in PRBMath too.

[smol-ninja]

Issues:

• Update EVM Version to Shanghai #926
• Update EVM Version to Shanghai v2-periphery#347