The nesting is quite deep at this point, it would probably make sense to break things up a little but that's orthogonal to this change.

Nice, looks good to me!

Removed blank lines following section comments.

In D26832#600476, @imp wrote:

Modulo the blank line thing, this looks good. It's arguably two different commits, but it's so small splitting it further at this stage likely has little benefit.

Dropped various whitespace fixes to preserve blame.

Actually works now after testing with fixed cxl driver.

Simplified error handling.

Fixed edge cases where first child is a skip node and where descendants of a skip node are also flagged with skip.

Don't skip nodes when iterating with NEXTNOSKIP.

Separated kern_sysctl changes to a different patch.
Use defined CTL_SYSCTL oid names.

The page fault in t4_get_lb_stats with my T580-LP-SO-CR was fun.
I have fixed CTL_SYSCTL_NEXTNOSKIP so it doesn't skip nodes, will update the diffs soon.

Ok I see the problem. We don't get nodes when iterating, so the skip node never gets seen for this to skip its descendants. Back to the drawing board, then.

Restored check that tx is enabled

Incorporated feedback

Nice work, just a few notes.

Updated diff to include man page, example code, build as a lua module instead of embedded in flua.
Diff is based on flua-libjail going in first.

I think this should be in a BUGS section

I intend to commit this once https://github.com/william-gr/py-sysctl/pull/8 has merged and tagged and the port has been updated.

Looks like I'm going to need to add this to py-sysctl, too. It's using sysctl(3) so won't benefit from the fixed sysctl(8). A few of the Python-based tools in OpenZFS use py-sysctl on FreeBSD to get at kstat sysctls, which are exactly what I have planned to SKIP.

Add 3lua section to contrib/mandoc/msec.in

Addressed some feedback.

Also visit all with tflag

Bumped to a version that works on 12 and 11, too.

Assuming it currently builds on 11.3 and 12.1 this shouldn't break anything, but I'll test it to be sure.

In D26258#583657, @adrian wrote:

Lol, who did that? :-)

Oh well!

Generally happy about the cleanup and the added flexibility in the build infrastructure.

Looks reasonably noninvasive. I'm surprised it's that easy :)