Foundry test reverts at PUSH0 with NotActivated and no REVERT opcode: set evm_version = 'shanghai' in foundry.toml

Symptom: a hand-written opcode contract (or any contract compiled for a newer EVM) reverts inside forge test, and in forge test --debug execution stops at PUSH0 without ever reaching a REVERT. The failure surfaces as a NotActivated error from revm.

Cause: the EVM revision the test runs against does not include the opcode. PUSH0 was introduced in Shanghai (EIP-3855); under a Paris/London revision it is not a valid opcode and revm aborts with NotActivated rather than a normal revert. Same class of failure for MCOPY, TSTORE/TLOAD, and BLOBHASH under a pre-Cancun revision.

Fix — set the EVM version explicitly in foundry.toml:

[profile.default] evm_version = 'shanghai' # or 'cancun'/'prague' as needed

Notes: this controls both compilation and the test EVM, so it also matters for handwritten bytecode that solc never sees. If you fork a real chain, match the revision to that chain's fork level, since deploying PUSH0 bytecode to a chain that has not activated Shanghai bricks the contract — a real deployment hazard on lagging L2s and sidechains. Conversely, if you target such a chain, pin the compiler to an older evm_version so solc stops emitting PUSH0.

Why it costs time: there is no revert string, no failing assertion, and the debugger's stop point looks like a stack or memory bug in your assembly rather than a configuration mismatch.

Source: adapted from a Stack Exchange answer by Patrick Collins (question by Patrick Collins), https://ethereum.stackexchange.com/questions/161891/received-notactivated-error-when-debugging-my-foundry-test — CC BY-SA 4.0