Explicity allow the use of CMake FetchContent in Solidity formula

[r0qs]

• Have you followed the guidelines for contributing?
• Have you ensured that your commits follow the commit style guide?
• Have you checked that there aren't other open pull requests for the same formula update/change?
• Have you built your formula locally with HOMEBREW_NO_INSTALL_FROM_API=1 brew install --build-from-source <formula>, where <formula> is the name of the formula you're submitting?
• Is your test running fine brew test <formula>, where <formula> is the name of the formula you're submitting?
• Does your build pass brew audit --strict <formula> (after doing HOMEBREW_NO_INSTALL_FROM_API=1 brew install --build-from-source <formula>)? If this is a new formula, does it pass brew audit --new <formula>?

---

Since Homebrew/brew#17075, CMake FetchContent is disabled by default . This made our formula to crash: https://github.com/Homebrew/homebrew-core/actions/runs/9175554982/job/25228895627?pr=172324.

This PR fix that by explicitly enabling it back.

See: Homebrew/brew#17187 (comment) for context.

[SMillerDev]

What does it need to download?

[r0qs]

What does it need to download?

Build dependencies like: fmtlib (https://github.com/ethereum/solidity/blob/8a97fa7a1db1ec509221ead6fea6802c684ee887/CMakeLists.txt#L53)

[carlocab]

This should use Homebrew fmt. https://docs.brew.sh/Acceptable-Formulae#stuff-that-requires-vendored-versions-of-homebrew-formulae

[SMillerDev]

It seems this is breaking for the exact reason we introduced this change. Upstream is vendoring formulae. Please adjust the formula to use the Homebrew versions.

[github-actions]

Thanks for contributing to Homebrew! 🎉 It looks like you're having trouble with a CI failure. See our contribution guide for help. You may be most interested in the section on dealing with CI failures. You can find the CI logs in the Checks tab of your pull request.

[r0qs]

It seems this is breaking for the exact reason we introduced this change. Upstream is vendoring formulae. Please adjust the formula to use the Homebrew versions.

Thanks for the information. But sadly I believe we cannot do that right now. The issue is that the current version of the Solidity compiler installed by this formula (i.e. 0.8.25) uses fetch content unconditionally, so we need to have the option -DHOMEBREW_ALLOW_FETCHCONTENT=ON here.

The mechanism to skip or not system libraries was only added in the recent Solidity release (i.e. 0.8.26, as you can see here: ethereum/solidity@b60ed2b). However the PR that bump the formula version is failing because of the fetch content restriction: #172324).

So I guess we would need to first just add the -DHOMEBREW_ALLOW_FETCHCONTENT=ON here and then we can switch to install the dependencies from homebrew in #172324, since that release also changed one dependency, i.e. we replaced jsoncpp by nlohmann-json. Or we can do all in the PR that bump solidity version, and close this one here, what do you prefer?

[SMillerDev]

The mechanism to skip or not system libraries was only added in the recent Solidity release (i.e. 0.8.26, as you can see here: ethereum/solidity@b60ed2b). However the PR that bump the formula version is failing because of the fetch content restriction: #172324).

Why not just make a PR that bumps the version and uses the Homebrew dependencies in the process?

[r0qs]

The mechanism to skip or not system libraries was only added in the recent Solidity release (i.e. 0.8.26, as you can see here: ethereum/solidity@b60ed2b). However the PR that bump the formula version is failing because of the fetch content restriction: #172324).

Why not just make a PR that bumps the version and uses the Homebrew dependencies in the process?

Sure, we can try to do that in #172324, so I will be closing this PR.

Does homebrew allow us to use specific versions of those dependencies? Sorry to ask that, but I'm not a macOS user and I don't use homebrew other than in our CI, so I'm not sure if I can specify a formula version with the depends_on, can I?

I'm asking this because we rely on specific version of those dependencies, for instance:

• fmtlib version 9.1.0
• nlohmann-json version 3.11.3
• range-v3 version 0.12.0

For instance, doing the following does not work:

  depends_on "fmt@9.1.0" => :build

Warning: No available formula with the name "fmt@9.1.0" (dependency of solidity).
==> Searching for similarly named formulae...
Error: No formulae found for fmt@9.1.0.

But the problem seems to be that such version is not available in homebrew repo, but if this is the case, how would we be able to use homebrew dependencies if we depend on some previous versions?

[SMillerDev]

No, the reasoning why that is unsupported is included in the docs for it: https://docs.brew.sh/Acceptable-Formulae#stuff-that-requires-vendored-versions-of-homebrew-formulae

[matheusaaguiar]

Isn't that too restrictive? I mean, sometimes it is difficult and time consuming to update every dependency every time there is a new version. Sometimes, there is third-party functionality that was deprecated and cannot be readily replaced/re-worked. Other times, we just want to use more stable/audited software and that means fixing a specific version.
Why not implement what was suggested here by CMake maintainer?
Also, from the docs, Increasingly, though: this can be (too) hard. Homebrew’s primary mission is to be useful rather than ideologically pure. If we cannot package something without using vendored upstream versions: so be it; better to have packaged software in Homebrew than not.

Don't get me wrong, not trying to be confrontational, just raising some questions/arguments.
Thanks.

[carlocab]

Why not implement what was suggested here by CMake maintainer?

We have implemented it. See:

Homebrew/brew#17075 (comment)
Homebrew/brew#17310

[matheusaaguiar]

We have implemented it. See:

Homebrew/brew#17075 (comment) Homebrew/brew#17310

Ah, my bad, I didn't see that one.