livecheck: really use referenced livecheck options

[samford]

---

• Have you followed our Contributing guidelines?
• Have you checked for other open Pull Requests for the same change?
• Have you explained what your changes do? Performance claims (e.g. "this is faster") must include brew benchmark results.
• Have you explained why you'd like these changes included, not just what they do?
• For bug fixes, have you given step-by-step brew commands to reproduce the bug?
• Have you written new tests (excluding integration tests)? Here's an example.
• Have you successfully run brew lgtm (style, typechecking and tests) locally?

---

• I did not use AI/LLM to create this PR, or I disclosed the tool/model below and reviewed its output; I did not attribute commits to AI and will answer maintainer questions and review comments myself without AI/LLM.

I implemented this fix by hand and confirmed it through manual testing. Afterward, I used Claude Code (Opus 5 high) to review the change and implement tests to cover the modified line.

The produced tests covered the primary scenarios involved in the fix but not everything. The tests exercised the code in a basic way (to produce coverage) but didn't check debug or [verbose] JSON output to confirm that the related code works as expected end-to-end. I reworked those tests until I was satisfied with them and added more tests to better test the fix.

I then expanded the tests to provide 100% coverage for all the logic in latest_version that relates to referenced livecheck blocks and used Claude Code to review the final changes. Overall, AI was helpful for reviewing but I could have implemented the tests entirely by hand from the start (though having AI produce a starting point saved me some typing).

---

When I introduced Livecheck::Options I added some logic in Livecheck::latest_version that was intended to fall back to a referenced package's options but that will never happen, as livecheck.options is never nil. Even if we used livecheck.options.presence, it wouldn't work as expected because we want to combine options from the initial livecheck block with the referenced check (including overriding an option that they both set). Basically, we should start with the referenced livecheck block options (if any) and then merge options from the initial livecheck block (so we can override, if necessary).

This was my original idea of how this should work (and part of the reason why I added Options#merge) but I didn't catch this oversight until I was testing a referenced livecheck block recently. This will become a little trickier if we add support for multi-level references (i.e., A references B which references C and options can be set in each) but this is fairly straightforward for now.

This addresses the issue by reworking how we handle referenced options to ensure we merge when a reference is used in a livecheck block. This also adds tests that fail without this fix as well as other tests that provide 100% coverage for reference-related code in latest_version.

[If there's any question, the T.absurd(ref_formula_or_cask) # simplecov:disable else branches only serve the purpose of getting those case statements to 100% coverage (which isn't possible without an else branch). This is something you'll see again in an upcoming PR where I'm expanding livecheck test coverage further.]

---

You can reproduce this issue by running brew livecheck --debug tableau (which outputs "URL Options: {user_agent: :browser}") and then compare the output to brew livecheck --debug tableau-prep (which omits the user agent option). With the changes in this PR, the user agent option from the referenced check is also present for tableau-prep, as intended. [The tableau-prep check intermittently works without the :browser user agent but this will become a more notable issue when we deprecate the user agent fallback logic in the near future, as the check would fail in that scenario.]

[Copilot]

🟡 Changes recommended

Conflicting POST options can cause a referenced value to override the initial block unexpectedly.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Fixes inheritance and overriding of Livecheck URL options from referenced formulae and casks.

Changes:

• Merges referenced and local Livecheck options.
• Adds exhaustive reference handling and regression coverage.

File summaries

File	Description
Library/Homebrew/livecheck/livecheck.rb	Merges referenced options and handles exhaustive reference types.
Library/Homebrew/test/livecheck/livecheck_spec.rb	Tests option inheritance, overrides, and default checks.

Review details

• Files reviewed: 2/2 changed files
• Comments generated: 1
• Review effort level: Balanced

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[samford]

Hmm, something like this would be a suitable workaround but this also reminds me that we can't set a nil option value in the initial livecheck block to unset an option that's inherited from a referenced livecheck block. In this context, livecheck.option and referenced_livecheck.option would both be Option objects and #merge calls #to_h internally, which calls #to_hash and that method filters out nil values. This is a quirk of how Sorbet has implemented #to_hash for T::Struct (iirc, something to do with what Stripe's code requires), so I may need to manually implement that method to work around this limitation.

However, that's something for a follow-up PR (we only have two livecheck blocks that use a reference with an override and both only use regex). I'll update this commit to include a related workaround for the flagged issue in a moment.

Edit: I ended up going down a different path to address this issue inside of Options instead. I'll push a commit sometime tomorrow when I'm done.

[MikeMcQuaid]

Looks good when 🟢, thanks!