In D24410#537510, @imp wrote:

Yikes!

I usually strip the tests part of patches since we don't do anything with them, but this is fine.

CI run https://cirrus-ci.com/task/6577255742701568

Successful amd64 CI build https://cirrus-ci.com/task/5249932466585600

In D24407#536889, @kib wrote:

Could you please extract two changes into separate commits:

1. change of the impg ps_strings type
2. addition of AUXARGS_ENTRY_PTR

Then the rest would be much cleaner.

Note that this will be committed in five commits (change ps_strings type, add and export new values, procstat support, _elf_aux_info, setproctitle).

OpenZFS will arrive and fix this soon enough

• Use a less generic name for the header (abi_compat.h).

In D24275#535453, @cem wrote:

For the generic CP() type copy operations, I'd suggest _Static_assert size checks. Exceptions can have an exceptional copy mechanism (maybe provided here).

• Improve whitespace and use a consistant style.

The values are also quite different if you diff the nm output. I think this whole thing will need to be rethought eventually, but removing -fcommon is probably a long ways off.

Current version passes universe (except for an unrelated config error on riscv).

CI run without the last mps fix: https://cirrus-ci.com/task/5297194320265216

unfoobar diff after arc malfunction.

• mpr: we need PTRIN unconditionally

• Another typo fix.

In D24265#533908, @emaste wrote:

The swig input is ${LLDB_SRCS}/bindings/lua.swig

Hmm, does this belong here or as part of liblua? It doesn't look like this binding depends on lldb source at all.

I've submitted this upstream https://reviews.llvm.org/D77333

• Typo fix, thanks @bcr.

• Fix issues reported by @imp.

LGTM. I'll commit shortly.

LGTM. I like that we'll be able to have functional decomposition rather than smooshing ever more things into a list.

Also successfully built with make -j16 -DWITHOUT_TESTS -DWITH_TESTS_SUPPORT buildworld

Successful CI run with updated patch: https://cirrus-ci.com/task/5744739794288640

• Take two: carry MK_TESTS_SUPPORT through regardless of MK_TESTS.

I believe what's happening is that MK_TESTS=no is causing MK_TESTS_SUPPORT to be implied in the sub make. I'm now testing a change where we pass the currently value of MK_TESTS_SUPPORT on through.

This definitely doesn't work. https://cirrus-ci.com/task/5639234392424448 (Breaks normal builds).

Here's the ports fix. https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=245107

In D24199#532259, @rgrimes wrote:

Thanks, and this should happen in the port too as you pointed out it was also doing the wrong thing.

Somewhat annoyingly, these fixes aren't -DNO_CLEAN safe as the changes are all in Makefiles.

• Use SUBDIR.${MK_FOO} variables to reduce .if clutter.
• Don't delete empty DESTDIR. Print deleted directories.

This feature was developed by @arichardson in CheriBSD. I've extracted it and tweaked it a bit.

Committed in rS359260.