310921 - AccessibleObjectFromPoint gives wrong object in some cases when text object is incorrectly not exposed

Status: RESOLVED FIXED

(Firefox :: Disability Access, defect)

Description

[Tim McHyde]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.8b5) Gecko/20050928 Firefox/1.4
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.8b5) Gecko/20050928 Firefox/1.4

AccessibleObjectFromPoint impl via IAccessible::accHitTest

Reproducible: Always

Steps to Reproduce:
Load this html into the latest FF 1.5 nightly and use your tester to get the
accessible object and rectangle as you move your mouse over each part of the
text and you'll see for the right half of the 2nd line it gives the rectangle
for the previou bold part..until you move to the next line of the non-bold part
where it finally figures out the right accessible/rectangle.

<div style="width:600">This part works fine right here <b>and this part works
fine in bold but keep going  to the right where this bold ends and </b>
GetAccessibleFromPoint here gives wrong node for previous bold part. but here
after the line wraps it gets the correct accessible node and rect.</div>

Comment 1

[Aaron Leventhal]

Attachment: Tim's testcase

Comment 2

[Aaron Leventhal]

Attachment: Use frame bounds checks instead of smallest area algorithm to find appropriate accessible in cases where they have overlapping bounds.. Also cleans up some code.

Comment 3

[Mike Kaply [:mkaply]]

Comment on attachment 198312 [details] [diff] [review]
Use frame bounds checks instead of smallest area algorithm to find appropriate accessible in cases where they have overlapping bounds.. Also cleans up some code.

r=mkaply

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 198312 [details] [diff] [review]
Use frame bounds checks instead of smallest area algorithm to find appropriate accessible in cases where they have overlapping bounds.. Also cleans up some code.

Given that the nsEnumRuleAccessible changes are part of a different patch,
sr=dbaron.

Comment 5

[Tim McHyde]

I'm not sure why this bug is left open after all this time since I thought it was fixed, but its good it was.  I'm seeing a new case of it on this page with my tester app linked above in both 1.5 RC2 and deer park 1.6a1:

http://askelm.com/tithing/thi005.htm

In the 2nd to last paragraph, if you wave the mouse over the words italicized words "free will contributions" at the end of the 4th line you will get the object for the previous half of the line ending with the words "was supported solely (at first) by".  If you move to the 5th line and beyond in this paragraph you get an Accessible object but it will not QI for ISimpleDOMText, which is an error.

Comment 6

[Aaron Leventhal]

Tim, I see the latest problem. I recommend attaching a simplified testcase.

Comment 7

[Tim McHyde]

Attachment: html page test case

As a simpler test case steps are to try to wave your mouse over the John 3:16 in this html page and like the other case Aaron was able to see this one wont give the right node.

Comment 8

[Aaron Leventhal]

Attachment: Text starting at the end of a line has a 0 width rect for first frame but isn't invisible

Comment 9

[Aaron Leventhal]

It's even worse than the original summary says -- changing it to reflect that.

Comment 10

[Aaron Leventhal]

Checked into 1.8: attachment 219901 [details] [diff] [review] "Text starting at the end of a line has a 0 width rect for first frame but isn't invisible"

Comment 11

[Tim McHyde]

Attachment: First test case that still fails (simplified version)

The last test case is verified as fixed, but the first test case still fails. 

After simplifying it I discovered it was due to some strange HTML markup: a DL tag without a DT tag, just used to indent apparently. With that DL tag there, you cannot get the correct node for the words in italics ("free will offerings", "free will contributions"). If you take it out, the italics nodes work fine.

Comment 12

[Aaron Leventhal]

Please open a new bug specifically about each case. Makes things easier for devs and QA.