Community
Participate
Working Groups
Test: 80/80 Errors: 0 Failures: 4
The failures seem related to the fact we don't remove the "lesser" binding on different platforms or at different levels, we just allow the "greater" binding to override it. org.eclipse.ui.tests.keys.BindingPersistenceTest.testSinglePlatform() org.eclipse.ui.tests.keys.BindingPersistenceTest.testBindingTransform() org.eclipse.ui.tests.keys.BindingPersistenceTest.testModifierWithPlatform() org.eclipse.ui.tests.keys.KeysPreferenceModelTest.testBasicConflicts() PW
It turns out there are two KeysTestSuite instances: * org.eclipse.ui.tests.keys.KeysTestSuite (plugin: org.eclipse.ui.tests) * org.eclipse.e4.ui.keybinding.tests.KeysTestSuite (plugin: org.eclipse.e4.ui.keybindings.tests) If I close either plug-in, and run the tests in the other, everything is green. Only with both plug-ins open do I get failures. It turns out the plug-in org.eclipse.e4.ui.keybindings.tests started out as a copy of the other - it redefines key bindings and commands already defined in org.eclipse.ui.tests. Removing those duplicate bindings allows both test suites to run while the other plug-in is open. The question is, if I delete the duplicate bindings, then I need to introduce a dependency between these two plug-ins - which way should it be. OR, I can 'rename' the bindings in one of these plug-ins so that they don't overlap with the other, and then review/modify all the test cases that then fail. Paul W: Any preference for what I should do?
(In reply to comment #2) > Paul W: Any preference for what I should do? Talked to Paul offline. We decided to go with removing the binding declarations for the e4 tests, and add a dependency to the legacy test plug-in (org.eclipse.ui.tests). Proposed fix pushed to Gerrit as: https://git.eclipse.org/r/21949
Created attachment 239919 [details] Diff of the 2 different test directories This is a generic diff of the 2 separate directories. If I'm reading it correctly, we could merge some extra tests from the e4 BindingPersistenceTest into the org.eclipse.ui.tests BindingPersistenceTest and then delete the e4 version completely. Sorry I didn't realize just how close there were before. PW
(In reply to comment #4) > Created attachment 239919 [details] > Diff of the 2 different test directories > > This is a generic diff of the 2 separate directories. > > If I'm reading it correctly, we could merge some extra tests from the e4 > BindingPersistenceTest into the org.eclipse.ui.tests BindingPersistenceTest and > then delete the e4 version completely. > > Sorry I didn't realize just how close there were before. > > PW I've abandoned the change mentioned in comment 3. Pushed a new change, that adds the 5 distinct JUnits from the e4 keybinding tests into org.eclipse.ui.tests, and deletes the org.eclipse.e4.ui.keybindings project: https://git.eclipse.org/r/22018
Released as http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/?id=292b5197d67fc653008b5d121f48c18add01616b Thanks Paul I also added http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/?id=948e492a932c1b48c18eb223eeaa569a4f5f0dc9 PW
Since I've added this back, we now get 6 failures on the mac: http://download.eclipse.org/eclipse/downloads/drops4/I20140218-0800/testresults/html/org.eclipse.ui.tests_macosx.cocoa.x86_5.0.html Paul, could you please have a look? PW
(In reply to Paul Webster from comment #7) > Since I've added this back, we now get 6 failures on the mac: > http://download.eclipse.org/eclipse/downloads/drops4/I20140218-0800/ > testresults/html/org.eclipse.ui.tests_macosx.cocoa.x86_5.0.html > > Paul, could you please have a look? > Looking right now. Some of the new tests have invalid assumptions about Mac key bindings. Preparing a patch right now.
