Community
Participate
Working Groups
3 of them do, that is. testOpenAntEditor1 testOpenAntEditor2 testOpenAntEditorNoFolding They do this in both streams, baseline, or "current". If they can not be fixed easily, I suggest they be commented out.
Created attachment 249503 [details] XML file that can be imported into JUnit. I've looked at the test cases, and the common problem is in super class, that has says: wait(2000); // NOTE: runnables posted from other threads, while the main thread waits here, are executed and measured only in the next // iteration So, problem number one: the method it is in is protected void measureOpenInEditor(IFile file) and it is not synchronized, so it will never "own" the monitor. Problem number two: I do not see anywhere in the class in question a "notifiy" counter part. So, even if made synchronized, I think it'd just always wait for 2 seconds, and then continue. Doubt if that's the intent. I wonder if this method was "copied" from elsewhere? And is basically incomplete?
I have found there are similar method in jdt.txt.performance.tests, but, different ... they have some mechanism to "runEventQueue" that might be to serve similar purpose? But, I've also found, on my local test machine, that adding "synchronized" to the method allows the test to run, and I presume still relatively valid results. One of them does provide to the "global finger print" graph, so assume they are important tests. Will attach patch for the synchronized solution.
Created attachment 249532 [details] Simple fix. Patch clarifies exactly what I mean.
(In reply to David Williams from comment #3) > Created attachment 249532 [details] > Simple fix. Patch clarifies exactly what I mean. Looks like we do need to Synchronize this method !!
Pushed to master: http://git.eclipse.org/c/platform/eclipse.platform.git/commit/?id=eccd2fea8b3928a5989600740e26eccc71a8df49 Pushed to 4.4.2: Bundle version update:
(In reply to Michael Rennie from comment #5) > Pushed to 4.4.2: > > > > Bundle version update: Nothing was pushed to 4.4.2, I just forgot to delete that part of the comment when I was changing the bug...
I think we should fix this for RC3, so we have at least some measurement of this test case at at the end of 4.2.2 ... more to "compare to" in future.
(In reply to Michael Rennie from comment #5) > Pushed to master: > > http://git.eclipse.org/c/platform/eclipse.platform.git/commit/?id=eccd2fea8b3928a5989600740e26eccc71a8df49 Cherry-picked to R4_4_maintenance with http://git.eclipse.org/c/platform/eclipse.platform.git/commit/?id=a81a8d738de1c329f29f4cc2a998b4a4a1f8dd25 I've also updated the bundle version with http://git.eclipse.org/c/platform/eclipse.platform.git/commit/?id=fc87bb52d92ff5c904855b527bdcc48e1f9b4308
(In reply to Dani Megert from comment #8) > (In reply to Michael Rennie from comment #5) > > Pushed to master: > > > > http://git.eclipse.org/c/platform/eclipse.platform.git/commit/?id=eccd2fea8b3928a5989600740e26eccc71a8df49 > > Cherry-picked to R4_4_maintenance with > http://git.eclipse.org/c/platform/eclipse.platform.git/commit/ > ?id=a81a8d738de1c329f29f4cc2a998b4a4a1f8dd25 > > I've also updated the bundle version with > http://git.eclipse.org/c/platform/eclipse.platform.git/commit/ > ?id=fc87bb52d92ff5c904855b527bdcc48e1f9b4308 While there's still issues displaying some of the Ant results, from RC3 performance tests, You can see the one "fingerprint" line, and now all 7 of unit the tests "pass". http://download.eclipse.org/eclipse/downloads/drops4/M20150204-1700/performance/performance.php
And I wanted to apologize ... I think I could have committed this fix, long ago, and just some how got it in my head that ant was in it's own repo ... but, this part is in "eclipse.platform", which, I can commit to.
