Yeah, that seems perfectly OK to me.

Apparently for me, _FAKE_KERNEL just never gets defined, so that part of the code never gets compiled. I have no idea where that define comes from. In the preprocessed lz4.i I simply do not see the offending code fragment at all.

In D27240#608620, @brooks wrote:

How did this work for anyone? I tripped on it in CheriBSD on aarch64 just now (this is with llvm ~11.0.0).

I've been using this for a while already, since otherwise makeman fails for me. :)

In D27071#604022, @yuri wrote:

Should `MK_LLVM_CXXFILT``` be made a default since it leads to a more stable c++filt?

Yes, I agree. __FreeBSD__ is mostly to be used for "am I on FreeBSD at all" checks, that it has a major version in it is maybe a little misleading. The only disadvantage of __FreeBSD_version is that you have to include <osreldate.h> at least.

This could also be fixed by consistently setting the -target flag when building sources. For instance, on stable/12 the default (x86_64-unknown-freebsd12.2 in my case) gives:

Might as well go full -O0. :)

While the Intel compiler support was an interesting addition in the past, it is long overdue to clean it up. Less code in cdefs.h is better! :)

I think this is fine for now. Upstream will want to centralize the define somewhere in __config, but we can take that discussion there.

Ah yes, this is OK. There's probably some more optimization possible, since these libraries build lots of duplicated object files, but I haven't had the time to look at that yet.

In D26044#586338, @zeising wrote:

@arichardson I'm not sure that is enough, there might be ports that include bsd.lib.mk from Makefiles in their respective tarballs.

The number of ports affected might be smaller than I thought though, I thought this was used for basically anything that's being linked on FreeBSD, but bsd.lib.mk isn't used in the ports infra (in ports/Mk/), which I thought was the case.

LGTM, I don't see any downsides. I sort of feel like it's asymmetric that the arguments aren't passed to the fini functions, but glibc also doesn't do that. :)

LGTM, nice catch!

LGTM. We can even MFC this.

LGTM. It works for me with the libwidevinecdm.so from the PR. I tried to generate an .so with weird phdrs using patchelf, but that didn't work. Not sure if there are other such tools...

This is fine. I understood there are some libc++ headers that do just ; as the body for something, and that triggers it.

Which builtins are those? As far as I know with rtld we should be careful not to just pull in external dependencies. @kib is usually guarding against that sort of stuff. :)

Thanks for this review, it's now effectively enabled via rS364732 for the change that went into upstream, and rS364733 to change src.opts.mk.