test/os/mac: add tests to validate pkg-config library versions

[Bo98]

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same change?
• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you written new tests for your changes? Here's an example.
• Have you successfully run brew style with your changes locally?
• Have you successfully run brew tests with your changes locally?

---

These tests assume the needed SDKs are correctly installed, i.e. brew doctor passes.
The CLT version installed should be the latest available for the running OS.
The tests do not check other OS versions beyond than the one the tests are being run on.

It is not possible to automatically check the following libraries for version updates:

• libedit (incorrect LIBEDIT_MAJOR/MINOR in histedit.h)
• uuid (not a standalone library)

Additionally, libffi version detection cannot be performed on systems running Mojave or earlier.

For indeterminable cases, consult https://opensource.apple.com for the version used.

[MikeMcQuaid]

Very nice 👏🏻

[Bo98]

Tested (and passing) on:

• macOS 10.15.6 Catalina
• macOS 10.14.6 Mojave

I can test on 10.13 High Sierra on one of the CI machines at some point.
I will test on 11.0 Big Sur once brew install-bundler-gems is working. Sorbet is the blocker there and I'm working with upstream to get that solved.

[MikeMcQuaid]

I can test on 10.13 High Sierra on one of the CI machines at some point.

Could consider making it one of these things that actually runs on the CI machines (like the brew doctor tests)?

[Bo98]

Possibly, though tests are mostly designed to ensure the pkg-config files are updated when things change in the OS, which won't happen for older OS versions.

I suppose it could be tested when the tests themselves change - if that is expected to happen.

[MikeMcQuaid]

Yeh, I guess I was thinking a test like this could be an easy way of avoiding manual verification. Could even be a (optional) brew doctor one?

[Bo98]

The effects of a wrong version in the *.pc files are often extremely minor. I don't think we've ever had a bug report of anything caused by that.

Something in brew doctor might make sense, but it should definitely be optional. I don't think we gain anything from making users run it that other checks don't already cover.

[MikeMcQuaid]

@Bo98 Yeh my main thoughts are:

• we already have CI for brew doctor we could piggy-back off
• this approach avoids having to install/run rspec for system-setup stuff
• checking the presence of files on the system feels more doctory than brew testsy in hindsight

[Bo98]

checking the presence of files on the system feels more doctory than brew testsy in hindsight

To clarify: although this test does have an exist check (mostly as a sanity check), the point of the tests here aren't to check whether the SDK is installed correctly but rather the *.pc files shipped in brew are correct.

[Bo98]

I can remove that exist check - that would make sense.

[Bo98]

TIL that adding tests can decrease coverage.

[MikeMcQuaid]

@Bo98 Lol simplecov. 🚢!

[MikeMcQuaid]

Thanks again @Bo98!

[Bo98]

Coverage thing seems to be because codecov is picking the wrong base commit. We generate the coverage report on 27559050c83a84fb64167e3518d076adb3e9d7ec, and that's the commit that gets sent to codecov. It's a merge commit of 86eca5a into 6cde372, generated by GitHub (ref/pull/8400/merge). While codecov knows to turn 27559050c83a84fb64167e3518d076adb3e9d7ec into 86eca5a instead, it incorrectly takes the parent of 86eca5a (af1e600) as the base commit rather than use 6cde372 that the merge commit based on.

[MikeMcQuaid]

@Bo98 Good catch, may be either needs a fix in the codecov gem or it's something we can pass by tweaking an environment variable.