In D23197#508966, @kevans wrote:

In D23197#508569, @brooks wrote:

I see that cheri also has a clean target in ^/sys/compat/cheriabi/Makefile; would it make sense to go ahead and add one to this common Makefile while I'm at it? It doesn't feel generally useful, but it also isn't harmful and the target would effectively just be rm -f ${GENERATED} at this point, and that Makefile could also probably use a similar (edit: cleanup, like these) with GENERATED_PREFIX=cheriabi_. (Even if it doesn't, I note that the sysent target is missing dependency on cheriabi_syscalls.c, not that it matters all that much)

This seems fine. I don't think there's much evidence that anyone currently cares about n32.

Very nice cleanup.

Compiles, but needs testing.

This seems sensible.

LGTM

I'd rather see use switch to static_assert since that's in C11 and C++11. It's a bit unfortunate that only C++17 has the single argument version though.

Looks fine, I agree with the comment on irc/slack that centralizing this in a .mk file likely makes sense. It's no much code, but it's a lot of duplication.

This seems fine and won't be hard to merge with our local changes.

I'm wondering what happens with libc's use of elf_aux_info...

Looks good to me other than the minor issue on sparc64.

The missing std*.h files are due to our inability to build FreeBSD when they are installed. We really need to land D17002 and MFC it so we can eventually start installing them.

This seems fine.

LGTM. I'll commit along with and update of the snapshot shortly

• Improve whitespace consistency.

• Add missing dependency

This seems like a reasonable approach.

Committed in rP518154 and rP518155

This review is born out of frustration with having to manually build git on our systems where we need git-subversion and prefer to use pkg in general.

LGTM

LTGM. Adding Alex in case he has concerns about bootstrapping on MacOS/Linux.

Passed tinderbox (sans mips and sparc64) overnight.

Fix typo.

LGTM

In D17002#489608, @arichardson wrote:

This change is required to build with an unmodified upstream LLVM since the default upstream clang builtin headers are incompatible with the ones in /usr/bin/include (see my earlier comments in this review).

--sysroot only affects the normal include paths, but does not change the search path for compiler-provided headers.
SYSTEM_COMPILER compiler should not make any difference here and will work as expected with /usr/bin/cc:

/usr/bin/cc -print-resource-dir
/usr/lib/clang/6.0.1

/usr/lib/clang/6.0.1/include contains all the necessary compiler-builtin headers (but we don't install the stddef.h/etc. headers that are incompatible)

It's probably worth making elf_freebsd_sysvec static, but this is fine.

LGTM

It's probably too late, but it would have been nice to pick one of long or int for the integer arguments...

The syscalls.master bits look correct now.

I think this is ready to go as is. PLIST cleanups can wait IMO.

This revision is long OBE. Commandeer so I can abandon it.

Works for me.

Over all this is a massive improvement. I was able to port a number of local changes quickly.