654831 - Consistent mochitest-a11y failure in test_dochierarchy.html

Status: RESOLVED WONTFIX

(Mozilla Labs :: F1, defect)

Description

[Philipp von Weitershausen [:philikon]]

All my recent try runs for fx-share have shown a mochitest-a11y failure in test_dochierarchy.html on all Linux and Windows builds, e.g.:

http://tbpl.mozilla.org/?tree=Try&rev=71e238ebbe8d
http://tbpl.mozilla.org/?tree=Try&rev=856cb75e7193

The base revisions on m-c were clean, so it's very likely something fx-share introduced.

Comment 1

[Philipp von Weitershausen [:philikon]]

OS X isn't showing any failures because accessibility is turned off there by default: https://mxr.mozilla.org/mozilla-central/source/configure.in#4835

Comment 2

[Philipp von Weitershausen [:philikon]]

I can reproduce this. The reason is that the <browser> element we define within the Share panel now shows up in the root accessible as a document, so the child count no longer is 1 but 2.

A really simple fix would be to change the test to expect 2 child docs, of course (with a comment as to why). I'm CCing MarcoZ who reviewed the original bug that added this test (bug 592913) to provide some insight.

Comment 3

[Philipp von Weitershausen [:philikon]]

Also CCing davidb who reviewed bug 592913, too.

Comment 4

[Marco Zehe (:MarcoZ)]

Why is this browser element defined?

Comment 5

[Philipp von Weitershausen [:philikon]]

(In reply to comment #4)
> Why is this browser element defined?

Most of the UI for the Firefox Share feature (previously known as F1 and available as an add-on) is loaded from the web into this browser element.

Comment 6

[Marco Zehe (:MarcoZ)]

Philipp,

Do you have a link to a try-server build and can give me some pointers to explain how this actually works so I can see what other implications this might have on screen readers before I comment further? Thanks!

Comment 7

[Philipp von Weitershausen [:philikon]]

(In reply to comment #6)
> Do you have a link to a try-server build and can give me some pointers to
> explain how this actually works so I can see what other implications this might
> have on screen readers before I comment further? Thanks!

Sure. You can try out the F1 add-on today (https://addons.mozilla.org/en-us/firefox/addon/f1-by-mozilla-labs/) which doesn't implement the exact same UI but is conceptually very similar.

A try build of the current state of the integrated Share feature can be found at http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-71e238ebbe8d/ (built yesterday, it's only a commit or two out of date)

Comment 8

[Marco Zehe (:MarcoZ)]

Philipp, yes it's fine to change the test to expect two documents instead of one.

While testing, I noticed quite a number of keyboard nav and markup oddity issues causing accessibility problems, the most obvious being that we need keyoard access to bring up that F1 panel without using the mouse.

Can you CC me to the bug that actually implements F1, or should I file a new bug for these issues?

Comment 9

[Philipp von Weitershausen [:philikon]]

(In reply to comment #8)
> Philipp, yes it's fine to change the test to expect two documents instead of
> one.

Thanks. I'll round up a patch.

> While testing, I noticed quite a number of keyboard nav and markup oddity
> issues causing accessibility problems, the most obvious being that we need
> keyoard access to bring up that F1 panel without using the mouse.

You should be able to use the F1 key (hence the name) to bring up the panel (both when using the add-on and with the try build).

> Can you CC me to the bug that actually implements F1, or should I file a new
> bug for these issues?

There are many. If you have any concerns, please file a bug in the same component that this bug is in, Mozilla Services -- Share: Firefox Client. Bug 651668 tracks the minimal set of things we need to address in order to land, bug 642684 tracks the minimal UX for this feature.

Comment 10

[Philipp von Weitershausen [:philikon]]

(In reply to comment #9)
> (In reply to comment #8)
> > Philipp, yes it's fine to change the test to expect two documents instead of
> > one.
> 
> Thanks. I'll round up a patch.

I thought this was going to take just 5 minutes, but then I realized that we need to make the test smart regarding whether MOZ_SERVICES_SHARE is enabled or not (if Share gets pref'ed off, we can't have failing tests obviously).

Comment 11

[Shane Caraveo (:mixedpuppy)]

we've moved to jetpack for the test infrastructure, and bug 654972 will address accessibility, closing this.