1107002 - testAboutHomeVisibility | Waiting for Toolbar to exit editing mode. -

Status: RESOLVED FIXED

(Firefox for Android Graveyard :: Testing, defect)

Description

[Geoff Brown [:gbrown]]

When I enable UITest exception handling and fix bug 1105795 and fix bug 1105887, testAboutHomeVisibility passes on Android 2.3, but still fails on Android 4.0 and Android x86 with:

http://ftp.mozilla.org/pub/mozilla.org/mobile/try-builds/[e-mail]-1a66c19534e5/try-android/try_panda_android_test-robocop-6-bm100-tests1-panda-build125.txt.gz

17:34:28  WARNING -  TEST-UNEXPECTED-FAIL | testAboutHomeVisibility | Waiting for Toolbar to exit editing mode. -
17:34:28     INFO -  0 ERROR Exception caught during test! - junit.framework.AssertionFailedError: TEST-UNEXPECTED-FAIL | testAboutHomeVisibility | Waiting for Toolbar to exit editing mode. -
17:34:28     INFO -  	at junit.framework.Assert.fail(Assert.java:47)
17:34:28     INFO -  	at org.mozilla.gecko.FennecMochitestAssert._logMochitestResult(FennecMochitestAssert.java:128)
17:34:28     INFO -  	at org.mozilla.gecko.FennecMochitestAssert.ok(FennecMochitestAssert.java:150)
17:34:28     INFO -  	at org.mozilla.gecko.tests.helpers.AssertionHelper.fAssertTrue(AssertionHelper.java:98)
17:34:28     INFO -  	at org.mozilla.gecko.tests.helpers.WaitHelper.waitFor(WaitHelper.java:59)
17:34:28     INFO -  	at org.mozilla.gecko.tests.components.ToolbarComponent.waitForNotEditing(ToolbarComponent.java:242)
17:34:28     INFO -  	at org.mozilla.gecko.tests.components.ToolbarComponent.dismissEditingMode(ToolbarComponent.java:180)
17:34:28     INFO -  	at org.mozilla.gecko.tests.testAboutHomeVisibility.testAboutHomeVisibility(testAboutHomeVisibility.java:38)
17:34:28     INFO -  	at java.lang.reflect.Method.invokeNative(Native Method)
17:34:28     INFO -  	at java.lang.reflect.Method.invoke(Method.java:511)
17:34:28     INFO -  	at android.test.InstrumentationTestCase.runMethod(InstrumentationTestCase.java:214)
17:34:28     INFO -  	at android.test.InstrumentationTestCase.runTest(InstrumentationTestCase.java:199)
17:34:28     INFO -  	at android.test.ActivityInstrumentationTestCase2.runTest(ActivityInstrumentationTestCase2.java:192)
17:34:28     INFO -  	at org.mozilla.gecko.tests.UITest.runTest(UITest.java:102)
17:34:28     INFO -  	at junit.framework.TestCase.runBare(TestCase.java:127)
17:34:28     INFO -  	at junit.framework.TestResult$1.protect(TestResult.java:106)
17:34:28     INFO -  	at junit.framework.TestResult.runProtected(TestResult.java:124)
17:34:28     INFO -  	at junit.framework.TestResult.run(TestResult.java:109)
17:34:28     INFO -  	at junit.framework.TestCase.run(TestCase.java:118)
17:34:28     INFO -  	at android.test.AndroidTestRunner.runTest(AndroidTestRunner.java:169)
17:34:28     INFO -  	at android.test.AndroidTestRunner.runTest(AndroidTestRunner.java:154)
17:34:28     INFO -  	at android.test.InstrumentationTestRunner.onStart(InstrumentationTestRunner.java:545)
17:34:28     INFO -  	at android.app.Instrumentation$InstrumentationThread.run(Instrumentation.java:1551)

Comment 1

[Geoff Brown [:gbrown]]

Attachment: screenshot from failure

This screenshot may be significant -- it appears there is no "edit cancel" view to click on. Is that a tablet vs phone difference?

Comment 2

[Geoff Brown [:gbrown]]

:margaret -- This test clicks on R.id.edit_cancel and times out waiting for R.id.url_edit_text to be not visible. It works fine in our Android 2.3 test environment but fails consistently on Android 4.2 x86 and Android 4.0. The screenshot (see attachment) suggests that edit_cancel is not present; do you know if that is a phone vs tablet difference? Or maybe a Gingerbread vs Honeycomb difference?

Comment 3

[:Margaret Leibovic]

(In reply to Geoff Brown [:gbrown] from comment #2)
> :margaret -- This test clicks on R.id.edit_cancel and times out waiting for
> R.id.url_edit_text to be not visible. It works fine in our Android 2.3 test
> environment but fails consistently on Android 4.2 x86 and Android 4.0. The
> screenshot (see attachment) suggests that edit_cancel is not present; do you
> know if that is a phone vs tablet difference? Or maybe a Gingerbread vs
> Honeycomb difference?

Yes, it looks like the new tablet UI doesn't have a cancel button. So it looks like this would be a problem for any UITest that uses ToolbarComponent.dismissEditingMode.

mcomella, do you have ideas about a fix we can land for the dismissEditingMode implementation on tablets?

Comment 4

[Michael Comella (:mcomella) [NI reported issues only: ex-Mozilla]]

(In reply to :Margaret Leibovic from comment #3)
> mcomella, do you have ideas about a fix we can land for the
> dismissEditingMode implementation on tablets?

If the EditText is focused, can hit back twice, else hit back once - I think that should work.

Comment 5

[Geoff Brown [:gbrown]]

Attachment: on tablet, go back

As suggested (thanks!).

Comment 6

[Geoff Brown [:gbrown]]

Attachment: avoid one more assertion

Unfortunately, with all my robocop patches applied, I still get another intermittent exception thrown on android 2.3 because view.getTranslationY was added in api 11. This removes getTranslationY and uses view.isShown() in place of the more specific visibility check.

That finally wraps up all of the issues except bug 1085837.

https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=649521fe765b

Comment 7

[Michael Comella (:mcomella) [NI reported issues only: ex-Mozilla]]

Comment on attachment 8536874 [details] [diff] [review]
avoid one more assertion

Review of attachment 8536874 [details] [diff] [review]:
-----------------------------------------------------------------

(In reply to Geoff Brown [:gbrown] from comment #6)
> Unfortunately, with all my robocop patches applied, I still get another
> intermittent exception thrown on android 2.3 because view.getTranslationY
> was added in api 11. This removes getTranslationY and uses view.isShown() in
> place of the more specific visibility check.

I'd prefer to have the more specific test - perhaps you can use org.mozilla.gecko.animation.ViewHelper to get the translation?

If not, can we use the more specific check on API 11+ and just isShown() on 2.3? Add a comment saying the check on 2.3 is not as thorough.

Comment 8

[Geoff Brown [:gbrown]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=9dd49c68402d

With the second suggestion (isShown on 2.3 + comment):

https://hg.mozilla.org/integration/mozilla-inbound/rev/79f04caf9dc9
https://hg.mozilla.org/integration/mozilla-inbound/rev/73dbbfed2b30

Comment 9

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/79f04caf9dc9
https://hg.mozilla.org/mozilla-central/rev/73dbbfed2b30