expectRevert cheat code doesn't work for internal calls

[cruzdanilo]

Component

Forge

Have you ensured that all of these are up to date?

• Foundry
• Foundryup

What version of Foundry are you on?

0.1.0 (39b6e39 2022-03-08T00:08:31.449073+00:00)

What command(s) is the bug in?

forge test

Operating System

macOS (M1)

Describe the bug

when trying to test internal calls (like libraries), expectRevert cheat code doesn't work.

example code:

pragma solidity ^0.8.4;

import { DSTest } from "ds-test/test.sol";
import { CheatCodes } from "./utils/CheatCodes.sol";

library Library {
    function willRevert() internal pure {
        revert("reason");
    }
}

contract RevertTest is DSTest {
    CheatCodes private constant VM = CheatCodes(HEVM_ADDRESS);

    function testExternal() external {
        VM.expectRevert("reason");
        this.willRevert();
    }

    function testInternal() external {
        VM.expectRevert("reason");
        willRevert();
    }

    function testLibrary() external {
        VM.expectRevert("reason");
        Library.willRevert();
    }

    function willRevert() public pure {
        revert("reason");
    }
}

forge test output:

Running 3 tests for RevertTest.json:RevertTest
[PASS] testExternal() (gas: 3762)
Traces:

  [3762] RevertTest::testExternal()
    ├─ [0] VM::expectRevert(reason)
    │   └─ ← ()
    ├─ [326] RevertTest::willRevert()
    │   └─ ← "reason"
    └─ ← ()

[FAIL. Reason: reason] testInternal() (gas: 3299)
Traces:

  [3299] RevertTest::testInternal()
    ├─ [0] VM::expectRevert(reason)
    │   └─ ← ()
    └─ ← "reason"

[FAIL. Reason: reason] testLibrary() (gas: 3233)
Traces:

  [3233] RevertTest::testLibrary()
    ├─ [0] VM::expectRevert(reason)
    │   └─ ← ()
    └─ ← "reason"


Failed tests:
[FAIL. Reason: reason] testInternal() (gas: 3299)
[FAIL. Reason: reason] testLibrary() (gas: 3233)

[gakonst]

Yep- this is because the vm.expectRevert call hooks on CALL, when we need it to hook on jumps etc. @onbjerg and I talked about this juts now, allowing this to work would require some deeper hooking into solc internals

[onbjerg]

Ref #432

[onbjerg]

Assigning myself to this as I want to work on it, but it is going to be a while until it lands since it also requires some changes in ethers-solc.

[onbjerg]

Tentatively added to stable milestone but it may prove infeasible in which case I will report back here

[mds1]

Closing in favor of generic JUMP support for a range of cheats, that will be tracked in #4439