RFR[8241072]: 'Reimplement Legacy DatagramSocket API'

[Daniel Fuchs]

Hi Alan,

On 12/04/2020 16:51, Alan Bateman wrote:
> I don't think SetGetSendBufferSize should be in the webrev. It has @bug 
> 8239355 so it's for a different issue.

Patrick will confirm but IIRC this test was added because coverage
testing reported that DS::setSendBufferSize was never called (presumably
existing tests might have been calling setOption instead).
@bug 8239355 is probably a copy/paste error and can be removed.

> I also don't think SendBufCheck should be in the webrev (btw: it also 
> has @bug 8239355). The underlying issue is that PDSI doesn't allow 
> sending IPv6 datagrams on macOS of size 65508-65527 (inclusive) bytes 
> without changing the default value of SO_SNDBUF. It's not an interesting 
> issue of course but I don't think the tests for the JEP should be adding 
> a test to make sure that the old implementation has this bug.

That's where I disagree because it's a slight difference in behavior
between the old implementation and the new - so I believe it's worth
having a test for it (even if does test the presence of a bug in the
old impl). I like this test because it does show how
every implementation behaves in one single place without having
to forage in every other test hierarchy to figure what their
behaviors are. The branch is the only place where you could have
such a test (since the new impl is only available here).

> Can't the 
> PDSI issue be fixed with a different issue? Alternatively adding a test 
> that checks that IPv6 datagrams of at least size 65507 can be sent on 
> configurations that have IPv6 enabled.

Sure - but is it even worth fixing?

best regards

-- daniel