Bug 74493 – [GTK] platform/gtk/accessibility/unknown-roles-not-exposed.html is flaky

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Philippe Normand]

Happens at least on the 64-bit Release bot. Diff:

--- /home/slave/webkitgtk/gtk-linux-64-release/build/layout-test-results/platform/gtk/accessibility/unknown-roles-not-exposed-expected.txt 
+++ /home/slave/webkitgtk/gtk-linux-64-release/build/layout-test-results/platform/gtk/accessibility/unknown-roles-not-exposed-actual.txt 
@@ -9,7 +9,7 @@
 TEST COMPLETE
 PASS root.role is 'AXRole: scroll pane'
 PASS webArea.role is 'AXRole: document frame'
-PASS iframe.role is 'AXRole: scroll pane'
-PASS iframe_body.role is 'AXRole: document frame'
-PASS iframe_button.role is 'AXRole: push button'
+FAIL iframe.role should be AXRole: scroll pane. Was AXRole: panel.
+FAIL iframe_body.role should be AXRole: document frame. Was AXRole: scroll pane.
+FAIL iframe_button.role should be AXRole: push button. Was AXRole: document frame.

Will update test_expectations accordingly.

[Mario Sanchez Prada]

(In reply to comment #0)
> Happens at least on the 64-bit Release bot. Diff:
> 
> --- /home/slave/webkitgtk/gtk-linux-64-release/build/layout-test-results/platform/gtk/accessibility/unknown-roles-not-exposed-expected.txt 
> +++ /home/slave/webkitgtk/gtk-linux-64-release/build/layout-test-results/platform/gtk/accessibility/unknown-roles-not-exposed-actual.txt 
> @@ -9,7 +9,7 @@
>  TEST COMPLETE
>  PASS root.role is 'AXRole: scroll pane'
>  PASS webArea.role is 'AXRole: document frame'
> -PASS iframe.role is 'AXRole: scroll pane'
> -PASS iframe_body.role is 'AXRole: document frame'
> -PASS iframe_button.role is 'AXRole: push button'
> +FAIL iframe.role should be AXRole: scroll pane. Was AXRole: panel.
> +FAIL iframe_body.role should be AXRole: document frame. Was AXRole: scroll pane.
> +FAIL iframe_button.role should be AXRole: push button. Was AXRole: document frame.
> 
> Will update test_expectations accordingly.

This is strange... I can't reproduce it in my laptop (64 bit release) nor in the 64bit debug bot. Perhaps it got fixed along the way?

If that's the case, perhaps unskipping and see how it behaves could be interesting

[Martin Robinson]

Sure, no harm in trying to unskip.

[Mario Sanchez Prada]

Committed r104590: <http://trac.webkit.org/changeset/104590>

[Mario Sanchez Prada]

(In reply to comment #3)
> Committed r104590: <http://trac.webkit.org/changeset/104590>

Unskipped again. Let's keep and eye on this to reopen this bug if we see the failure happening again in the bots

[Mario Sanchez Prada]

Reopening, as it has been randomly failing again in the bots. Added again to test_expectations.txt:
http://trac.webkit.org/changeset/104819

[Joanmarie Diggs]

Created attachment 252398 [details]
Patch

[Joanmarie Diggs]

The attached patch should make this test more reliable and cross-platform friendly. Chris, are you aware of any platform where exposing an object with an unknown role is actually the right thing to do?

[WebKit Commit Bot]

Comment on attachment 252398 [details]
Patch

Clearing flags on attachment: 252398

Committed r183842: <http://trac.webkit.org/changeset/183842>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.