Good catch. Looks like those have been dead code ever since r292373.

In D21287#466723, @cem wrote:

This approach looks technically correct to me, and it simplifies the logic, but there is definitely some microbenchmark cost to exiting to userspace every retry instead of doing so "a few times" in the kernel (with some algorithm). But this is maybe just premature optimization on my part.

Accomodate buggy or malicious fuse servers that always return ERANGE for
FUSE_LISTXATTR. Instead of restarting within the VOP; return ERESTART.
That way the blocked process will be killable.

In D21434#466423, @sucanjan_gmail.com wrote:

I would also like to ask for help with copyright. What should I use if some code is not mine (mainly the #define directives) and some is?

Well, it fixes the build. But now the program segfaults.

The build fails on sparc64 and powerpcspe with this error. Weirdly, it passed on powerpc and powerpc64; I don't know why:

/home/asomers/freebsd/base/head/sbin/ping/ping.c: In function 'main':
/home/asomers/freebsd/base/head/sbin/ping/ping.c:923: warning: cast increases required alignment of target type

You still need to add entries to UPDATING and ObsoleteFiles.inc. Other than that, it looks good.

This revision needs to be rebased since the changes to -n and -H.

In D21377#465322, @sucanjan_gmail.com wrote:

It will take me some time to find out why. On my system it builds. Could you please send me your main.c? It's interesting that there are double definitions of the functions.

Now it doesn't even build:

/usr/home/somers/freebsd/base/head/sbin/ping/main.c:249:1: error: redefinition
      of 'main'
main(int argc, char *argv[])
^
/usr/home/somers/freebsd/base/head/sbin/ping/main.c:59:1: note: previous
      definition is here
main(int argc, char *argv[])
^
/usr/home/somers/freebsd/base/head/sbin/ping/main.c:331:1: error: redefinition
      of 'usage'
usage(void)
^
/usr/home/somers/freebsd/base/head/sbin/ping/main.c:141:1: note: previous
      definition is here
usage(void)
^
2 errors generated.

Oh, and you need to update rescue/rescue/Makefile, too.

Don't forget to add an entry to UPDATING and to ObsoleteFiles.inc

One more thing: you should normally add an entry in UPDATING letting people know that the options have changed so they can update their scripts. I'll go ahead and write it this time.

Uhh, the requirement would be that I need to read the entire review, not just the most recent commit.

Please add a copyright header with a $FreeBSD$ tag to the top of ping6_test.sh. Also, add an entry to etc/mtree/BSD.tests.dist.

Please add a copyright header with a $FreeBSD$ tag to the top of ping_test.sh

In D21351#464706, @cem wrote:

Hm, I'm not sure I like this being the new default. I guess Linux does it, but there are or used to be many networks where reverse DNS was totally broken (as in, silently dropped). It is nice that ping is mostly a test of IP (ICMP) and not the higher level UDP/TCP protocols DNS uses.

Is there a very short timeout on rDNS lookup failure or on broken networks will ping just hang for 30 seconds at a time?

I'm going to reword the man page entry a bit. Other than that, this change looks good.

In D21292#464559, @sucanjan_gmail.com wrote:

Yes, the TTL will be filtered out. Both ping and ping6 tests will be sending only one packet.

I think it would be good to add -t 1 (timeout in seconds) for one packet just to be sure that the test case will not wait for the reply forever (I don't know if ATF has some timeout).

I would recommend leaving -H in. Even though it's the default, defaults change. Script authors don't like to rely on defaults; they like to specify exactly the behavior that they want.

This looks mostly good, but I have two concerns:

1. Should you filter out hop limit the same way you filtered out ttl from the ping tests?
2. The test case takes more than 2 seconds to run. That doesn't sound like much, but it adds up over more than 7000 tests. I suggest reducing the packet count to 1. If you also want to test > 1 packet, then you can add a second test case that requires root privileges and uses -i to reduce the wait.

This needs a rebase now.

In D21338#464272, @sucanjan_gmail.com wrote:

I'm not sure if I understand you correctly.

This is ping's use of F_NUMERIC flag in pr_addr():

	if (options & F_NUMERIC)
		return inet_ntoa(ina);

	hp = cap_gethostbyaddr(capdns, (char *)&ina, 4, AF_INET);

	if (hp == NULL)
		return inet_ntoa(ina);

	(void)snprintf(buf, sizeof(buf), "%s (%s)", hp->h_name,
	    inet_ntoa(ina));
	return(buf);

and this is ping6's use of F_HOSTNAME in pr_addr():

	if ((options & F_HOSTNAME) == 0)
		flag |= NI_NUMERICHOST;

	if (cap_getnameinfo(capdns, addr, addrlen, buf, sizeof(buf), NULL, 0,
		flag) == 0)
		return (buf);
	else
		return "?";

ping doesn't do reverse lookup. So doesn't this change actually increase the differences between the utilities? I think a better change would be to implement -H for ping.

Don't forget to add an entry to etc/mtree/BSD.tests.dist

This test is certainly better than nothing. I wish we had something better, but I can't come up with a superior test myself. I'll accept this in lieu of something perfect.

That's the whole of the compiler's error output. If you want to reproduce it locally, you should be able to do it with env ARCH=powerpc TARGET_ARCH=powerpc make buildworld. I recommend adding -DNO_CLEAN if you do it more than once. And don't forget the appropriate -j X, or you'll be waiting all day.

This change breaks the build on powerpc, powerpc64, and sparc64:

In D21289#463535, @sucanjan_gmail.com wrote:

It seems that that are far more variable parts in the output than I thought. I think the exact ttl value and ordering of responses cannot be relied on either. Is it OK having a test like this with all of that numeric information filetered out?

In D21292#463402, @sucanjan_gmail.com wrote:

Could I use getaddrinfo command for that? Like this:

if getaddrinfo -f inet6 localhost 1>/dev/null 2>&1; then
    atf_add_test_case ping6_c3

Is this a full revocation of r350857?

Remove debugging detritus