shared: only fall back to -march=native on supported architectures

[alebcay]

• 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 typecheck with your changes locally?
• Have you successfully run brew tests with your changes locally?

---

On Linux, -march=native is added in the optflags when building from source (and not for a bottle). This works fine for the case where users install brew in a custom prefix on a supported architecture, but breaks the case where the user is building from source for running under Linux on an unsupported architecture for which the -march=native flag is not valid (I specifically ran into this issue while tinkering with brew under Ubuntu on RISC-V).

Introduce a new optimization flag option none and adjust effective_arch to only fall back on :native if running on Intel or ARM (which are not the only architectures which support -march=native, but it does cover all existing supported use cases and leaves room for current development/experimentation happening with brew on Linux arm64), and falls back on the more conservative :none option if not running one of those architectures.

[MikeMcQuaid]

On Linux, -march=native is added in the optflags when building from source (and not for a bottle).

Just putting it out there: everyone who felt very strongly about this being the case has left the project. Other maintainers may still feel strongly about this but: I think it's worth reopening the discussion about removing this behaviour. Reasoning:

• breakage like this
• it's inconsistent with macOS (that doesn't use -march=native like this)
• it makes source builds inconsistent with bottles which harms debugging

[carlocab]

Seems fine to me.

• breakage like this

This kind of breakage (using brew on RISC-V) seems exotic enough to not be worth making decisions that will affect many other users. If there are other types of more common breakage, we can discuss those, but I don't want to base it on this one.

• it's inconsistent with macOS (that doesn't use -march=native like this)

macOS and Linux have different baselines -- on macOS, we do -march=westmere or -march=nehalem, which are reasonable. The baseline for Linux is -march=core2, which is pretty bad.

• it makes source builds inconsistent with bottles which harms debugging

I don't think this has really caused us any significant difficulty in the past few years.

[MikeMcQuaid]

Looks good, thanks @alebcay!