289006 - Support required and invalid states (leaving open for ATK support)

Status: RESOLVED WORKSFORME

(Firefox :: Disability Access, defect)

Description

[Aaron Leventhal]

Both DHTML accessibility and XForms have a concept of required and invalid
states. ATK has an invalid state, but MSAA does not. Neither have a required state.

**> Bill, can you add a required state to ATK?

In MSAA we are currently planning to 'shoehorn' the states into 2 states that
are rarely/never used today:
STATE_ALERT_LOW = required
STATE_ALERT_HIGH= invalid
This has nothing to do with MSAA's ROLE_ALERT or EVENT_ALERT.
At some point we may create an IAccessible2 interface which supports extended
states, rather than using these alert states.

Comment 1

[Bill Haneman]

Hi Aaron:  REQUIRED state sounds like a good idea.  I'll raise an ATK RFE.

Comment 2

[Aaron Leventhal]

(In reply to comment #1)
> Hi Aaron:  REQUIRED state sounds like a good idea.  I'll raise an ATK RFE.

Bill, no rush, but if you can get me a constant for required soon then I can
support it in the first checkin this week.  Otherwise we'll do it in a separate
round, but that will require extra reviews, etc.

Comment 3

[Bill Haneman]

ATK_STATE_REQUIRED is now available in atk version 1.10.0, cvs HEAD. (I haven't
released 1.10.0 tarball yet).

Comment 4

[Aaron Leventhal]

(In reply to comment #3)
> ATK_STATE_REQUIRED is now available in atk version 1.10.0, cvs HEAD. (I haven't
> released 1.10.0 tarball yet).

Thanks Bill, is it a #define?
Do you know if I should be able to write:
#ifdef ATK_STATE_REQUIRED
    if (aState & nsIAccessible::STATE_REQUIRED)
        atk_state_set_add_state (state_set, ATK_STATE_REQUIRED);
#endif

Or is there a version-based #define I should use?

Comment 5

[Bill Haneman]

It's not a #define, its in an enum.  So you'll have to check for a specific ATK
version at build time and use some compilation guards, I think.

However, since it's just a C enum value, if you build a version based on ATK
1.10.0, it will evaluate to something >= ATK_STATE_LAST_DEFINED, so it should be
safe at runtime; these enums were designed to be extensible that way.

Comment 6

[Aaron Leventhal]

> You'll have to check for a specific ATK version at build time and use some 
> compilation guards, I think.
Okay, do you know the exact #ifdef I should drop in?

Comment 7

[Aaron Leventhal]

Hmm, neither George Kraft nor I could find anything in /usr/include/atk-1.0/atk
to indicate which version of ATK is being used. Doesn't seem like we can have a
precompiler directive unless you add something. Can you add something we can use
or let us know what other options you can think of?

Comment 8

[Aaron Leventhal]

Code freeze is tonight, so I just checked in support for DHTML states required
and invalid in bug 288989. I didn't have an answer on the preprocessor directive
and I just went with this for now:

#ifdef MAI_HAS_ATK_STATE_REQUIRED
    if (aState & nsIAccessible::STATE_REQUIRED)
        atk_state_set_add_state (state_set, ATK_STATE_REQUIRED);
#endif

Obviously that's not right. Leaving it open and assigning to Louie Zhao to get
the right preprocessor directive in.

Comment 9

[Bill Haneman]

the MAI_HAS_ATK_STATE_REQUIRED #define should be keyed to autoconf checks, using
pkgconfig.  If atk version (from pkgconfig) >= 1.9.2, them
MAI_HAS_ATK_STATE_REQUIRED should be set to something non-false.
I'll release atk 1.9.2 today, meanwhile if you build atk from cvs, the version
will be 1.9.2.

Comment 10

[Bill Haneman]

correction: I see that I revved to 1.10.0 when I added this constant, so the
version test should be '>= 1.10.0'

Comment 11

[Trevor Saunders (:tbsaunde)]

We expose these states through xpcom, and map them to states on windows and atk, so close as worksforme

Comment 12

[alexander :surkov (:asurkov)]

right, worksforme