Hi Gary, ok, I would agree that a change like globally moving from dbgSysInetAddr to something like dbgsysPToN should at least be handled by a different change.
Best regards Christoph From: Gary Adams [mailto:gary.ad...@oracle.com] Sent: Montag, 5. Februar 2018 20:15 To: gary.ad...@oracle.com Cc: Langer, Christoph <christoph.lan...@sap.com>; OpenJDK Serviceability <serviceability-dev@openjdk.java.net>; OpenJDK Build <build-...@openjdk.java.net>; OpenJDK Networking <net-...@openjdk.java.net> Subject: Re: JDK-8080990: libdt_socket/socket_md.c(202) : warning C4996: 'gethostbyname': Use getaddrinfo() or GetAddrInfoW() One more to fix to cover the remaining test failures I was seeing. Previously, using inet_addr, there was a single -1 return that needed to be checked. Now that we're using inet_pton, there is a -1 and a 0 error return to be considered. My preference would be to leave the dbgSysInetAddr name unchanged and have the low level check for inet_pton errors to simply return the -1 that was previously checked. This specific issue is about the deprecation of library functions on windows. I'd recommend not upgrading the other platforms at this time until a complete update is done for ipv6 support. I'll post a new webrev when I've retested using the updated inet_pton error checking. ... Hi Gary, > Here's a revised webrev > > > http://cr.openjdk.java.net/~gadams/8080990/webrev.01/index.html<http://cr.openjdk.java.net/%7Egadams/8080990/webrev.01/index.html> > > Still testing ... > > Using shutdown() fixed problems reported by the > java/nio/channelSocketChannel tests. The fix looks good. I would think we should rename dbgsysInetAddr to dbgsysPToN and use inet_pton also for the Unix/Linux platforms. It would be the better choice there as well, though we still only support IPv4 in libdt_socket. > > I also noticed prior use of getaddrinfo for "localhost" was not calling > freeaddrinfo. Thanks for spotting/fixing that. Best regards Christoph
