The combination of #16619 and #16960 lead to some test failures in bazel itself when building targets that depend on the darwin_universal_binary rule added in java PR.

The failure looks like this:

(20:06:01) ERROR: /private/var/tmp/_bazel_buildkite/4b66776e1a7e7db9d1f29690af6b43c8/external/local_config_cc/BUILD:66:19: in cc_toolchain_suite rule @local_config_cc//:toolchain: cc_toolchain_suite '@local_config_cc//:toolchain' does not contain a toolchain for cpu 'darwin_x86_64'

And is caused by the fact that if you want to build non-host platform macOS binaries you must use the new toolchain that lives in the apple_support repo, which bazel does not. I don't have full context on the java issue so I wanted to file this to discuss possible solutions. I see at least 2 options:

1. Make the builtin cc toolchain on macOS support cross-compiling binaries for x86_64/arm. The downside of this is that it adds complexity to the cc toolchain that's macOS specific and the goal of migrating the cc toolchain to the other repo was partially to remove that complexity from bazel itself. I think at the very least it would require some custom setup of the toolchain for macOS (to setup 2 toolchains), as well as adding at least 1 new feature for adding the right -target flags, which are just not passed and the compiler picks the default today
2. Try to unblock java using a separate binary on the different hosts, specifically this comment from the java PR made it sound like that might be possible:

That's what I tried first, but it was a more invasive change. If we were to have add a new repo for apple silicon, I think we should rename the existing one to "darwin_intel" or "darwin_x86_64" (leaving it as "darwin" would be misleading/incorrect). So I tried that, and it ended up getting very messy. There are a lot of assumptions in code and tests that there is only a single darwin platform (i.e. platform == OS).

I wonder if this change #17547 also would help with that.

I'm happy to take on the changes here soon so we can ideally avoid reverts, but I'm hoping for some feedback before choosing a path here.

// @googlewalt, @comius, @hvadehra