[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [egit-dev] SWTBot tests
|
Hey,
Could you try doing the following change?:
@Test
public void testSearchDirectoryWithBareRepos() throws Exception {
deleteAllProjects();
+ shutDownRepositories();
clearView();
refreshAndWait();
Regards,
robinst
Laurent Goubet wrote:
> A memory profiler helped :).
>
> within the tests,
> org.eclipse.egit.ui.view.repositories.GitRepositoriesViewRepoHandlingTest.testSearchDirectoryWithBareRepos()
> tries to delete the test directory, directory that contains the test
> repositories in the user home. This recursively deletes all files it
> can find. Deep down, there is a pack file that needs to be deleted.
>
> This pack file is held open by
> org.eclipse.jgit.internal.storage.file.WindowCache . A savage hack
> to close() and removeAll(PackFile) on that particular file before
> the call to "delete" allows the tests to pass.
>
> I do not know these parts of JGit enough to really know what to do
> from there. Any suggestion on how to properly clear the repository?
>
> Laurent
>
> On 21/03/2013 10:56, Laurent Goubet wrote:
>
>
> I can't say "always the same" for sure, but since I have from 70 to
> 90 failures every run, some are failing every time, yes :).
>
> I am currently trying to figure out how I could detect the leak, but
> lack the tools. Currently trying out with a memory profiler if there
> are suspicious input/output streams on that file.
>
> Laurent
>
> On 21/03/2013 10:35, Alex Blewitt wrote:
>
>
> On 21 Mar 2013, at 09:03, Laurent Goubet wrote:
>
>
>
> I usually end up with a lot of errors, most of them because JGit/EGit
> does not manage to remove some of the files from its test
> repositories. Disabling my anti-virus does not help. My latest run
> ended up in 79 errors and 1 failure, and left the following files on
> my fs :
> Unlike Linux, Windows doesn't let you delete a file if something else
> has it open for read. It's possible that these tests are exposing a
> file handle leak which would not manifest itself on Linux or MacOSX.
>
> Is it the same tests that are failing each time?
>
> Alex
>
>
>
> _______________________________________________
> egit-dev mailing list egit-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/egit-dev
>
>
> _______________________________________________
> egit-dev mailing list
> egit-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/egit-dev
>