After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 570551 - (ARIA) UIUC Number Guessing Game Alert Dialog not presented when it appears
(ARIA) UIUC Number Guessing Game Alert Dialog not presented when it appears
Status: RESOLVED FIXED
Product: orca
Classification: Applications
Component: general
unspecified
Other opensolaris
: Normal normal
: 2.24.4
Assigned To: Willie Walker
Orca Maintainers
Depends on:
Blocks: 404403
 
 
Reported: 2009-02-04 21:42 UTC by Willie Walker
Modified: 2009-02-24 21:47 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
First pass (2.19 KB, patch)
2009-02-12 19:42 UTC, Willie Walker
none Details | Review
Rev 2 along with a new regression test (6.61 KB, patch)
2009-02-12 21:33 UTC, Willie Walker
committed Details | Review

Description Willie Walker 2009-02-04 21:42:38 UTC
1) Go to http://test.cita.uiuc.edu/aria/alertdialog/alertdialog1.php
2) Enter a number and press return
3) An "Alert Box" dialog appears.  Orca doesn't automatically present the contents.

NOTE: You can arrow to the contents and Orca will present them.
Comment 1 Willie Walker 2009-02-11 21:50:33 UTC
This one seems nicer to work with than the Yahoo/YUI alert - we at least get an alert box where the "Close" button is part of the hierarchy.  In addition, this particular example gives focus: to the alert item, making it even easier for us to figure something out.
Comment 2 Willie Walker 2009-02-12 19:42:00 UTC
Created attachment 128583 [details] [review]
First pass

Here's a first pass at a fix.  It does these things:

1) Creates a new _getSpeechForAlert method for Gecko's speech generator that expands the EOC's for an alert if it can (nice idea, Joanie!).

2) Treats the ROLE_ALERT role specially in Gecko's script.py:onFocus.  What it does is put the caret context where we'd like it, but makes sure the script logic for handling focus sees that a ROLE_ALERT got focus.  Without that, the code falls through to the code that stuffs the focus *and* caret to the first caret context inside the alert, making us miss that the ROLE_ALERT got focus.

It seems to work OK, though I'm noticing indeterminate/inconsistent behavior when arrowing around the alert when it first pops up.  At best, I can arrow left once and then arrow right and go through the alert.  At worst, I can't arrow at all.  

I'll play some more...
Comment 3 Joanmarie Diggs (IRC: joanie) 2009-02-12 20:32:38 UTC
I'm still messing with caret nav. I'll look at the arrowing issue. (I have a patch I'm regression testing which might address it.)
Comment 4 Willie Walker 2009-02-12 21:00:43 UTC
I have another patch and regression test coming up.
Comment 5 Willie Walker 2009-02-12 21:33:22 UTC
Created attachment 128594 [details] [review]
Rev 2 along with a new regression test

This new patch does things slightly differently, and should hopefully be somewhat insensitive to the caret nav code that Joanie is making.  There are a couple oddities I'm seeing, though, but I don't think they are related to this bug and/or patch:

1) The expandEOCs call isn't putting spaces where I might expect them.  In particular, we're getting "Alert Box12..." and not "Alert Box 12..."  But, we also do everything we can to get rid of extra spaces.  So, I might try something a little different in the new speech generator method.

2) When dismissing the dialog, I'm noticing that the braille cursor isn't going back to the text box.  Instead, it's not being displayed at all.  If I type enough to make sure the caret moves, though, the braille cursor ends up jumping there.  This might be something that will be sensitive to Joanie's caret work.

Full regression tests running now as a sanity check...
Comment 6 Willie Walker 2009-02-13 00:16:34 UTC
(In reply to comment #5)
> Full regression tests running now as a sanity check...

This doesn't introduce any new issues to the regression tests and it pylints well.
Comment 7 Joanmarie Diggs (IRC: joanie) 2009-02-13 00:39:12 UTC
Then I say go ahead and check it in. I'm still struggling with getting mine right.
Comment 8 Willie Walker 2009-02-13 00:56:15 UTC
(In reply to comment #7)
> Then I say go ahead and check it in. I'm still struggling with getting mine
> right.

Okie dokie!  Committed to trunk and gnome-2-24.  The minor nits can wait.
Comment 9 Willie Walker 2009-02-24 21:47:53 UTC
Closing.  Weĺl open new bugs if people run across them.