ceylon.net tests hang forever if server can't be started

[quintesse]

I already had something running on port 8080 so the following happened when running the SDK tests:

.
.
.
[ceylon-test] running: test.ceylon.net::testPathMatcher
[ceylon-test] running: test.ceylon.net::testServer
[ceylon-test] Starting on 127.0.0.1:8080
[ceylon-test] Exception in thread "Thread-123" java.net.BindException: Address already in use
[ceylon-test]   at sun.nio.ch.Net.bind0(Native Method)
[ceylon-test]   at sun.nio.ch.Net.bind(Net.java:433)
[ceylon-test]   at sun.nio.ch.Net.bind(Net.java:425)
[ceylon-test]   at sun.nio.ch.ServerSocketChannelImpl.bind(ServerSocketChannelImpl.java:223)
[ceylon-test]   at sun.nio.ch.ServerSocketAdaptor.bind(ServerSocketAdaptor.java:74)
[ceylon-test]   at sun.nio.ch.ServerSocketAdaptor.bind(ServerSocketAdaptor.java:67)
[ceylon-test]   at org.xnio.nio.NioXnioWorker.createTcpConnectionServer(NioXnioWorker.java:182)
[ceylon-test]   at org.xnio.XnioWorker.createStreamConnectionServer(XnioWorker.java:243)
[ceylon-test]   at ceylon.net.http.server.internal.DefaultServer.start$canonical$(DefaultServer.ceylon:175)
[ceylon-test]   at ceylon.net.http.server.internal.DefaultServer.start(DefaultServer.ceylon:139)
[ceylon-test]   at ceylon.net.http.server.internal.DefaultServer$1httpd_.run(DefaultServer.ceylon:206)
[ceylon-test]   at java.lang.Thread.run(Thread.java:745)

And the process never stops, the tests never finish.

[quintesse]

@matejonnet is this something you can take a look at? Perhaps Status needs to be extended with an error? Or always signal stopped at least. Or perhaps skip waitTestToComplete() if the server is in an invalid state somehow. You probably know best :)

[quintesse]

Moving to 1.2.3

[quintesse]

@matejonnet any ideas you could give us? This issue continues to exist and is a bit nasty.

[matejonnet]

This should solve the issue https://github.com/ceylon/ceylon-sdk/pull/619

[quintesse]

It does, thanks!

[quintesse]

So it seems this was only partly fixed, because on the CI server we still have the same problem:

.
.
.
[ceylon-test] running: test.ceylon.http.client::RequestLocalServerTest
[ceylon-test] running: test.ceylon.http.client::RequestLocalServerTest.defaultGet
[ceylon-test] Starting on 127.0.0.1:59746
[ceylon-test] Failed to start server.
[ceylon-test] Exception in thread "Thread-143" ceylon.http.server.ServerException "Failed to start server."
[ceylon-test]   at ceylon.http.server.internal.DefaultServer.start$canonical$(DefaultServer.ceylon:190)
[ceylon-test]   at ceylon.http.server.internal.DefaultServer.start(DefaultServer.ceylon:126)
[ceylon-test]   at ceylon.http.server.internal.DefaultServer$1anonymous_0_.run(DefaultServer.ceylon:215)
[ceylon-test] Caused by: java.net.SocketException: Permission denied
[ceylon-test]   at sun.nio.ch.Net.bind0(Native Method)
[ceylon-test]   at sun.nio.ch.Net.bind(Net.java:463)
[ceylon-test]   at sun.nio.ch.Net.bind(Net.java:455)
[ceylon-test]   at sun.nio.ch.ServerSocketChannelImpl.bind(ServerSocketChannelImpl.java:223)
[ceylon-test]   at sun.nio.ch.ServerSocketAdaptor.bind(ServerSocketAdaptor.java:74)
[ceylon-test]   at sun.nio.ch.ServerSocketAdaptor.bind(ServerSocketAdaptor.java:67)
[ceylon-test]   at org.xnio.nio.NioXnioWorker.createTcpConnectionServer(NioXnioWorker.java:182)
[ceylon-test]   at org.xnio.XnioWorker.createStreamConnectionServer(XnioWorker.java:243)
[ceylon-test]   at ceylon.http.server.internal.DefaultServer.start$canonical$(DefaultServer.ceylon:181)
[ceylon-test]   ... 2 more
Build was aborted
Aborted by Jenkins Admin
Finished: ABORTED

And after this error the test again hangs forever. So I'm reopening this.

(Btw, the part about "Aborted by Jenkins Admin" was me, like 2 hours after the tests had been started. So it obviously wasn't going to recover by itself)

[quintesse]

As an aside, I think this code related to the problem that @matejonnet improved the last time:

DefaultServer.ceylon:Line 187

} catch (IOException e) {
    print("Failed to start server.");
    notifyListeners(stopped);
    throw ServerException("Failed to start server.", e);
}

should perhaps be changed to catching any Exception, not just IOExceptions. Because if there's some other, runtime, problem then we'd want to listeners to get notified as well, right?

[matejonnet]

Sounds right to catch all.