Foundry’s v1.10.0 release notes, published November 20, 2026, describe a change that can affect existing projects: evm2 replaces revm across Forge, Anvil, Cast and Chisel. Before upgrading, check tests and project configuration rather than assuming results will match v1.9.0.
Check existing projects
- Regenerate gas snapshots. The evm2 change can alter gas numbers. Fork warmth and transient storage also no longer leak from
setUp, whilevm.setArbitraryStorageleaves slots cold. - Review tests that expect a revert. A halting call under
vm.expectRevertnow consumes all forwarded gas, as it does on-chain, except when the test makes the call directly as a non-static call withisolate = true. If subsequent test work runs out of gas, the release notes suggest forwarding a fixed amount, for exampletarget.f{gas: 100_000}(). - Check remappings and test discovery. A conflicting dependency remapping now applies only inside that dependency; adding the old mapping to the project’s root remappings restores v1.9-style resolution project-wide. Only suites under
src,testandscriptrun by default.forge testwarns about project suites it skips; widen the test selection if those suites should still run. - Recheck invariant assumptions. Snapshot fuzzing is on by default, so a run may resume from a state reached by an earlier run rather than starting from
setUp. Handlers orafterInvarianthooks that assume a fresh state each run may fail.
The release also adds pinned Solar releases: solc = "solar:0.3.0" selects that release, which Forge downloads if missing. Forge can pass a thread count to Solar, including with forge build -j 4. Other additions include minimized counterexamples for failing fuzz and invariant tests, symbolic equivalence checks, and tooling for Tempo T14 configurable accounts. On Tempo chains where T14 is active, cast wallet new creates a configurable account; use --eoa for a conventional account.
The release notes report forge test performance gains versus v1.9.0 of 4.5× on Morpho Blue, 3.5× on Spark PSM and 2.1× on Aave v4. Those are results for the named workloads, not a guarantee for every project. A practical upgrade check is to rerun your own tests, compare regenerated gas snapshots and inspect any changed remapping, discovery or invariant behavior.

