Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [eclipse-dev] Automated test results delayed

Thanks Sonia.

The browser that is left opened is not the same VM as Eclipse, and is being
run as a separate process.  The communication between Eclipse process and
browser process occurs through standard input and output.  The process is
being sent a command to close, and the command is not executed.  I think
there is either a deadlock in interprocess communication caused by a change
in a setup, or there is not enough resources on the machine to run the
other process smoothly.  Performing the tests that failed takes over 20MB,
6 new java thread, plus any native threads that the embedded IE internally
creates.  If there is any output from the browser process it is being
collected in .metadata\.plugins\\browser.log, not the main
Eclipse log.

The reason tests block after the failure is that Eclipse process refuses to
exit after tests are finished.  All threads in Eclipse used to communicate
with browser process are daemons (I ensured this yesterday) and die when
Eclipse VM exits, but for some reason Eclipse process does not return with
a child process left alive.

If you have time to pursue this issue further I would suggest the

1. Check .metadata\.plugins\\browser.log
2. Reboot machine and run only help tests.
3. Run eclipse manually, launch help, close Eclipse without closing help
window.  The window should close, and all eclipse/java processes should
4. Enable debugging options: = true
Repeat help test, or scenario 3. with -debug.  Send us

I think, as long as there is no test failures reported with the build, the
results of this investigation are not interesting for all eclpse-dev
subscribers.  Let's take further discussion off-line.

Konrad Kolosowski
Eclipse Help System

                      Dimitrov/Ottawa/IB        To:       eclipse-dev@xxxxxxxxxxx                                                       
                      M@IBMCA                   cc:                                                                                     
                      Sent by:                  Subject:  Re: [eclipse-dev] Automated test results delayed                              
                      03/19/2003 10:06                                                                                                  
                      Please respond to                                                                                                 

When I try rerunning the help tests via an rsh connection from our build
machine, I can reproduce the "hang".  I have attached the contents of the
.log file and the console output.  Help team, please advise.


     [echo] Running
     [java] ----- testIEBrowser
     [java] testIEBrowser: setUp...
     [java] testIEBrowser: tearDown...

     [java] ----- testLiveHelp
     [java] testLiveHelp: setUp...
     [java] testLiveHelp: tearDown...


   Roberts/Ottawa/IBM@IBMCA             To:                               
   Sent by:                     eclipse-dev@xxxxxxxxxxx                   
   eclipse-dev-admin@xxxxxxxxxx         cc:                               
   g                                    Subject:        [eclipse-dev]     
                                Automated test results delayed            
   03/19/2003 08:15 AM                                                    
   Please respond to                                                      

Hi folks,

The automated test results are delayed again.

When I came in this morning the Windows Help tests where "paused" with an
open window.  Manually closing the window has resumed the tests.  They will
be posted as soon as they are available.  I anticipate about 90 minutes.
However, the build is available for you to start your sniff tests on at:

While this is not very informative, here is a screen shot of the state of
the Windows test machine when I got in.  I also included a screen shot
showing version info for IE on that computer.  The colour shift is not
important.  Just downsizing the colour depth of the screen capture.

Let me know if I can provide any more info.


#### ieversion.JPG has been removed from this note on March 19, 2003 by
Sonia Dimitrov
#### hang1.JPG has been removed from this note on March 19, 2003 by Sonia
(See attached file: .log)

Attachment: .log
Description: Binary data

Back to the top