Re: [PATCH] for Review: Fix for fix for PR 13102

[Mohan Embar <gnustuff at thisiscool dot com>]

Hi Michael,

>I really wonder why people send some patches just before I can finish
>the same work ?

Just trying to cause trouble. You haven't seen the end of me yet....

>> I ran into problems when running your patch with
>> my network test. I think the problem is due
>> to the use of s.impl in ServerSocket.implAccept()
>> given that impl hasn't been initialized yet.
>> However, you made Socket.getImpl() private.
>
>Problems compiling ?

>This solution really confuses more then it helps. IMO the bast way is
>one getImpl() and the appropriate code everywhere using it.

getImpl() needs to be at least package-private in order for
ServerSocket to get at it. But if you make it package-private,
it conflicts with NIOSocket.getImpl(). That's why I renamed it.
But I also found it unnecessary that five methods caught
and ate an exception when that can be done in a helper
method.....

>> With this patch, my networking test now passes
>> with the (linux,mingw) cross compiler, but I
>> don't have a Linux native compiler to test with.
>> (I could build one if you want.)
>
>I'm only testing with Mauve. You should really put your tests into
>Mauve to let other people share your work.

Apparently this slipped by Mauve then. I'm sorry
I haven't looked into Mauve yet, but the barrier for me
is higher since to get Mauve to work with Windows, I have to
do all kinds of FTP and Telnet daemon things that I haven't
had the inclination to do yet.

However, my testsuite is published with my build and
as of the last build, my testsuite runs on Linux too. That
means that the legions of Open Source hackers can take
these and port them to Mauve anytime they want to.

It looks like you're in this code already, so let me know if
and how you want me to follow up on this.

-- Mohan
http://www.thisiscool.com/
http://www.animalsong.org/