illegal hardware instruction running tailwind since version 4 on Mac

[digitalcohesion]

What version of Tailwind CSS are you using?

3.4.16 works fine
4.0.0-beta.5 and 4.1.10 both fail with an illegal hardware instruction error

What operating system are you using?

For example: macOS sequoia on a 2012 MacBook Pro

Reproduction URL

The error is produced by simply running the tailwindcss-macos-x64 binary

Describe your issue

I ran into a problem first when testing the latest release candidate of Phoenix LiveView. Projects were failing to compile with a tailwind exited with 132 error. It was suggested on the elixir forums I should test the tailwind binary directly.

I downloaded the latest 3.6.14 release and ran it. It runs fine and shows the usage and options. Then I tested the latest 4.1.10 and it fails to run with an illegal hardware instruction error. I then tested the earliest release of 4 I could find on github and it also failed with an illegal hardware instruction error.

As a sanity check I set up a debian 12 vm on my MacBook Pro and downloaded the 4.1.10 version for linux-x64 and it runs fine in the VM so it seems to be something specific about the way the Mac version is built and not an inherent need for a particular hardware instruction.

[wongjn]

Could I check you mean 3.4.16? I don't believe there is a Tailwind of version 3.6.14.

[thecrypticace]

How are you running macOS Sequoia on a 2012 MacBook Pro? Did you patch the installer? (OCLP?) It's definitely not "supported" and that itself could be pert of the problem.

I suspect this is related to use of unsupported vector instructions and a check has been omitted. We use the baseline build of Bun for macOS builds which should work around things like this but unfortunately, I do not have hardware that old to test on.

[digitalcohesion]

Yes, sorry, 3.4.16.

[digitalcohesion]

If you can help me get things setup I can test further on my machine. I'm not familiar with much more than the basic usage of tailwind, as it's default for liveview, but I'd be willing to spend some time on this issue if I can help get it fixed. Alternatively I might be able to get a VM or another machine setup I could give access to for testing if that would be easier.

[thecrypticace]

Yeah, in this case I might actually need access to comparable hardware as it's a CPU fault that bubbles up to throw that error. I'd end up running the CLI through a debugger to figure out what instruction(s) are doing this. But given that the actual hardware + OS configuration is unsupported by Apple (they only support Sequoia on 2018 and later) actually testing this would be rather complicated.

[digitalcohesion]

Yeah, in this case I might actually need access to comparable hardware as it's a CPU fault that bubbles up to throw that error.

I sent you an email. I can provide access to a computer that demonstrates the problem if you're willing to troubleshoot further.