Community
Participate
Working Groups
N20070408-0010 Several performance test suites are marked with DNF but there is no explanation for these DNFs. I would expect at least: - the test(s) that DNFed - the reason of the DNF (valued of the timeout, test had to be killed, etc.) (Marking this as critical since the same tests DNFed for several performance runs now, and no progress can be made on fixing this without more information).
*** Bug 181724 has been marked as a duplicate of this bug. ***
DNFs mean -the tests hung and the vm timed out -the VM crashed -other issues The end result is that the JUnit test result file is not generated in xml format and therefore the xml cannot be transformed into html for the build results web page. There isn't a way to programmically determine the reason for the DNF. As for the jdtcore tests in this build Search for Running org.eclipse.jdt.core.tests.performance.AllPerformanceTests in http://download.eclipse.org/eclipse/downloads/drops/N20070408-0010/performance/consolelogs/linux.gtk.perf_consolelog.txt and you can see that the tests timed out. Here is the log associated with that specific test suite http://download.eclipse.org/eclipse/downloads/drops/N20070408-0010/performance/linux.gtk.perf/org.eclipse.jdt.core.tests.performance.AllPerformanceTests.txt For the jdt core linux perf tests, it looks like the tests didn't complete either. Search for Running org.eclipse.jdt.core.tests.performance.AllPerformanceTests in http://download.eclipse.org/eclipse/downloads/drops/N20070408-0010/performance/consolelogs/winxp.perf_consolelog.txt See the log http://download.eclipse.org/eclipse/downloads/drops/N20070408-0010/performance/win32perf/org.eclipse.jdt.core.tests.performance.allperformancetests.txt In both cases, it looks like the tests hung on Running FullSourceWorkspaceModelTests#testCloseProjects... Does this help? Were you able to replicate this problem locally on similar hardware - especially the same amount of RAM? The machines in question are linux.gtk.perf RHEL 3.0 Sun 1.4.2_10 (2 GHz 512 MB) win32perf Win XP Sun 1.4.2_10 (2 GHz 512 MB)
Kim, What is the value for the time out? Is this timeout a time limit to run all the JDT/Core tests (ie. all those run with AllPerformanceTests.xml Ant script) or is this a time limitation for each test? Looking at time values for all JDT/Core tests, I guess the answer is the first one... So, as we added some new tests and increase time of some existing ones, it seems that we have now reached the limit on the slowest boxes... A possible fix could be that timeout was a little higher on slow boxes.
The default timeout is 7200000 milliseconds or 2 hours in the libary.xml of org.eclipse.test. You can override this in your test plugin's test.xml.
OK, so back in JDT/Core land to modify the timeout value...
Add 30mn to global timeout should fix the problem. Done by adding following line: <property name="timeout" value="9000000" /> to "performance-suite" target in org.eclipse.jdt.core.tests.performance/test.xml file... Release for 3.3 M7 in HEAD stream and also in perf_32x branch
I fail to understand : - how this bug has changed title (the fact that there is no explanation of DNF has not been fixed) - how we can take more that 2 hours to index the full source workspace
(In reply to comment #7) > - how we can take more that 2 hours to index the full source workspace I take that back. I now understand that the whole suite takes more than 2 hours. So Frederic's change seems reasonable.
(In reply to comment #7) > I fail to understand : > - how this bug has changed title (the fact that there is no explanation of DNF > has not been fixed) Since this bug has been hijacked, Frederic can you please open new bug against releng to request that the DNF should be explained with at least some the timeout value.
(In reply to comment #9) > (In reply to comment #7) > > I fail to understand : > > - how this bug has changed title (the fact that there is no explanation of DNF > > has not been fixed) > Since this bug has been hijacked, Frederic can you please open new bug against > releng to request that the DNF should be explained with at least some the > timeout value. > Regarding the severity of this bug, I assigned this bug to me immediately as soon as I found the explanation looking at the log file indicated by Kim in comment 2. I wanted this problem to be fixed for this week-end performance tests run. As Kim already said there's no way to add explanation when tests DNF, I thought it wasn't necessary to open a bug for this. The only remaining question I have is: how can I find easily the link to this kind of page? I tried to find it everywhere in performance results HTML pages and wasn't able to find any. Kim, can you help on this point?
Entered bug 184437 against releng to show the value of the timeout.
Verified for 3.3M7 with I20070427-0010