In message <4bf4f9a4.6050...@gmail.com>, Regis writes: > > On 2010-05-20 16:46, Mark Hindess wrote: > > > > In message<4bf4ec09.3090...@gmail.com>, Regis writes: > >> > >> On 2010-05-19 20:45, Catherine Hope wrote: > >>> I also see these 2 tests failing in the same way. It seems to be > >>> caused by Regis' commit 944119 "SocketChannelImpl.SocketAdapter's > >>> remote address should be updated after channel connected" on > >>> 14/05/2010. It could be that the tests are invalid though - > >>> the first one is checking that the SocketAddress returned by > >>> getRemoteAddress() is a different object to the one that was passed > >>> to connect(), though as SocketAddress is immutable would this cause > >>> a problem? > >>> > >> > >> The test case expect behaviors to be exactly same with RI, but I don't > >> think it's a big deal, I intend to delete this assert. > >> > >> for - testSocket_NonBlock_BasicStatusAfterConnect, isConnected() > >> should check first, following patch can fix it. And again, the test > >> failed at same line as BasicStatusAfterConnect. > >> > >> Following patch can fix these failures, I'd like to commit it if > >> second committer agree this. > > > > -1 > > > > I am +1 for the patch to SocketChannelImpl.java because this fixes > > behaviour. > > > > However, I am -1 for the patch to SocketChannelTest.java because we > > don't fix broken tests during code freeze. > > All right, how about this (create new InetSocketAddress instance everytime):
-1 Fix the behaviour for testSocket_NonBlock_BasicStatusAfterConnect. Do not make any changes for the other failure - i.e. leave it failing - and remove it after the code freeze. -Mark.
