Caleb Maclennan

> [kspalaiologos/bzip3/runs/6434386258?check_suite_focus=true#step:4:1136](https://github.com/kspalaiologos/bzip3/runs/6434386258?check_suite_focus=true#step:4:1136) - something is also wrong with the FreeBSD build. FreeBSD doesn't have time for you ;-) ``` time: ./: Permission denied ``` I suggest not mixing benchmarky things...

Hmm, that might be a `bmake` handling difference here, it makes fewer passes setting variables along the way. Try this on for size (`git am < x.patch`): ```diff From bc3b5a35c59311c43702383ef9e0cd9cf9d53228...

Still no. I'll work on that on my fork...

Nope, that warning is unrelated (and `nawk` is installed), but since it is also *expected* to return non-zero there I'll suppress that warning while I'm at it.

Gah! Apparently `$

I don't know what rabbit hole I just fell down. BSD documentation for make suggests that support for `$

Thanks @ypsvlq a little testing suggests you are correct. I'll make accommodation. I already know how it just wasn't nice not to understand *why*.

@kspalaiologos I think forcing `gmake` as seen in bc19e13 was a mistake, we *should* be testing with BSD's native make and we should keep it fully portable so that every...

On the Windows front, I have some unfinished work in the `autotools-windows` branch on my fork you might be interested in. Using `gitbash.exe` and clang on the `windows-latest` platform I...

Using the Windows host for testing would be pretty easy. Cross compile, post binary artifacts, have another job dependent on the first, download artifacts, test a roundtrip with it. Also...