8231259: (dc) DatagramChannel::disconnect re-binds socket to the wildcard address (macOS)

[Chris Hegarty]

> On 19 Nov 2019, at 10:59, Alan Bateman <Alan.Bateman at oracle.com> wrote:
> 
> On 19/11/2019 09:54, Chris Hegarty wrote:
>> :
>> 1) The new dup2 wrapper around the native call seems more suited to sun.nio.ch.Net, rather than an instance method on DCI.
> sun.nio.ch.Net is network so not the right place. We should potentially move it to NativeDispatcher, didn't seem to be worth doing as part of this change but I can do.

Yes, NativeDispatcher ( the location of preClose ) is the right place for this new method.

>> 
>> 2) Why change disconnect0 from a static method to an instance method? The implementation does not need access to the DCI instance? Same for dup2, as above.
> The native methods were originally written as instance methods, second parameter is a jobject. All I've done is fix the declaration to match the method. I have other changes come that will replace the receive0 method and that would be a good time to change the methods to be static, otherwise we would have a mix of static and instance that leads to the mistakes like disconnect0.

I look at this the other way around. The bug is that Java_sun_nio_ch_DatagramChannelImpl_disconnect0 has `jobject this`, rather than ` jclass clazz`. I find no issue with mixing static and non-static, it’s clear that the non-static require to poke in the receiver, otherwise they should be static.  I take it from your comment that the intent is to make receive0 static too ( in the near future ), so why not just fix ` jclass clazz` now, rather than changing disconnect0 to be non-static and then back to static again.

-Chris.