Tidy up write side.
It cannot be the same as the read side as there is no registration for write interest. There is only write operations and the
callbacks associated with them.
Fixed by calling tryFillInterested() rather than fillInterested() to
cope with the race between reads scheduling read interest and
setWriteListener() that also executes code in
HttpChannelState.unhandle() that wants to schedule read interest.
+ Reverting change to HttpTester.parseResponse(Input)
+ Providing new HttpTester.parsePartialResponse(Input)
+ InsufficientBytes tests no longer assert content strings with invalid
characters (this was breaks the surefire report xml)
+ Upgrading to jetty-test-helper 4.0
+ Removing use of org.eclipse.jetty.toolchain.test.SimpleRequest
+ Removing use of org.eclipse.jetty.toolchain.test.http.SimpleHttpParser
+ Removing use of org.eclipse.jetty.toolchain.test.http.SimpleHttpResponse
+ Updating long since deprecated (and now removed) known quirky methods
in jetty-test-helper and the test classes.
* Updated Maven plugins to versions that support JDK 9.
* Added jdk9 profiles to the build files.
* Introduced modules jetty-alpn-java-client and jetty-alpn-java-server
containing a pure JDK 9 implementation of ALPN.
* Wired ALPN connection factories (client and server) to use the proper
ALPN implementation based on the JDK platform version (8 or 9).
The fix notifies the transport when a reset frame is received,
allowing the transport to fail the write callback which then notifies
the application, either by throwing (in case of blocking writes) or
by calling error listeners.
Also added a guard, in HttpChannel.handle() for the ERROR_DISPATCH case,
that checks if the response is already committed, and if so, abort
the transport - similar to what's already there for 9.4.
Introduced isAllowed(InetAddress, HttpServletRequest) so that it
would be possible to make access decisions also based on request
information such as context, path, headers, etc.
