Apache OpenOffice (AOO) Bugzilla – Issue 62723
Open Dialog "read-only" checkbox has incorrect accessibility information.
Last modified: 2006-07-18 12:27:34 UTC
Start up Writer. Bring up the Open dialog. Look at it with at-poke. In particular, look at the "Read-Only" checkbox at the bottom of the dialog. This has a "LABELLED_BY" relationship to the File Type: combo box which is incorrect. If that wasn't present, screen readers such as Orca would be able to correctly use the "Read-Only" text for the checkbox to speak/braille it's true meaning.
I really want to add a keyword of "accessibility" to this bug, but I don't see how to do that from the web page I'm given at http://www.openoffice.org/issues/show_bug.cgi?id=62723
FileOpen dialog is Framework component.
TM->FS: Please have a look, thanks !
fs->pl: The Z-Order in the dialog is - "File type" (fixed text) - filter list (list box) - "Read-only" (check box) Nonetheless, the check box'es GetLabeledBy returns the fixed text. It should return NULL, shouldn't it?
s/Z-Order/Taborder/, yes ? If so, that is not broken; a label is considered the last FixedText, FixedLine or GroupBox in taborder before the control itself. Whether that is a good algorithm may be debatable, but that works as designed. This algorithm basically was decided by MT, so maybe we should discuss this with him.
The only places which use this function are the A11Y implementations. For those cases, the current semantics is nonsense, as it contradicts the sematics of the A11Y LABELED_BY relationship. So, we can either - have the current GetLabeledBy method function in VCL, and translate its results in A11Y-conformant results in all A11Y implementations OR - fix the VCL-version of GetLabeledBy to be A11Y-conformant. To me, the latter sounds like the better solution.
That is beside the point, what is your proposed new algorithm ? Do checkboxes label themselves ? Do checkboxes not have labels ? What other controls should be changed ? Remember I'm not the customer of GetLabeledBy; i can probably do a lot as long as you come up with some requirements.
fs->mt,pl: The text in Window::GetLabeledBy currently reads: // search for a control that labels this window // a label is considered the last fixed text, fixed line or group box // that comes before this control; with the exception of push buttons // which are labeled only if the fixed text, fixed line or group box // is directly before the control I could imagine the following solutions: - what is done for buttons here should also be done for check boxes. This would fix this particular bug here, but does not scale to similar problems. - A fixed text, fixed line, or group box is *only* considered a label of a control if it immediately precedes the control. The idea is: if the order is "fixed text", "control A", "control B", then either the fixed text labels control A (but then most probably *not* control B, at the same time), or control A is a label control itself, then it labels control B. - All places which use GetLabeledBy for A11Y relationships are modified so that they check that "GetLabeledBy( control )->GetLabelFor() == control" holds true. - We explicitly implement A11Y's LABELED_BY / LABEL_FOR relationships for all controls in all our dialogs. Okay, just kidding. What do you think?
fs->pl: as talked about, please discuss a better algorithm with MT. After my vacation, I'm willing to join the discussion, if necessary :)
As per mt's decree check boxes and radio buttons do not have accessibility labels anymore. pl->es: please have a preliminary look at this issue; i didn't find any problems, however i'm not an accessibility guru like you. I built linux and windows install sets for CWS vcl57, so you can check whether this causes any accessibility related problems.
fixed in CWS vcl57
Just for clarification: My "decree" was to check how standard gnome dialogs behave here. So I assume that PL figured out that they don't have lables (because they are 'self labled').
please verify in CWS vcl57
Verified in CWS vcl57
Ok in src680m168
*** Issue 63984 has been marked as a duplicate of this issue. ***
*** Issue 63985 has been marked as a duplicate of this issue. ***
*** Issue 63981 has been marked as a duplicate of this issue. ***
*** Issue 63980 has been marked as a duplicate of this issue. ***