Re: RFC: InetAddress.lookup() change (Win32 parts untested)

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Gary" == Gary Benson <gbenson@redhat.com> writes:

Gary> The attached patch (which needs to be applied after my patch with
Gary> subject "RFC: Untested Win32 InetAddress tweak") changes lookup to
Gary> always return an array, and to create address objects using
Gary> InetAddress.getByAddress() which makes Inet4Address or Inet6Address
Gary> objects accordingly.

I skimmed your patch and it all looked reasonable to me.  I'm sure at
this point you know this area better than I do anyway :)

I do have a couple things to point out.  I hate to add on, but these
things really would be helpful.

First, InetAddress is not merged with Classpath.  Could you take a
look at merging it and removing our divergence?  Every divergence here
is a maintenance burden and a place we can mess up when importing new
versions of Classpath... we should only have them when they are really
necessary, and I don't think InetAddress should fall into that
category.

The helper method allocArray hasn't been needed for eons; there is
equivalent CNI code.

Unfortunately doing this change will make your Windows problem worse,
not better :(.  So I guess we should hold off on the bigger change
here until the 4.2 branch is made, and just go with your smaller
change for now.

Gary> The change is to a part of InetAddress.lookup() that is identical
Gary> in both Posix and Win32 versions, but as before I don't have the
Gary> ability to test the Win32 part of this patch.  Could someone please
Gary> test it for me?

If no one does this, I would suggest filing a PR, attaching the
Windows-specific part of the patch, and committing the rest.

Tom