392818 - getTextAtOffset() broken for all but the first line in a multi-line entry

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Joanmarie Diggs]

Steps to reproduce:

1. Type some text into a multi-line entry (e.g. in bugzilla's Description or comments) so that it wraps onto subsequent lines.
2. Launch Accerciser and select the entry in the hierarchy.
3. Move the caret to some location within the entry.
4. In the iPython Console type 

	text=acc.queryText()
	text.getTextAtOffset(text.caretOffset, Accessibility.TEXT_BOUNDARY_LINE_START)

Expected results:  Regardless of the location of the caret when you perform step 4, the current line of text would be returned along with the appropriate starting and ending offsets.

Actual results: The current line of text (along with the appropriate offsets) is returned only if you're on the first line of the entry.  If you start from subsequent lines, an empty string is returned along with starting and ending offsets of 0.

Notes:  It's not just limited to TEXT_BOUNDARY_LINE_START.  All of the text boundary types return empty strings and offsets of 0 with the notable exception of TEXT_BOUNDARY_CHAR.  That boundary works as expected on the first line of an entry (i.e. returns a single character and appropriate offsets).  On subsequent lines it will return the text from your present location to the end of the entry, along with the corresponding offsets.

Comment 1

[Aaron Leventhal]

Ginn, can you look?

Comment 2

[alexander :surkov (:asurkov)]

The problem is GetRelativeOffset() retunrs -1 because the given frame is not primary frame (RenderedToContentOffset fails).

Comment 3

[alexander :surkov (:asurkov)]

Attachment: patch

Comment 4

[Aaron Leventhal]

Comment on attachment 277892 [details] [diff] [review]
patch

Ginn, could you review this? I'm currently mostly away until next week.

Comment 5

[alexander :surkov (:asurkov)]

Attachment: patch2

updated to trunk

Comment 6

[Ginn Chen]

+    nsIFrame *endFrame = GetPosAndText(startOffset, endOffset, nsnull, nsnull,
+                                       nsnull, getter_AddRefs(endAcc));

Is it correct? or missed one nsnull ?

168   nsIFrame* GetPosAndText(PRInt32& aStartOffset, PRInt32& aEndOffset,
169                           nsAString *aText = nsnull,
170                           nsIFrame **aEndFrame = nsnull,
171                           nsIntRect *aBoundsRect = nsnull,
172                           nsIAccessible **aStartAcc = nsnull,
173                           nsIAccessible **aEndAcc = nsnull);

Comment 7

[alexander :surkov (:asurkov)]

(In reply to comment #6)
> +    nsIFrame *endFrame = GetPosAndText(startOffset, endOffset, nsnull, nsnull,
> +                                       nsnull, getter_AddRefs(endAcc));
> 
> Is it correct? or missed one nsnull ?

Correct because latest nsnull is not necessary as argument with default value. Did I get you right?

Comment 8

[Ginn Chen]

I was confused by giving "endAcc" to "aStartAcc".
I think we intend to do that.

I'm still wondering if it is the best way to solve it.
-- to get primary frame from nsIAccessible

Comment 9

[alexander :surkov (:asurkov)]

(In reply to comment #8)
> I was confused by giving "endAcc" to "aStartAcc".
> I think we intend to do that.

It's the same logic like endFrame to startFrame (that is return value).

> I'm still wondering if it is the best way to solve it.
> -- to get primary frame from nsIAccessible
> 

Really I don't know. It's probably because I'm not in deep of frames, layout etc.

Comment 10

[alexander :surkov (:asurkov)]

checked in

Comment 11

[Marco Zehe (:MarcoZ)]

I can confirm that this is now indeed fixed! I am happily typing and see the text scrolling and am able to review it just fine. Good work!