Tests failing due to foundry recently disabling expectRevert for internal functions

[r0qs]

What version of PRBMath are you using?

v4.1.0

What version of Solidity are you using?

0.8.29-8c54f4d

Describe the bug

We noticed in our external tests (https://app.circleci.com/pipelines/github/ethereum/solidity/37890/workflows/821e4683-ca57-48b4-a751-8246720aa1db/jobs/1740901/parallel-runs/0/steps/0-108) that some PRB-Math tests started failing with the latest versions of Foundry after foundry-rs/foundry#9537 was merged. The PR disables vm.expectRevert for internal functions by default, causing tests like test_RevertWhen_Positive_GtMaxPermitted to fail with the following error:

[FAIL: call didn't revert at a lower depth than cheatcode call depth] test_RevertWhen_Positive_GtMaxPermitted() (gas: 4033)

The previous behavior can be restored by enabling the allow_internal_expect_revert setting in foundry.toml (false by default):

[default]
allow_internal_expect_revert = true

Or by adding the following comment to affected tests (which is around 92 of them - I didn't check them all though):

/// forge-config: default.allow_internal_expect_revert = true
function test_RevertWhen_Positive_GtMaxPermitted() external whenNotZero {
    SD59x18 x = EXP_MAX_INPUT + sd(1);
    vm.expectRevert(abi.encodeWithSelector(PRBMath_SD59x18_Exp_InputTooBig.selector, x));
    exp(x);
}

[PaulRBerg]

Thanks for the report @r0qs. Can you please confirm if the suggestion is to set allow_internal_expect_revert to true in the PRBMath foundry.toml, or in the user's own configuration?

cc @andreivladbrg @smol-ninja — we will likely face the same issue in the Sablier repos when the stable version of Foundry gets this change.