I did do a quick eyeball of the results before entering the jira - I guess the "goto fail" in getaddrs caught my eye - that skips the freeaddrinfo call, but now looking again it seems thats an unlikely case (oom).
Feel free to close, it might be good to document this somewhere, perhaps a src/c/valgrind.txt file that details the exceptions in case someone else runs valgrind? Isn't there some way to pass known exceptions to valgrind? Perhaps that's a better way (commit the exceptions to svn).
anyway - looks like we are good for 3.1.1, feel free to close.