Re: Firefox 4 beta 8 -> get_accChild() returns E_INVALIDARG

You're right. As I said I filed bug
https://bugzilla.mozilla.org/show_bug.cgi?id=624648 for this issue.

Alex.


On Tue, Jan 11, 2011 at 3:57 PM, Sylvain WERDEFROY
<[email protected]> wrote:
> The point is that the behaviour has changed between beta 6 ans 8.
> Could you check the get_accChild modifications ? Perhaps I am wrong...
>
> Sylvain
>
> Le 11 janv. 2011 à 08:45, Alexander Surkov
> <[email protected]> a écrit :
>
>> Then here's a problem since they are 0-based in Firefox crossplatform layer.
>>
>>
>> On Tue, Jan 11, 2011 at 3:33 PM, Sylvain WERDEFROY
>> <[email protected]> wrote:
>>> Not sure, but I think indices are 1-based. The zero value is CHILDID_SELF.
>>>
>>> Sylvain
>>>
>>> Le 11 janv. 2011 à 08:05, Alexander Surkov
>>> <[email protected]> a écrit :
>>>
>>>> Hi, Sylvain.
>>>>
>>>> Indices are zero-based, though if 5th is BR then there's still a
>>>> problem. I'll take a look.
>>>>
>>>> Thank you.
>>>> Alex.
>>>>
>>>>
>>>> On Tue, Jan 11, 2011 at 2:56 PM, Sylvain <[email protected]> wrote:
>>>>> Hi Alexander,
>>>>>
>>>>> Here is the html page: www.swsso.fr/demo/
>>>>>
>>>>> If you explore this page with a tool such as inspect32.exe (free
>>>>> Microsoft tool), you can see that the content window has 6 childs.
>>>>> The first 5 childs are <BR> objects.
>>>>> The 6th is a form object. Inspect32 succeeds getting this object
>>>>> because it uses accNavigate method.
>>>>>
>>>>> I would like to get this object with a simple loop form 1 to 6
>>>>> invoking the get_accChild() method:
>>>>> - 1 to 5 returns OK (<BR> objects).
>>>>> - 6 returns E_INVALID_ARG.
>>>>>
>>>>> Hope this helps.
>>>>>
>>>>> Thank you.
>>>>>
>>>>> Sylvain
>>>>>
>>>>> On 11 jan, 04:06, Alexander Surkov <[email protected]> wrote:
>>>>>> Hi, Sylvain.
>>>>>>
>>>>>> I could suppose beta 6 and beta 8 expose different accessible trees
>>>>>> making your code fail. As David said we need a test case to see what
>>>>>> happens. I don't really think it's trivial as you assumed.
>>>>>>
>>>>>> Thank you.
>>>>>> Alex.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Mon, Jan 10, 2011 at 10:08 PM, Sylvain <[email protected]> wrote:
>>>>>>> Hi all,
>>>>>>
>>>>>>> This code was OK on beta 6, but fails on beta 8 (not tested on beta
>>>>>>> 7). This code simply tries to get the first child of an object:
>>>>>>
>>>>>>> hr=AccessibleObjectFromWindow(w,(DWORD)OBJID_CLIENT,IID_IAccessible,
>>>>>>> (void**)&pTopAccessible);
>>>>>>> hr=pTopAccessible->accNavigate(0x1009,vtStart,&vtResult); //
>>>>>>> NAVRELATION_EMBEDS = 0x1009
>>>>>>> pIDispatch=(IDispatch*)vtResult.lVal;
>>>>>>> hr=pIDispatch->QueryInterface(IID_IAccessible, (void**)&pAccessible);
>>>>>>> hr=pAccessible->get_accChildCount(&lCount);
>>>>>>> //  ==> value of lCount is 6 in my example
>>>>>>> vtChild.vt=VT_I4;
>>>>>>> vtChild.lVal=1;
>>>>>>> hr=pAccessible->get_accChild(vtChild,&pIDispatch); => hr=S_OK
>>>>>>> ...
>>>>>>> vtChild.lVal=6;
>>>>>>> hr=pAccessible->get_accChild(vtChild,&pIDispatch); => hr=E_INVALIDARG!
>>>>>>
>>>>>>> Do you have an idea? Not sure, but I guess there is a bug in the
>>>>>>> Firefox code: I suppose there is a test like this in the
>>>>>>> get_accChild() method:
>>>>>>> if (vtChild.lVal >= lCount) return E_INVALIDARG; // where lCount is
>>>>>>> the number of childs.
>>>>>>> As the childs are 1 based and not 0 based, the right code should be:
>>>>>>> if (vtChild.lVal > lCount) return E_INVALIDARG; // where lCount is the
>>>>>>> number of childs.
>>>>>>
>>>>>>> Thanks!
>>>>>>
>>>>>>> Sylvain
>>>>>>> _______________________________________________
>>>>>>> dev-accessibility mailing list
>>>>>>> [email protected]
>>>>>>> https://lists.mozilla.org/listinfo/dev-accessibility
>>>>>
>>>>> _______________________________________________
>>>>> dev-accessibility mailing list
>>>>> [email protected]
>>>>> https://lists.mozilla.org/listinfo/dev-accessibility
>>>>>
>>>
>