Fixup with current versions

[MasonProtter]

This should resolve some warnings we were getting and fix the broken native_llvm_module(funcs::Array) which was breaking the WASM integration test.

Tests now pass locally for me, but looking at the CI logs before this PR, I saw some other failures in the CI that I don't see locally, not sure yet what the deal is with those.

This also doesn't address the failures on nightly.

[brenhinkeller]

Oh awesome!!

[MasonProtter]

So the status is

• Julia Nightly -- CI fails
• Julia Nightly -- Integration fails
• Julia v1.9.0-rc3 -- CI passes
• Julia v1.9.0-rc3 -- Integration passes
• Julia v1.8 -- CI passes
• Julia v1.8 -- Integration fails

Since 1.9 is about to release, what do we think of just abandoning 1.8 and focusing on 1.9 and nightly?

[brenhinkeller]

Since 1.9 is about to release, what do we think of just abandoning 1.8 and focusing on 1.9 and nightly?

Fine by me! I haven't used 1.8 in months haha. After you merge this I'll update the integration tests on the StaticTools side to use the same versions.

[MasonProtter]

The fixes for nightly are pretty easy, other than that clang() is unhappy with us. If we just use the system cc instead of clang() the tests pass locally for me. Hopefully we can fix up whatever is causing that, but if not, no big deal because we already use the system cc on MacOS.

[MasonProtter]

Okay, it seems that it was either that my julia build was too stale or my local manifest was too stale. Apparently the clang issue was already solved earlier, lets see how the test suite does now.

[MasonProtter]

The remaining v1.8 failures seem to be about v1.8 failing to sufficiently optimize stuff away resulting in dynamic dispatches.

[MasonProtter]

Got it going locally by removing some invocations that used --compile=min. This'll slow down some tests but oh well.

[brenhinkeller]

Oh nice! Ahh, I hadn't thought about the consequences of --compile=min in those scripts!