DefaultHttpProxy leaves infinite alive non-daemon threads after stopping about littleproxy

[nasis]

I'm having exactly the same problem. Sample code when used with WebDriver. I've tested that both on 4.0 and 5.0 SNAPSHOT

public class GroupExample {

public static void main(String[] args) throws InterruptedException {

    int port = 9090;
    HttpProxyServer proxyServer = new DefaultHttpProxyServer(port);
    proxyServer.start();

    Proxy proxy = new Proxy();
    proxy.setProxyType(Proxy.ProxyType.MANUAL);
    String proxyStr = String.format("localhost:%d", port);
    proxy.setHttpProxy(proxyStr);
    proxy.setSslProxy(proxyStr);

    DesiredCapabilities capability = DesiredCapabilities.firefox();
    capability.setCapability(CapabilityType.PROXY, proxy);

    String urlString = "http://www.yahoo.com/";
    WebDriver driver = new FirefoxDriver(capability);
    driver.manage().timeouts().pageLoadTimeout(30, TimeUnit.SECONDS);

    driver.get(urlString);

    driver.close();
    System.out.println("Driver closed");

    proxyServer.stop();
    System.out.println("Proxy stopped");
}

}

[isopov]

nasis, I used http://code.google.com/p/proxoid/source/browse/trunk/#trunk%2Fproxylight%2Fsrc%2Fcom%2Fmba%2Fproxylight proxy for tests. Maybe it can be useful to you too.

[nasis]

Thanks. It looks to simple though. I need to record the resources content as well. Looks like I'll need to add lots more code.

[myleshorton]

I'm not able to reproduce this with the 0.5-SNAPSHOT guys. I added the commits under #37 by mistake, but basically converted your tests to junit in EndToEndStoppingTest.java. Both the HttpClient version and the WebDriver version stop fine in those tests.

[nasis]

Why no try the main program as it is without JUnit?
Lets see how that goes.
Where you able to reproduce it on 0.4? Probably the same issue.

On Wed, Aug 29, 2012 at 10:23 PM, myleshorton [email protected]:

I'm not able to reproduce this with the 0.5-SNAPSHOT guys. I added the
commits under #37 #37 by
mistake, but basically converted your tests to junit in
EndToEndStoppingTest.java. Both the HttpClient version and the WebDriver
version stop fine in those tests.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/36#issuecomment-8137734.

[isopov]

I have debugged your Unit Test. The problem is present there, but is not visible. What is going on, if we run Unit tests:

1. First test creates Proxy and stops it. But several non-daemon threads are left after this proxy.
2. Second tests creates Proxy and stops it. But another pack of non-daemon threads are left after this proxy.
3. After all tests pass - Junit runner calls System.exit(0) and all threads are stopped.

If we run the same logic in main() method without calling System.exit(0) JVM will not finish it execution after Proxy.stop(). I have written simple illustration for this logic here: https://gist.github.com/3524700

So I think the issue should be renamed in "DefaultHttpProxy leaves infinite alive non-daemon threads after stopping".

[nasis]

Well at least we agree that there is a problem.
I don't know how to change the subject of a post but if you want I can
repost on a new name.
However we will then probably loose all these comments.

On Thu, Aug 30, 2012 at 12:19 PM, Ivan Sopov [email protected]:

I have debugged you Unit Test. The problem is present here, but is not
visible. What is going on, if we run Unit tests:

1. First test creates Proxy and stops it. But several non-daemon threads
   are left after this proxy.
2. Second tests creates Proxy and stops it. But another pack of non-daemon
   threads are left after this proxy.
3. After all tests pass - Junit runner calls System.exit(0) and all
   threads are stopped.

If we run the same logic in main() method without calling System.exit(0)
JVM will not finish it execution after Proxy.stop(). I have written simple
illustration for this logic here: https://gist.github.com/3524700

So I think the issue should be renamed in "DefaultHttpProxy leaves
infinite alive non-daemon threads after stopping".

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/36#issuecomment-8153685.

[adamfisk]

Yup seeing that now guys -- thanks. Looking into it.

[adamfisk]

This should be all set now guys. I also added your straight main(...) test to the EndToEndStoppingTest to test it that way too.

[adamfisk]

Hmn...now the HttpClient test actually fails, I think because LittleProxy doesn't close the connection to the client following HTTP 1.1 rules. Reopening.

[adamfisk]

OK - this should be good to go. I keep mistakenly committing to #37, but the commit that fixes this is 93628b4

[adamfisk]

This is also deployed to the latest snapshot -- thanks for the help guys.