Re: [platform-releng-dev] Stacktrace for DNFs in consolelog

[David M Williams <david_williams@xxxxxxxxxx>]

Well, even better, at least for us in WTP, since we do already use "your" test framework ... we will get this (nearly) for free once we move up to a build that contains it. Thanks! Its not often releng makes the "New and Noteworthy"! :) but I think this is truly noteworthy. It will be a great help.

Thanks!



Markus Keller ---04/20/2011 01:37:07 PM---The patch at https://bugs.eclipse.org/bugs/attachment.cgi?id=192770&action=""> pretty

From: Markus Keller <markus_keller@xxxxxxxxxx>
To: "Eclipse platform release engineering list." <platform-releng-dev@xxxxxxxxxxx>
Date: 04/20/2011 01:37 PM
Subject: Re: [platform-releng-dev] Stacktrace for DNFs in consolelog
Sent by: platform-releng-dev-bounces@xxxxxxxxxxx

---

The patch at https://bugs.eclipse.org/bugs/attachment.cgi?id=192770&action="">

pretty much sums it up:

In the Eclipse test framework, the EclipseTestRunner parses the command line and invokes JUnit in the test VM.
I just added another command line argument "-timeout <millis>", and then I start a deamon thread that dumps the stack 30s before the timeout elapses.
In our case, the timeout is given by the Ant script that calls <java fork="true" timeout="${timeout}" ...>.

I've been thinking about this for quite some time, and I would have loved to put it into the Ant task itself (the one which kills the sub-process after the timeout). Unfortunately, I found no good way to do that. Since you cannot easily add code to the target VM from the outside, I wanted to use the jstack executable to get the dumps. But it's not easy to find this executable on all platforms (might not even be present if you're just running with a JRE), and there's also no official way to get the process ID from a java.lang Process.

An alternative technique could be to just wrap the main class, i.e. replace it with another class that sets up the timer, removes the timeout argument from the command line, and then calls the original main class.

HTH,
Markus

From:	David M Williams <david_williams@xxxxxxxxxx>
To:	"Eclipse platform release engineering list." <platform-releng-dev@xxxxxxxxxxx>
Date:	2011-04-20 17:14
Subject:	Re: [platform-releng-dev] Stacktrace for DNFs in consolelog

---

Hey Markus, do you have any code/scripts/techniques/pointers that others could use to "get stackdumps shortly before a plug-in test VM gets killed in automated tests". This is something I've always wanted to do in WTP, but never had the time ... so, if you have any time-saving tips, that'd be great! I'm sure for many.

Thanks,


Markus Keller ---04/20/2011 10:59:03 AM---As a side effect of the efforts to track down https://bugs.eclipse.org/342053 , you now get stackdu

From: Markus Keller <markus_keller@xxxxxxxxxx>
To: "Eclipse platform release engineering list." <platform-releng-dev@xxxxxxxxxxx>
Date: 04/20/2011 10:59 AM
Subject: [platform-releng-dev] Stacktrace for DNFs in consolelog
Sent by: platform-releng-dev-bounces@xxxxxxxxxxx

---

As a side effect of the efforts to track down https://bugs.eclipse.org/342053 , you now get stackdumps shortly before a plug-in test VM gets killed in automated tests.

E.g. for the failing org.eclipse.debug.tests_win32.win32.x86_6.0 in the N-build, the Console Output Log
http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000/testresults/consolelogs/win32-6.0_consolelog.txt
now contains:

  [echo] Running org.eclipse.debug.tests.AutomatedSuite. Result file: C:\buildtest\N20110419-2000\eclipse-testing/results/win32.win32.x86_6.0/org.eclipse.debug.tests.AutomatedSuite.xml.
 [java] EclipseTestRunner almost reached timeout '7200000'.
 [java] Thread dump at 2011-04-20 06:17:47 -0400:
...
 [java] java.lang.Exception: main
 [java]                  at org.eclipse.swt.internal.win32.OS.WaitMessage(Native Method)
 [java]                  at org.eclipse.swt.widgets.Display.sleep(Display.java:4650)
 [java]                  at org.eclipe.debug.tests.viewer.model.ContentTests.testLabelUpdatesCompletedOutOfSequence1(ContentTests.java:191)
 [java]                  at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
 [java]                  at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
 [java]                  at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
 [java]                  at java.lang.reflect.Method.invoke(Method.java:597)
 [java]                  at junit.framework.TestCase.runTest(TestCase.java:168)
 [java]                  at org.eclipe.debug.tests.viewer.model.ContentTests.runTest(ContentTests.java:84)
...
  [java] FocusControl:
 [java] Tree {}
 [java] Shell {}
 [java] Shells:
 [java] visible: Shell {Java - Eclipse SDK}
 [java] visible: Shell {}

While this doesn't explain why the tests hung up, it at least gives a hint at where to start investigating, and it lists the Focus Control (together with its ancestors) and the open Shells. The whole dump is repeated 2 seconds later. Since the code that prints the SWT infos uses "display.syncExec(..)", this may or may not influence a hung main thread.

Note that the org.eclipse.test plug-in has not been tagged yet with this change, that's why you only see this in the N-build.

Markus


BTW: The jdt.ui.tests failures have been fixed in HEAD.

From:	Kim_Moir@xxxxxxxxxx
To:	platform-releng-dev@xxxxxxxxxxx
Date:	2011-04-20 13:48
Subject:	[platform-releng-dev] [eclipse-build]Build N20110419-2000 (Timestamp: 201104192000):Build is complete. Test failures/errors occurred.

---

Build N20110419-2000 (Timestamp: 201104192000):  Automated JUnit testing is complete.  Test failures/errors occurred in the following:  

org.eclipse.debug.tests_win32.win32.x86_6.0
org.eclipse.jdt.ui.tests_linux.gtk.x86_6.0
org.eclipse.jdt.ui.tests_macosx.cocoa.x86_5.0
org.eclipse.jdt.ui.tests_win32.win32.x86_6.0
org.eclipse.ui.tests_win32.win32.x86_6.0


HTTP Download:

http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000 


Test result files (hint: copy/paste or drag-and-drop to the JUnit view):

http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000/testresults/xml/org.eclipse.debug.tests_win32.win32.x86_6.0.xml
http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000/testresults/xml/org.eclipse.jdt.ui.tests_linux.gtk.x86_6.0.xml
http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000/testresults/xml/org.eclipse.jdt.ui.tests_macosx.cocoa.x86_5.0.xml
http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000/testresults/xml/org.eclipse.jdt.ui.tests_win32.win32.x86_6.0.xml
http://download.eclipse.org/eclipse/downloads/drops/N20110419-2000/testresults/xml/org.eclipse.ui.tests_win32.win32.x86_6.0.xml

_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev

_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev

_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev