Community
Participate
Working Groups
team.cvs tests listed below fail occasionally. ProjectSetImporterTests#testImportMultipleProjects ResourceMapperTests#testCacheBase These tests fail on all machines, but usually on only one during a build. The reason of failures are server timeouts. Failures just started couple week ago and were not related to any code change in the team/cvs plug-ins.
It looks like it's failing creating this file testCacheBase-1284717450668 We have three test machines connecting to the same cvs test server. Does the test suite create folders with a randomly generated filename? Could there be file contention from one of the other test machines?
(In reply to comment #1) > Does the test suite create folders with a randomly generated filename? The project name is suffixed with System.currentTimeMillis()[1]. Except for the project name, folders and files have their names hardcoded. > Could there be file contention from one of the other test machines? Given the above I think it's the case. [1] org.eclipse.team.tests.ccvs.core.EclipseTest.getUniqueTestProject(String)
I looked at this for today and completely reinitialized the CVS test repo. It seems like the configuration was hosed for pserver. I reran the tests from last night and they all passed.
Unfortunately, it happened again. ResourceMapperTests#testCacheBase failed again in the latest I-build: http://download.eclipse.org/eclipse/downloads/drops/I20101130-0900/testresults/html/org.eclipse.team.tests.cvs.core_win32.win32.x86_6.0.html You already know about failing ProjectSetImporterTests#testImportMultipleProjects, see bug 246547.
I was rerunning the tests today and they passed. I asked you this before but the name of the filename that's created in the test cvs repo is created according to the timestamp, correct? I don't know if multiple test suites running at once against the same repo is causing issues. Another note is that this machine was updated to a faster one a few months ago, not sure if this is causing file contention as well.
(In reply to comment #5) > I asked you this before but the name of the filename that's created in the test cvs repo is created > according to the timestamp, correct? Yup, see comment 2.
ResourceMapperTests#testCacheBase passed in I20101205-2000 and I20101206-1300. Results for I20101206-1800 are still pending.
Also failed again in M20110209-1607.
Failed again in N20110218-2000. Please make the test more robust or delete it as it causes noise in the build results.
I would suggest to make the test invulnerable to this particular failure. I would leave the bug open and log the problem in the build logs when it happens again. It would help to notice whether any Kim's changes in the configuration help. Tomek, please make the change.
Created attachment 189681 [details] Test fix If the exception being thrown in the test is caused by InterruptedIOException the fix prevents the test from failing but log the exception.
Created attachment 189682 [details] mylyn/context/zip
The fix has been applied to HEAD (with minor changes). Available in build >=N20110224-2000. I'm leaving the bug open as suggested in comment 10.
I think this bug can be closed.
