Approved

Fine with me

I doubt an exp-run is necessary, as only three binutils are built today, as, ld, objdump (on some subset of architectures), and if anything does break I expect it would most likely be found by tinderbox.

Although we still need to make sure that everything still works with this change.

No objection from me.

In D16140#342262, @bz wrote:

What about (what do others think):

• leave the non static and not rename them to *_GLOBAL_*
• rename the static to STATIC_DPCPU_DEFINE

That way grep()ing for DPCPU_DEFINE will still give us all of them, there's less noise in the change, and only the problematic case is (will be) dealt with.

I'm fine with this as a NFC first step

LGTM

Just tried building lang/gcc7 on my gets-less dev system and see:

Some suggestions but overall I think this is fine. I can futz with it after commit otherwise.

Update with kib@ and jhb@ feedback

I requested an exp-run in PR222796. I think we could deal with most of the failures post-commit, but a few gcc ports fail and we should probably address those in advance.

Any further comment @andrew?

I wonder if we also want to restore ibcs2's int return value?

Review feedback from @kib

In D15687#339170, @emaste wrote:

The libssp subdir should be added only if both MK_SSP and MK_GCC are true; someone who sets WITHOUT_SSP but WITH_GCC should not get it.

For reference, BSDL __stack_chk_fail in Android https://android.googlesource.com/platform/bionic/+/ics-mr1-release/libc/bionic/ssp.c and NetBSD http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/misc/stack_protector.c?annotate=1.9&only_with_tag=MAIN.

cd to the VCS topdir before invoking git diff

Hrm, this only works if git diff is run at the top level dir; needs further investigation.

In D15952#337808, @chuck wrote:

Bah, this one was named linux32_trans_osrel() so my search missed it. I'll add that to this patch as well.

I'm fine with either check (<= 0 or @cem's version) with the same change applied to linux32_sysvec.c

In D15952#337749, @chuck wrote:

I'm only finding two copies: sys/i386/linux/linux_sysvec.c and sys/amd64/linux/linux_sysvec.c but have updated the patch to fix those copies.

We will also need to address the four copies of linux_trans_osrel (but after further investigation if desired, I have not checked exactly how the value they calculate is used).

In D15952#337697, @cem wrote:

Right, I suggested using LINUX_KERNVER() to do this in my initial comment :-).

Ah, top level Linux Makefile emits a definition of KERNEL_VERSION using shifts.

Ah, I see the comment in the bug now - yes, if Linux uses shifts we should too. It would be good to put a reference in the commit that mentions the Linux macro/etc. that calculates it though.

In D15952#337607, @cem wrote:

Ed, can you explain it in a little more detail?

In D15858#336639, @chuck wrote:

In D15858#336162, @emaste wrote:

See linux_map_osrel

This probably fits better with the existing code, and I'm inclined to move that direction. Is using linux_kernver() the way to access this value?

I'm still getting the handle of the process here. In my local repo, these are two separate git commits, are you saying the correct way to do this would have been two separate arc diff commands for each of my git commits?

I took care of the NFIT man page addition in rS335459 - it was my fault :)

In D15904#336964, @mat wrote:

What matters is that it is available on all supported releases :-)

Another option: upon discussion with @dim we will likely end up installing llvm's llvm-objdump as objdump. It's broadly compatible, but the output format etc. might be slightly different.

On second thought I'm not sure about this: --print-file-name options typically prefix all output lines with the file name.

This looks fine to me. It's probably worth adding a comment explaining the use of ${NONEXISTENT} though.

In D15904#336884, @mat wrote:

Mmmm,

$ readelf --print-file-name -r /usr/bin/mktemp
readelf: unrecognized option `--print-file-name'

Reported in the PR -- nm might be more useful for this script.

LGTM

Would it make sense to set some new variable based on TOOLS_PREFIX, MK_LLD, MK_LLD_BOOTSTRAP so that the same expression doesn't need to be repeated?