On 2/7/18 12:55 PM, gary.ad...@oracle.com wrote:
On 2/7/18 12:31 PM, Chris Hegarty wrote:
Gary,
http://cr.openjdk.java.net/%7Egadams/8080990/webrev.02/
I think the replacement of WSASendDisconnect with
shutdown(SD_SEND) should be fine. I do note that there
is another usage of WSASendDisconnect in
java.base/windows/native/libnet/net_util_md.c.
Thanks for the reference.
I'm not sure how that one slipped by unnoticed.
I found in make/lib/NetworkingLibraries.gmk the BUILD_LIBNET
target includes a number of specific disabled warnings. In particular,
4996 which suppresses all deprecated function warnings.
I think I'd like to get the current changes pushed to remove the specific
WINSOCK_DEPRECATED_NO_WARNINGS and then file another bug to
come back and look at selective makefiles that use a blanket 4996
suppression of deprecated warnings.
[ Maybe you want to separate out the changes in java.base
( NIO and NET ) from the serviceability changes? Up to
you. ]
Actually, this change is about removing the compilation flag globally
in flags.m4 and any sources that need to be updated
to replace the deprecated functions.
Curious about the specific hints you have chosen to use.
In other areas we have the following:
hints.ai_flags = AI_CANONNAME;
hints.ai_family = AF_INET;
[ Not saying that what you have is incorrect, just questioning
if you need to specify the socket type and protocol ]
One of the benefits of getaddrinfo is that it can return a list of
addresses from the name service. In the places we were using
gethostbyname we were looking for just one ipv4 address.
By adding additional constraints on the supplied hints, it can help
reduce the list of returned addresses.
-Chris.
