Re: [platform-releng-dev] 4.6.0 N-Build: N20160116-1500: 20 failures fro

["David M Williams" <david_williams@xxxxxxxxxx>]

> Out of curiosity, has anybody looked at configuring
the surefire jobs to take a screenshot on failed tests?

Not that I know of. But, FYI, we do not use "surefire"
during our production build tests. Just Ant to invoke the JUnit tests on
an SDK. 

We do, though, have screen shots taken when there is a
"timeout" of a test. See 
https://wiki.eclipse.org/Platform-releng-faq#Troubleshooting_tests_that_crash_or_time_out.2C_aka_.22.28-1.29DNF.22

I assume you know to look at "detailed logs"?
Such as, 
http://download.eclipse.org/eclipse/downloads/drops4/N20160116-1500/testresults/macosx.cocoa.x86_64_8.0/org.eclipse.e4.ui.tests.UIAllTests.txt

Some people write "extra data" when there is
test failure, that can help debug "remote" tests (without hooking
up a remote debugger). 

One thing I noticed in the detailed log (referenced above)
is a a message about "Failed to persist contents
of part". Is that a normal part of the test?
Or an indication of a problem? If unexpected, would that then cause other
tests to fail? 

We have, in the past, had troubles with running out of
disk space on that Mac. Usually there is a huge failure if that happens,
but if we are (were) "on the edge" of being out of space, perhaps
that would account for some of your observations? If you think so, you
might want to add some diagnostic info if the same test fails intermittently
that prints out "disk in use" at the time of the failure. (File.getTotalSpace(),
... .getFreeSpace() ... .getUsableSpace()). 

HTH





From:      
 Brian de Alwis <briandealwis@xxxxxxxxx>
To:      
 "Eclipse platform
release engineering list." <platform-releng-dev@xxxxxxxxxxx>,
Date:      
 01/18/2016 03:06 PM
Subject:    
   Re: [platform-releng-dev]
4.6.0 N-Build: N20160116-1500:        20  
     failures from ep46N-unit-mac64
Sent by:    
   platform-releng-dev-bounces@xxxxxxxxxxx

---

Sorry David, I should have been explicit.  The nightly
builds have been failing, particularly on OS X but sometimes Linux, with
the Platform/UI tests.  The N20160116-1500 build was particularly
bad (see below), but then they go away.

Out of curiosity, has anybody looked at configuring the
surefire jobs to take a screenshot on failed tests?

Brian.


org.eclipse.e4.ui.tests.workbench.PartRenderingEngineTests.testBug372226
org.eclipse.e4.ui.tests.workbench.MWindowTest.testCreateWindow
org.eclipse.e4.ui.tests.workbench.MWindowTest.testContextChildren
org.eclipse.e4.ui.tests.workbench.PartFocusTest.testFocusChangesOnExplicitPartActivation
org.eclipse.e4.ui.tests.workbench.PartFocusTest.testNoActivationOnExplicitInPartWidgetSelection
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInActivePartOnly
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInWindowOnly
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerSwitchToInactivePartInOtherPerspectiveWithPartHandlers
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInActivePerspectiveAndWindow
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInActivePerspectiveOnly
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerSwitchToInactivePartInOtherPerspectiveWithPerspectiveHandlers
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInActivePartAndPerspectiveAndWindow
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerSwitchToInactivePart
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInActivePartAndWindow
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerInActivePartAndPerspective
org.eclipse.e4.ui.tests.workbench.HandlerActivationTest.testHandlerSwitchToInactivePerspective


On 17-Jan-2016, at 10:22 PM, David M Williams <david_williams@xxxxxxxxxx>
wrote:


> Date: 01/17/2016 07:49 PM
> Subject: Re: [platform-releng-dev] 4.6.0 N-Build: N20160116-1500:
20
> failures from ep46N-unit-mac64
> Sent by: platform-releng-dev-bounces@xxxxxxxxxxx
> 
> These build failures are baffling (and unreproducible) 

= = = = = = 

> and have been happening for a while.  

Happening for a while? N20160116-1500 is the first time 
I have seen the '16 failures' in e4.ui.tests. 
Or do you mean "every now and then, for a while"? 
You are not confusing these with bug 484441 are you? (It was 
fixed as of last Tuesday). 

= = = = = = 

> Is there something about the Mac test machine? 

Yes. It's always something. 

Usually something that requires a restart 
(such as an "update" or something similar). 
Occasionally some "run away test" is still running from some
previous run. There are various diagnostics and screen shots "captured"
during the 
tests but I can not find where "we" (me and Markus) wrote how
to make use of 
those, so I will have to try looking again later. 

= = = = = = 

Eclipse webmasters automatically restart Hudson on Sunday, early AM, but
am not sure 
if that includes the Mac slave, or not. (I know it does not include the
Windows 
slave.) 

So if we see a couple of times in a row, we typically open a bug for the
webmaster to
"take a look at it" and see if there is a prompt open on the
desktop or an erroneous 
process running. 

= = = = = = 

If you mean something besides anything I have mentioned, the normal procedure
is to 
open a bug so all the right people can get involved to track it down. 
Or, if I have completely misunderstood what you are asking about, please
ask again. 

= = = = = = 

Thanks, 


_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev