don't permit freeaddrinfo(NULL). now the behavior is consistent

across {free,net,open}bsd.

both rfc2553 and X/Open spec are silent about the behavior,
and there's no strong consensus either.  i think library should NOT be
forgiving in this case, to promote development of more robust 3rd-party
codebase (code works on "freeaddrinfo(NULL) = SEGV" will work on
"freeaddrinfo(NULL) is okay" environment, but not the other way around).
only issue i have now is NRL freeaddrinfo() compatibility, which permits
freeaddrinfo(NULL).

1 files changed, 2 insertions, 2 deletions

diff --git a/src/lib/libc/net/freeaddrinfo.c b/src/lib/libc/net/freeaddrinfo.c
index 40534f4422..30fbecb805 100644
--- a/src/lib/libc/net/freeaddrinfo.c
+++ b/src/lib/libc/net/freeaddrinfo.c
@@ -39,11 +39,11 @@ freeaddrinfo(ai)
 {
         struct addrinfo *p;
 
-        while (ai) {
+        do {
                 p = ai;
                 ai = ai->ai_next;
                 if (p->ai_canonname)
                         free(p->ai_canonname);
                 free((void *)p);
-        }
+        } while (ai);
 }