Community
Participate
Working Groups
frequent DNF's (Did Not Finish) in this test suite in the 3.3 builds
Comment from Ian: David, It would be very helpful to me if you could capture a dump so we could try to determine what the JSF core test suite is getting hung up on – it runs consistently green on my system, and I have been unable to determine what the issue could be. Thanks, - Ian
Looking at the recent history, appears this can happen in following suites. It seems when they run, it only takes a few minutes. org.eclipse.jst.jsf.core.tests.AllTests 169 103.296 org.eclipse.jst.jsf.context.symbol.tests.AllTests 62 60.467
Created attachment 180918 [details] 4 javacore dumps Here's 4 dumps ... each taken a few minutes apart ... after the jsf.core tests had been running about an hour. I took a quick glance at first two, and appears that some kind of work is going on (not literally a frozen hang). Perhaps its one of those "deadlock breaking" algorithms that keeps interrupting each other, but not really making progress?
I also wanted to add, I run the builds and tests on a "local" machine (mostly to test build scripts) but the tests always seem to finish there. (there's a few consistent errors, so not sure that machine is set up right to test jsf ... but they do finish each time (in the 5 or 6 I still had as history on that machine). So, what's different on build.eclipse.org? The tests (on production machine only) are (now) signed. That can effect classloading ... but seems that would just make the tests consistently longer. The other difference is the "display" set up. (would that effect "core" tests?) In the logs on production machine, I see a several messages about Xlib: extension "RANDR" missing on display "127.0.0.1:1001.0". (what ever that means). Also, there's many messages in there that say [Fatal Error] :-1:-1: Premature end of file. I assume that's not normal, and you don't see locally? Not sure where that comes from? reading a tag lib or something?
Thanks for the dumps and the info, David - I shall attempt to decipher what is happening. As for "[Fatal Error]...", I found it to be coming out of org.eclipse.wst.server.core.internal.ServerPreferenceModifyListener.removeLockedServerRuntimePreference(Preferences), and it has been present in JSF Tools and adopter testing for many weeks. What's very strange is that I was sure (positive!) I had logged a bug, but now I cannot find it...and I'm not *that* old. I'll log one now.
Logged: Bug 327849 - ServerPreferenceModifyListener producing "[Fatal Error]..." log output, but there are no actual perceived failures.
Mass Triage: Deferred from Indigo
Changing the Target Milestone to future
New Gerrit change created: https://git.eclipse.org/r/114253
New Gerrit change created: https://git.eclipse.org/r/114259
Gerrit change https://git.eclipse.org/r/114259 was merged to [master]. Commit: http://git.eclipse.org/c/jsf/webtools.jsf.git/commit/?id=9addae63cf9222e5553204eee908430540d02f25
Gerrit change https://git.eclipse.org/r/114253 was merged to [master]. Commit: http://git.eclipse.org/c/jsf/webtools.jsf.git/commit/?id=a4bee171911548ad140497b9c7a203907d322233
