feat(cheatcodes): add vm.getFoundryVersion()

[leovct]

Motivation

Closes #8036

Implement a new cheatcode to get the foundry version: vm.getFoundryVersion.

Solution

The current versioning scheme incorporates the cargo version, the git SHA and the UTC timestamp of the build to simplify differentiating by version number until v1 is released.

The version format follows this structure: <cargo_version>+<git_sha>+<build_timestamp>.

For example, a typical version might look like this: 0.2.0+faa94c384+202407110019.

Test

Install forge locally.

cargo install --path ./crates/forge --profile release --force --locked
alias forgedev="$(pwd)/target/release/forge" 

Run GetFoundryVersion test.

$ pushd testdata
$ forgedev test --match-contract GetFoundryVersion
[⠊] Compiling...
[⠢] Compiling 1 files with Solc 0.8.18
[⠆] Solc 0.8.18 finished in 101.72ms
Compiler run successful!

Ran 1 test for default/cheats/GetFoundryVersion.t.sol:GetFoundryVersionTest
[PASS] testGetFoundryVersion() (gas: 12955)
Suite result: ok. 1 passed; 0 failed; 0 skipped; finished in 425.29µs (111.42µs CPU time)

Ran 1 test suite in 148.77ms (425.29µs CPU time): 1 tests passed, 0 failed, 0 skipped (1 total tests)

[mattsse]

cool, this should work because the vergen env var is set when the forge binary is compiled

doc nit

[mds1]

Do linux/mac/windows binaries all have the same UTC timestamp? I think they are built separately in CI and so likely will have the same date but different timestamps.

The failure mode here is I see my foundry version is 202407110019 and assert >= 202407110019, but the linux version is 202407110018 and fails that check. I think it's ok if we just document this and suggest only comparing against >= YYYYMMDD0000, i.e. just zero out the timestamps

[leovct]

Do linux/mac/windows binaries all have the same UTC timestamp? I think they are built separately in CI and so likely will have the same date but different timestamps.

The failure mode here is I see my foundry version is 202407110019 and assert >= 202407110019, but the linux version is 202407110018 and fails that check. I think it's ok if we just document this and suggest only comparing against >= YYYYMMDD0000, i.e. just zero out the timestamps

We can also format the build timestamp to %Y%M%D (e.g. 20240711)?

[mds1]

I thought about this but sometimes two releases are put out in a day, e.g. when there is an import or urgent fix. I think the current approach is fine as long as we document it. Thanks for this PR! :)

[mattsse]

lgtm, pending @DaniPopes re vergen dep

@leovct do you want to submit the new cheatcode to forge-std? foundry-rs/forge-std#589 nvm this is updated automatically

[mattsse]

I assume this is fine, given that we also use this in other crates and the build overhead should be minimal