We're no longer using monthly.cgi

No objection. Please make sure that the releng documentation gets updated to make sure it gets turned back on again (either when making the stable branch or the release branch)

They are indeed not broken which is why I originally abandoned the review, but reopened it based on cem's feedback - as written there's some opportunity for confusion.

Seems fine to me.

Fine with me

Would you rebase/regenerate this patch so that those pieces that have been split out and committed separately drop out of the review?

In D20974#457245, @greg_unrelenting.technology wrote:

Would be nice to test on ThunderX2 :)

We could instead hook it into BROKEN_OPTIONS in bsd.opts.mk but I'm not sure how to use bsd.compiler.mk and bsd.linker.mk there.

Please post patch to elftoolchain-developers too

Please post patch to elftoolchain-developers too

Would you post the patch to elftoolchain-developers too?

initial updates incorporating kib feedback

I can confirm this attaches on my eMAG but I cannot verify the watchdog functionality - were you able to test it?

@greg_unrelenting.technology do you have a git repo available with your work in it? It'd be slightly easier for me to just add your repo and grab changes directly.

A Cirrus-CI build is running now.

I have a CI build running with this change now and will commit if that completes successfully.

Use incdir

The functions here don't match the canonical versions in sys/net/ethernet.h and seem unusual, but aren't actually broken since there's no wrapping on the u8 additions. That said I'm still happy to commit this if it seems more standard.

meh

further updates from kib

Simplify some boolean conditionals per kib

upload correct diff

LGTM

In D18028#456257, @crest_bultmann.eu wrote:

I applied the patch to FreeBSD 12.0 on a T470s. ...

Seems reasonable. Was originally introduced in rS173375 but Clang is the default compiler and we can assume that contemporary GCC won't have "the -O2 issues."