Bug 36977 – aria-label doesn't work on image map area

• Component: Accessibility
• Status: RESOLVED FIXED

[chris fleizach]

@aria-label value should be recognized as the label, but it's only determined from @alt

[chris fleizach]

Created attachment 52322 [details]
patch

[chris fleizach]

http://trac.webkit.org/changeset/56930

[Eric Seidel (no email)]

This test looks to be very flaky judging by bug 37005.

[Adam Barth]

Committed r56965: <http://trac.webkit.org/changeset/56965>

[Adam Barth]

Rolled out because this caused aria-liveregions-notifications to become very flaky (or at least that's our best guess).

[chris fleizach]

i have a hard time believing this test is causing flakiness.

[Adam Barth]

You're welcome to look at the buildbot logs.  It's the only accessibility related change near when the test started to become flaky.  We might well be wrong, in which case we can roll it back in.

[chris fleizach]

re-added patch now that leopard release is not running aria-live-region-notification test, which failed last time this was added. that test is the flaky one, and this patch does not affect that code path. so no reason to stop legitimate development

http://trac.webkit.org/changeset/57167

[Eric Seidel (no email)]

This test failed on Snow Leopard again just now:
http://build.webkit.org/results/SnowLeopard%20Intel%20Leaks/r57184%20(5687)/platform/mac/accessibility/aria-liveregions-notifications-diffs.txt

Sorry for commenting on 3 bugs, not sure which one is actually the active one.

[chris fleizach]

i'll see if i can make a less flaky test tonite

(In reply to comment #9)
> This test failed on Snow Leopard again just now:
> http://build.webkit.org/results/SnowLeopard%20Intel%20Leaks/r57184%20(5687)/platform/mac/accessibility/aria-liveregions-notifications-diffs.txt
> 
> Sorry for commenting on 3 bugs, not sure which one is actually the active one.

[Adam Barth]

(In reply to comment #8)
> re-added patch now that leopard release is not running
> aria-live-region-notification test, which failed last time this was added. that
> test is the flaky one, and this patch does not affect that code path. so no
> reason to stop legitimate development

There is some interaction between your patch and that test because without your patch the test passes consistently but with your patch, the test fails intermittently.  Perhaps we should try to understand the interaction before disabling tests?