I've seen that when using ATF, ${PWD} is preferred over $(pwd) (efficiency, I guess).

Very nice!

Thank you!

I think this is not sufficient (just to avoid a missing mac.conf entry):
I will test it thoroughly later, but for now:

# jail -c path=/ name=D54786 persist
# jls -s

Nice! I may steal a few ideas for some other similar tests.
I made a small observation, that can be ignored, I'm just thrilled with my recent discovery of atf_check -r.

After this refactoring, the test syslogd_test:forward becomes very flaky. It should be all fixed in the following revision.

But I also have one more question: one of the warnings is also about unexisting .Xr sysctl 2. As I can see, there is no sysctl(2) in FreeBSD, only sysctl(3). But the man pages talks about it in the context of syscalls, so it seems that we can't just change sysctl(2) to sysctl(3) since sysctl(3) is not a syscall. What sould we do about that?

In D54761#1251514, @tembun_bk.ru wrote:

In D54761#1251476, @jlduran wrote:

If possible, try to fix all the mandoc -T lint share/man/man4/capsicum.4 recommendations while here.

Sure. One of the warnings is: STYLE: referenced manual not found: Xr pdwait4 2. As I can see, pdwait(2) syscall is about to be introduced by https://reviews.freebsd.org/D54592 soon. So shall we keep the reference to pdwait4(2) (but change it to pdwait(2))?

Thanks. The "fi" ligature was probably an artifact from copy/pasting from the PDF article.
If possible, try to fix all the mandoc -T lint share/man/man4/capsicum.4 recommendations while here.

I'll just submit the trivial rdr_action_head() fix and abandon the rest.
I'll wait for the test timeouts dust to settle and simply try sleeping a bit more once I gather all the flaky tests.
Thank you for your input!

In D54681#1250340, @ivy wrote:

ideally these should have a package= tag, but i'm leery about adding those without more testing, so that can wait until someone (me) comes along and adds package tags for everything in this file.

I imagined so. Thank you!

I think this has already been fixed in 742d93da842afa34c5268e96c6a3f95ad44ae770.

I'll add the comment upon commit:

/* https://reviews.llvm.org is deprecated and this review is never going to be updated or completed */

Thanks.

In D54695#1249657, @kp wrote:

I'm not sure this is sufficient. It is still possible for tcpdump to have started, but not gotten to the point of actually opening the pflog device.

I've had a very quick look at the tcpdump code, and I think it looks like we could potentially rely on the existence of a capture file (so -w <file>). I think that file only gets created after we've done pcap_open()/pcap_setfilter(). That'd require a bit of test-reworking though, because we'd be saving a pcap file, which we'd have to translate back to text to run the checks on.

In D54682#1249624, @ivy wrote:

i think you also want to add /usr/share/pkg/triggers. currently only usr.bin/mandoc installs a file here, but more things will in the future.

This might be an overkill, but I would like to know if a similar helper function could help avoid failure intermittence (flakiness) in some tests. I reproduced this test natively on a very fast aarch64 machine, and it did not fail once, comparing the output of a successful pass vs. a failed one makes me wonder if it is not sleeping enough: