571613 - Pass accessibility tests (mochitest-a11y) without assertion failures

Status: RESOLVED WORKSFORME

(Core :: Disability Access APIs, defect)

Comment 1

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

If there is urgency for bug 404077, we should probably immediately quiet these via preprocessor. Thoughts?

Comment 2

[Jesse Ruderman]

There isn't urgency for bug 404077.  I don't think it's happening immediately, and it's likely to work the same way as reftests, where each test is annotated with the number of assertions it currently triggers.

Comment 3

[Jesse Ruderman]

Bug 404077 has landed, with initial annotations.  mochitest-a11y wasn't the worst:

http://mxr.mozilla.org/mozilla-central/search?string=expectAssertions&find=accessible

Comment 4

[Jesse Ruderman]

And those all seem to be for a small number of assertions, which I can tell my fuzzer to ignore.

###!!! ASSERTION: Can't create an accessible for the document!: 'parentDocAcc', file accessible/src/base/DocManager.cpp, line 389

###!!! ASSERTION: Flush during accessible tree update!: '!accService->IsProcessingRefreshDriverNotification()', file ../../../layout/base/nsPresShell.cpp, line 3817

###!!! ASSERTION: Incorrect results for GetTextHelper: '(finalStartOffset < aOffset && finalEndOffset >= aOffset) || aType != eGetBefore', file accessible/src/generic/HyperTextAccessible.cpp, line 970

###!!! ASSERTION: Incorrect results for GetTextHelper: '(finalStartOffset <= aOffset && finalEndOffset > aOffset) || aType == eGetBefore', file accessible/src/generic/HyperTextAccessible.cpp, line 971

###!!! ASSERTION: Event other than SHOW and HIDE fired for plain text leaves: 'type == nsIAccessibleEvent::EVENT_SHOW || type == nsIAccessibleEvent::EVENT_HIDE', file accessible/src/atk/AccessibleWrap.cpp, line 945

Comment 5

[Jesse Ruderman]

It would be nice to have bugs on all of those.  It looks like bug 850973 will take care of a few.

Comment 6

[Trevor Saunders (:tbsaunde)]

> ###!!! ASSERTION: Can't create an accessible for the document!:
> 'parentDocAcc', file accessible/src/base/DocManager.cpp, line 389

bug 847636 should probably fix all of those.

> ###!!! ASSERTION: Flush during accessible tree update!:
> '!accService->IsProcessingRefreshDriverNotification()', file
> ../../../layout/base/nsPresShell.cpp, line 3817

that might well be bug 809871.  I haven't checked, but if those are all from bug 809871 you might well be able to not ignore that one for fuzzing now so long as you don't try and fuz xul.

> ###!!! ASSERTION: Incorrect results for GetTextHelper: '(finalStartOffset <
> aOffset && finalEndOffset >= aOffset) || aType != eGetBefore', file
> accessible/src/generic/HyperTextAccessible.cpp, line 970
> 
> ###!!! ASSERTION: Incorrect results for GetTextHelper: '(finalStartOffset <=
> aOffset && finalEndOffset > aOffset) || aType == eGetBefore', file
> accessible/src/generic/HyperTextAccessible.cpp, line 971

surkov's working on cleaning up text crap now, so I'm tempted to see where these assertions stand after that atleast forthe ones not fixed by bug 850973.

> ###!!! ASSERTION: Event other than SHOW and HIDE fired for plain text
> leaves: 'type == nsIAccessibleEvent::EVENT_SHOW || type ==
> nsIAccessibleEvent::EVENT_HIDE', file accessible/src/atk/AccessibleWrap.cpp,
> line 945

I seem to remember looking into that at some point, but don't really remember the details.  In any case that's not a terribly worrying assertion.

(In reply to Jesse Ruderman from comment #5)
> It would be nice to have bugs on all of those.  It looks like bug 850973

I'm sort of afraid such bugs won't be terribly useful for the same reason its not terribly useful to file warning bugs that you don't intend to fix.

Comment 7

[Jesse Ruderman]

On Mac at least, it's down to this pair:

###!!! ASSERTION: Incorrect results for GetTextHelper: '(finalStartOffset <= offset && finalEndOffset > offset) || aType == eGetBefore', file accessible/src/generic/HyperTextAccessible.cpp, line 989

###!!! ASSERTION: Incorrect results for GetTextHelper: '(finalStartOffset < offset && finalEndOffset >= offset) || aType != eGetBefore', file accessible/src/generic/HyperTextAccessible.cpp, line 988

And two assertions outside of accessible/:

###!!! ASSERTION: GetElementById had some kind of spasm.: 'Error', file content/xul/content/src/nsXULPopupListener.cpp, line 383

###!!! ASSERTION: Uh, IsInModalState() called w/o a reachable top window?: 'Error', file dom/base/nsGlobalWindow.cpp, line 7345

Comment 8

[Trevor Saunders (:tbsaunde)]

(In reply to Jesse Ruderman from comment #7)
> On Mac at least, it's down to this pair:

I saw the same on linux locally a couple days ago.

a couple weeks ago there was still a assert in states/test_frames.html but it was also outside accessible/ though I don't remember what it was about off hand.

Comment 9

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

We done here?

Comment 10

[alexander :surkov (:asurkov)]

technically no I believe, we just put them into "law", however I'm not sure if we need this bug, might be handy as meta one though.

Comment 11

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

If we intend to file bugs (or have them already) to hang off this one then it makes sense to keep it -- otherwise not.

Comment 12

[Ryan VanderMeulen [:RyanVM]]

Is this bug still useful?

Comment 13

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

Nopers