672912 - Some accessibility tests scroll window

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Neil Deakin]

This can cause later tests that send events to fail as the element the event is firing on is scrolled out of view.

events/test_aria_menu.html is one such test, others within the events directory seem to also cause this, although I haven't narrowed down which ones.

This issue is affecting some fixes to arrow panels I want to make.

Comment 1

[alexander :surkov (:asurkov)]

Do you mean test window (iframe) or main window?

Some accessibility methods do scrolling, maybe this is what we should live with (for example, make tests sensible to scroll position scroll the window too).

Some tests do scrolling (like events/test_aria_menu.html), it's done here to make sure the element can be clicked, otherwise iirc test fails.

Comment 2

[Neil Deakin]

With the patch in bug 524545 applied, I get a timeout in the test states/test_comboboxes.xul

The main window is scrolling. This has been an issue with other tests (such as bug 667034). One solution, if possible, is to cancel scroll and mousescroll events.

Comment 3

[alexander :surkov (:asurkov)]

This test uses DOMNode.scrollIntoView(true) which seems to cause a problem. Anything to be fixed in a11y tests?

Comment 4

[Marco Zehe (:MarcoZ)]

For some of the tests, we even need potential scrolling to happen, to test for correct same page anchor jumps, for example. We need a scrolling_start event to be fired so screen readers know that the browser just performed a jump to a same page anchor from a link.

Comment 5

[Neil Deakin]

Ah, ok, scrollIntoView scrolls all parent frames as well.

I wonder if other tests have had this issue.

The simplest fix may be to change the tests that use scrollIntoView to scroll back again.

Comment 6

[David Bolter [:davidb] (NeedInfo me for attention)]

(In reply to comment #5)
> The simplest fix may be to change the tests that use scrollIntoView to
> scroll back again.

What's the simplest way to do that?

Comment 7

[Neil Deakin]

You should be able to use window.top.scrollTo(0, 0);

Comment 8

[alexander :surkov (:asurkov)]

Attachment: patch

Enn, can you check if it helps please?

Comment 9

[Trevor Saunders (:tbsaunde)]

Comment on attachment 549121 [details] [diff] [review]
patch

YEAH, THAT SHOULD FIX THIS INVOKER CAUSING A SCROLL  THAT DOESN'T GET UNDONE

Comment 10

[alexander :surkov (:asurkov)]

landed http://hg.mozilla.org/mozilla-central/rev/3a78019c34e5

Comment 11

[Neil Deakin]

Yes, the tests are ok now.