Hide Forgot
Description of problem: libtirpc.so.1.0.10 uses gethostbyname, which may impact IPv6 support Version-Release number of selected component (if applicable): libtirpc-0.2.1-3.el6 How reproducible: deterministic Steps to Reproduce: 1.RPMDIFF for libtirpc-0.2.1-3.el6 compared to libtirpc-0.2.1-1.el6 Actual results: VERIFY libtirpc lib/libtirpc.so.1.0.10 on i686 ppc s390 uses function gethostbyname, which may impact IPv6 support VERIFY libtirpc lib/libtirpc.so.1.0.10 on i686 ppc s390 uses function inet_ntoa, which may impact IPv6 support VERIFY libtirpc lib64/libtirpc.so.1.0.10 on x86_64 ppc64 s390x uses function gethostbyname, which may impact IPv6 support VERIFY libtirpc lib64/libtirpc.so.1.0.10 on x86_64 ppc64 s390x uses function inet_ntoa, which may impact IPv6 support Expected results: rpmdiff passes Additional info:
The callers of these functions look like they come from either legacy code in the library (AUTH_DES) or some other seldom-used stuff. I doubt this will make a big difference in practice, but I'll take a closer look and discuss it with upstream folks. Maybe we can just rip this code out, or ipv6-ify it. The strange bit here though is that this code hasn't really changed from the very early versions of libtirpc. Is this test new or did it change in some way? I'm not sure why right offhand that libtirpc-0.2.1-1.el6 would have had a different result.
Given that we're this late in the dev cycle and I still don't have anything upstream. I'm going to push this out to 6.3. It's not a problem likely to affect anyone using the more frequently traveled codepaths in libtirpc. What I'm probably going to end up doing is adding a configure switch that enables the AUTH_DES code and then have that turned off by default. It's hard to imagine anyone really using it in this day and age, but that's always hard to predict. The part that has me stumped at the moment is how to handle the header files, so it may take a little while to get this straight.
I've been leaving it open in the hopes I'll get the time to work on it, but so far I haven't. Given that this is only in some (very) legacy code, I suggest we just close this WONTFIX and move on. Doing so -- please reopen if you think this is something that needs to be addressed.