Testfailure fix for MacOS (DirtyStateEditorSupportIntegrationTest) #3275
Conversation
|
@LorenzoBettini I hope this helps for fixing the test problem. Do you have time to check if its ok now? |
|
@mehmet-karaman looks like it fixes the problem, and for sure fixes the arguments passed to assert methods. |
mehmet-karaman
deleted the
Fix_flaky_test_macos_for_refresh_editor_on_sync
branch
|
Need to reproduce this myself and fix this instead of trial and error approach.. |
|
Yesterday I've spend the whole day with experimenting on mac. It runs successful in a junit plugin test, but fails on maven build. I've added several debug outputs tried to understand the state of the object and when the change happens, but it seems that there is no way without trying to understand with debugging in tycho surefire. |
|
If it fails in the Maven build constantly, maybe an explicit refresh? |
|
explicit Refresh of the IFile didn't helped.. It still fails.. There must be an Eclipse Job, which is scheduled but not executed.. |
|
Does it systematically fail? |
|
Mostly they fail, but sometimes they are succesful..
The Debug output: This is the constellation where it is succesfull.. I will compare it with the failing test.. |
|
Just a few thoughts:
|
|
Maybe unrelated, but please when posting snippets use the markdown syntax instead of posting screenshots of code ;) |
By "wait" I mean "join". |
|
It causes a deadlock, because some of these jobs are waiting for the ui thread.. So I have to do a kind of polling. I will remove that break check that again. |
|
I've executed the maven build several times and it seems the polling helped, the test is always successful and not flaky anymore. The asserts have to wait until all scheduled jobs are done (except the periodic jobs). I am going to clean up the code, execute that build several time and if its still ok.. i will create a new PR :) Thank your for your patience @LorenzoBettini |
|
The solution still fails sometimes, depending on the state of my mac.. I will try to run the junit plugin test with the same arguments as the maven build is using.. i must be able to debug this.. |
|
@memin20 please don't waste too much time on that. Given all your experiments it might also be a lost cause: there must be severe synchronization issues in the platform itself and we don't seem to be able to do anything about that.. |
|
Should we disable this test for macos? |
|
I'd agree. @szarnekow wdyt? |
This patch adds a delay (waitForReconciler-method call before checking the document content) and it should fix the synchronization problem which causes flaky tests under mac os.