ATF is incompatible with capsicum

[asomers]

@jansucan discovered that if an ATF test case enters capability mode with cap_enter, then it won't be able to write the results file. It seems like this would be easy to fix, but simply opening up the results file at the top of atf_tc_run. Thoughts? cc @jmmv @ngie-eign .

[cemeyer]

I discovered and fixed this in 2017: https://github.com/cemeyer/atf/commit/317c80628513337af41b29fceeaa62fbaf498a02 . Jmmv won't take my change because I can't sign Google's CLA on my employer's behalf.

[asomers]

That's lame. Do we need to fork atf?

[cemeyer]

Sure, a freebsd/atf fork is fine with me. If jmmv's fork has activity (it's pretty idle) we can always pull new material down.

[asomers]

Who has the authority to create new repos in the freebsd organization?

[cemeyer]

Not me. I'd ask Ed.

[krytarowski]

What is new with this fork? In NetBSD I would love to rely on plain C rather than fancy C++. C++ creates some burden on the system. It happened to be harder to get C++ right, sanitized etc than execute some preexisting regression tests that are mostly C code.

[cemeyer]

I doubt we would make any immediate move away from C++.