(In reply to Paul Elder from comment #8) > (In reply to Paul Webster from comment #7) > > Since I've added this back, we now get 6 failures on the mac: > > http://download.eclipse.org/eclipse/downloads/drops4/I20140218-0800/ > > testresults/html/org.eclipse.ui.tests_macosx.cocoa.x86_5.0.html > > > > Paul, could you please have a look? > > > > Looking right now. Some of the new tests have invalid assumptions about Mac > key bindings. Preparing a patch right now. Here's the fix. Tested on Mac and Windows. Paul W, can you test on Linux? https://git.eclipse.org/r/22260
(In reply to Paul Elder from comment #9) > > Here's the fix. Tested on Mac and Windows. Paul W, can you test on Linux? > > https://git.eclipse.org/r/22260 Released this fix as http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/?id=eee261ed514929a5321b7b3687478ba0dcb3b378 The tests run clean for me on Linux and there's a failure unrelated to keybindings on mac. PW
The build says there are 6 test failures again, although they ran clean locally on the mac. I'll test locally again: http://download.eclipse.org/eclipse/downloads/drops4/I20140303-2000/testresults/html/org.eclipse.ui.tests_macosx.cocoa.x86_5.0.html PW
If you see these 2 failures, the reason is bug 429592 (but I don't know what made them fail in the last couple of test runs, but not in I20140303-2000): - org.eclipse.ui.tests.keys.Bug43538Test.testCtrlSpace(Bug43538Test.java:63) - org.eclipse.ui.tests.keys.Bug43610Test.testShiftAlt(Bug43610Test.java:62)
(In reply to Markus Keller from comment #12) > If you see these 2 failures, the reason is bug 429592 (but I don't know what > made them fail in the last couple of test runs, but not in I20140303-2000): > > - org.eclipse.ui.tests.keys.Bug43538Test.testCtrlSpace(Bug43538Test.java:63) > - org.eclipse.ui.tests.keys.Bug43610Test.testShiftAlt(Bug43610Test.java:62) My local tests on a mac switch between all green, only those 2 failing, and 6 failing like in the last I-build test.
The two KeyDown tests are no-ops when the workbench window (of the running tests) doesn't have focus: In that case, SWT doesn't deliver the Display#post(...) events, and the tests don't detect that (-> green). On the Mac, I see only these two fail (due to bug 429592).
(In reply to Markus Keller from comment #12) > - org.eclipse.ui.tests.keys.Bug43538Test.testCtrlSpace(Bug43538Test.java:63) > - org.eclipse.ui.tests.keys.Bug43610Test.testShiftAlt(Bug43610Test.java:62) Fixed with http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/?id=c4d4b0663ab4533018de913045e0baa6d65360f4
I'm setting the target back to M7. If can't spend time to fix them, then we should disable them on the Mac. In N20140406-2000 there were again 5 failures.
I've disabled some of the tests as they aren't important now, but the transform test still needs to be investigated. PW
fixed. PW
testBindingTransform failed in the latest I build. PW
This still needs to be investigated but it's not a regression (the tests were added M6 but this test has been failing since Kepler). Deferred. I've disabled the test for now with http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/?id=477b44770d995d87751fd7fbbff1f51406b72557 PW
Those 3 tests are failing right now on Linux in official build but do not fail if I run them locally: BindingPersistenceTest.testAboutBindingEmacs BindingPersistenceTest.testAboutBinding BindingPersistenceTest.testAboutBindingIn3x The failures are caused by two assertions: assertNotNull(bindingService.getPerfectMatch(KeySequence.getInstance("F12"))) and assertNotNull(bindingService.getPerfectMatch(KeySequence.getInstance("ALT+R"))) Any idea?
New Gerrit change created: https://git.eclipse.org/r/47567
Gerrit change https://git.eclipse.org/r/47567 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/?id=356175fe9ea00d59ffc307df1c2917f5c22f843f
(In reply to Eclipse Genie from comment #23) > Gerrit change https://git.eclipse.org/r/47567 was merged to [master]. > Commit: > http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/ > ?id=356175fe9ea00d59ffc307df1c2917f5c22f843f Close VerifyDialog instances left open after the UIWizardsAuto and DeprecatedUIWizardsAuto tests. Those opened dialogs affected somehow key binding tests below. BindingPersistenceTest.testAboutBinding - OK now BindingPersistenceTest.testAboutBindingIn3x - OK now BindingPersistenceTest.testAboutBindingEmacs - still failing If one excludes either org.eclipse.ui.tests.dialogs.UIWizardsAuto or org.eclipse.ui.tests.dialogs.DeprecatedUIWizardsAuto or both (included in the UIAutomatedSuite) all tests pass. So there is still something in those wizard tests which makes BindingPersistenceTest unhappy. I've merged the commit withoit review since changes should not affect other tests (no extra failures), the changes are only in the test code and minimal.
(In reply to Andrey Loskutov from comment #24) > I've merged the commit withoit review since changes should not affect other > tests (no extra failures), the changes are only in the test code and minimal. Working on test without review is fine, tests, examples and demos are not part of the endgame plan.
(In reply to Eclipse Genie from comment #23) > Gerrit change https://git.eclipse.org/r/47567 was merged to [master]. > Commit: > http://git.eclipse.org/c/platform/eclipse.platform.ui.git/commit/ > ?id=356175fe9ea00d59ffc307df1c2917f5c22f843f Seems to be fixed now in http://download.eclipse.org/eclipse/downloads/drops4/I20150511-2130/testResults.php
