Apple

	/*  check for mapped or compat addresses */
	i = IN6_IS_ADDR_V4MAPPED(addr);
	j = IN6_IS_ADDR_V4COMPAT(addr);
	if ((i != 0) || (j != 0))
	{
		a4.s_addr = addr->__u6_addr.__u6_addr32[3];
		sprintf(tmp, "::%s%s", (i != 0) ? "ffff:" : "", inet_ntoa(a4));

On Jan 4, 2016, at 12:23 PM, Jeff Johnson <email@hidden> wrote:

I can reproduce (on 10.10 FWIW).

It seems to be inet_ntop that's failing. If you look at the memory, inet_pton seems to give the correct result.

The strange thing is that Apple's open source code works correctly:

http://www.opensource.apple.com/source/xnu/xnu-3248.20.55/libkern/net/inet_ntop.c

So, apparently whatever code is in the OS is not the same as in the open source.

-Jeff

On Jan 4, 2016, at 8:30 AM, Robin Hack <email@hidden> wrote:

Hi mighty brains!

I would like to report possible bug in libSystem.
While porting nss_wrapper (part of cwrap.org) project one with my test fails when I tried to work with ipv6 in form: ::12.

After close debugging it looks like libSystem contain bug which is reproducible by attached reproducer.

Core problem:
Looks like one of inet_*() functions are not able to translate ipv6 addresses in form ::XX where XX are numbers.

Intresting fact is: ::1 works as expected.

Steps to Reproduce:
1) compile attached file libSystem-bug.c
2) run

Expected Results:
Output of attached reproduceer. Same results on (FreeBSD, OmniOS, Linux):

PASS: match with '::1' found!
result (pointer): ::12
result (buffer): ::12
PASS: match with '::12' found!

Actual Results:
PASS: match with '::1' found!
result (pointer): ::0.0.0.18
result (buffer): ::0.0.0.18
FAIL: no match with '::12' found!

Version:
10.9 and 10.11 (with latest updates).

Notes:
* I tested this behavior across different operating systems (FreeBSD, OmniOS - aka Solaris, Linux Fedora 20).
* I already opened bug 23739955 however I feel that this bug report would be better here.

<libSystem-bug.c> _______________________________________________