Technologie & IA
Foundry 1.10.0 switches execution engines for Ethereum testing
The Foundry project released version 1.10.0 on November 19. The new evm2 interpreter replaces revm across all four tools in the suite. The release notes call for regenerating gas measurements and flag changes in how tests are discovered and run.

evm2 becomes the common engine across the suite
Forge, Anvil, Cast, and Chisel use the same execution engine in Foundry 1.10.0. The change affects development environments that simulate transactions or execute contracts locally. According to the release notes, the tools that observe execution now follow only the instructions they use. The project also reduced the overhead of some calls, a change that can recur many times during a test suite.
In the suites used by the Foundry team, Forge’s test command ran 4.5 times as fast as version 1.9.0 on Morpho Blue, 3.5 times as fast on Spark PSM, and 2.1 times as fast on Aave v4. These measurements come from the project. The three results show that the gain varies with the code and tests being run. On the 1.46 million-step Aave v4 test, the time until the debugger was ready fell from 1.96 to 1.40 seconds.
Projects need to recalculate their gas measurements
Foundry calls for regenerating gas benchmarks. Gas is the unit that measures the work required to execute an operation. Some test states affecting that cost could persist after setup when a scenario ran on a forked network. The notes say those temporary states now carry over differently. A change in recorded gas can therefore come from the test environment even if the contract itself has stayed the same.
The team also flags a change in how import aliases are resolved. In a conflict, an alias set by a dependency may stop applying elsewhere in the project. Forge also limits default test discovery to the source, test, and script directories. It warns when it skips project suites. Teams comparing two runs can therefore verify that the scope and dependencies match before interpreting a change in runtime or output.
Invariant testing resumes from states already reached
Campaigns that test properties meant to hold through a sequence of operations can now start by default from a state reached in an earlier run. Foundry says this lets them explore more unusual states. Test routines that assumed a fresh initial state for every run may need to be adapted. If a test fails, the tool also tries to shorten the sequence that caused the property to fail.
The release also lets teams pin an exact version of Solar, a Solidity compiler. Forge fetches and verifies the requested version when it is missing. This gives teams an explicit setting for reproducing a build independently of changes to the testing engine. Version 1.10.0 therefore brings several changes that projects can observe separately in builds, test suites, and recorded measurements.
