Browser accessibility improvements so that screen readers can access more...

Browser accessibility improvements so that screen readers can access more
complicated webpages without problems.

First, WebAccessibility now works around a "multiple inheritance problem"
in WebCore::AccessibilityObject where the same node appears as a child of
multiple parents. For example, a table cell appears as a child of both a
row and a column. This is solved by having each WebAccessibility parent
check whether the child lists itself as an ancestor. If not, it notes the
child's id only in a separate vector, so each child appears fully only once.

Second, BrowserAccessibility now has internal reference counting, which allows
BrowserAccessibilityManager to update any subtree while maximally reusing
as many objects as possible. This fixes many screen reader interaction
problems!

All of this new functionality is tested with new cross-platform tests.

BUG=67192
BUG=67620
TEST=Adds new unit tests and a browser test.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=77316

[Chris Guillory, 2011-03-06 21:45:13]

This is great! Some minor comments.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility.h (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility.h:68: void ClearChildren();
Seems like it would make more sense to call this function DetachChildren.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager.cc (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:70:
focus_->InternalReleaseReference(false);
Why call InternalReleaseReference separately for focus_? Isn't focus_ is either
a descendant of root_ or is equal to root_.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:238: if (focus_)
Why is focus_ reference counted now?

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:288: // Build a
new tree, reusing old nodes if possible. Any nodes that's
Nit: Grammar, Any nodes -> Each node.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:298: if (!focus_
|| !focus_->instance_active())
Perhaps this logic can be moved into the loop above or even inside of
InternalReleaseReference. I think keeping focus_ as a non-refcounted pointer
would be simpler.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:304: void
BrowserAccessibilityManager::RecursiveGetNodesAndClearChildren(
Should this recursive function be a member of BrowserAccessibility?
void BrowserAccessibility::DetachChildren(std::vector<BrowserAccessibility*>&
children);

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:337:
instance->UpdateParent(parent, index_in_parent);
Can UpdateParent call InternalAddReference in it's new child.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager.h (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.h:135: const
WebAccessibility& acc_obj,
Change acc_obj to src.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager_unittest.cc
(right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:28:
virtual void InternalAddReference() OVERRIDE {
Why override this function?

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:32:
virtual void InternalReleaseReference(bool recursive) OVERRIDE {
Why override this function?

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:83:
checkbox.name = UTF8ToUTF16("Checkbox");
L"Checkbox" won't work here?

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:212:
tree2_root.children.push_back(tree2_child1);
Nit: Why not push_back tree1_child1?

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_win.h (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_win.h:67: 
Nit: Extra newline?

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility.cc
File webkit/glue/webaccessibility.cc (right):

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility...
webkit/glue/webaccessibility.cc:390: // Don't add children are invalid thus
preventing a crash.
Unrelated: Should be "Don't add children that are"..

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility.h
File webkit/glue/webaccessibility.h (right):

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility...
webkit/glue/webaccessibility.h:201: // Returns true if |ancestor| is one of the
ancestors of |child|.
Nit: Nit: "one of the" -> "an"

[dmazzoni, 2011-03-07 16:51:44]

Thanks for the review!

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility.h (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility.h:68: void ClearChildren();
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Seems like it would make more sense to call this function DetachChildren.

Good idea, that's much more clear.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager.cc (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:70:
focus_->InternalReleaseReference(false);
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Why call InternalReleaseReference separately for focus_? Isn't focus_ is
either
> a descendant of root_ or is equal to root_.

I did this to avoid the scenario where UpdateNode deletes the node that had
focus. Doing the check once at the end of UpdateNode seemed cleaner than
checking if a node was equal to focus_ every time the reference count was
decremented.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:288: // Build a
new tree, reusing old nodes if possible. Any nodes that's
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Nit: Grammar, Any nodes -> Each node.

Done.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:298: if (!focus_
|| !focus_->instance_active())
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Perhaps this logic can be moved into the loop above or even inside of
> InternalReleaseReference. I think keeping focus_ as a non-refcounted pointer
> would be simpler.

It seems wasteful to check if every element is equal to focus_ inside the loop
when we can just check it once at the end...plus, checking it inside the loop
would require checking whether the node is equal to focus_ AND whether the
node's reference count just reached zero. This just seems simpler to me.

Another possible benefit: if focus changes and we ever want to "unfocus" the old
focused element, this keeps it around long enough for us to do this.

One reason to possibly do this: I'm considering storing certain attributes of a
node only if it's focused, to be less wasteful. For multi-line text editing, for
example, the line breaks only need to be sent if it's focused.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:304: void
BrowserAccessibilityManager::RecursiveGetNodesAndClearChildren(
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Should this recursive function be a member of BrowserAccessibility?
> void BrowserAccessibility::DetachChildren(std::vector<BrowserAccessibility*>&
> children);

Good idea! This is cleaner.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:337:
instance->UpdateParent(parent, index_in_parent);
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Can UpdateParent call InternalAddReference in it's new child.

Hmmm, I'm not sure this necessarily makes sense to always do this.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager.h (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.h:135: const
WebAccessibility& acc_obj,
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Change acc_obj to src.

Done.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager_unittest.cc
(right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:28:
virtual void InternalAddReference() OVERRIDE {
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Why override this function?

Oh, that was for debugging. Deleted.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:32:
virtual void InternalReleaseReference(bool recursive) OVERRIDE {
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Why override this function?

Done.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:83:
checkbox.name = UTF8ToUTF16("Checkbox");
On 2011/03/06 21:45:13, Chris Guillory wrote:
> L"Checkbox" won't work here?

On Linux, wstring and string16 are not the same.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:212:
tree2_root.children.push_back(tree2_child1);
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Nit: Why not push_back tree1_child1?

I prefer clarity over brevity here - to make sure I didn't have any typos I
wanted to make sure that the first section was only tree1 and the second section
was only tree2.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_win.h (right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_win.h:67: 
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Nit: Extra newline?

Done.

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility.cc
File webkit/glue/webaccessibility.cc (right):

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility...
webkit/glue/webaccessibility.cc:390: // Don't add children are invalid thus
preventing a crash.
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Unrelated: Should be "Don't add children that are"..

Done.

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility.h
File webkit/glue/webaccessibility.h (right):

http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility...
webkit/glue/webaccessibility.h:201: // Returns true if |ancestor| is one of the
ancestors of |child|.
On 2011/03/06 21:45:13, Chris Guillory wrote:
> Nit: Nit: "one of the" -> "an"

Done.

[Chris Guillory, 2011-03-07 18:30:08]

On 2011/03/07 16:51:44, Dominic Mazzoni wrote:
> Thanks for the review!
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> File chrome/browser/accessibility/browser_accessibility.h (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility.h:68: void ClearChildren();
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Seems like it would make more sense to call this function DetachChildren.
> 
> Good idea, that's much more clear.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> File chrome/browser/accessibility/browser_accessibility_manager.cc (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager.cc:70:
> focus_->InternalReleaseReference(false);
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Why call InternalReleaseReference separately for focus_? Isn't focus_ is
> either
> > a descendant of root_ or is equal to root_.
> 
> I did this to avoid the scenario where UpdateNode deletes the node that had
> focus. Doing the check once at the end of UpdateNode seemed cleaner than
> checking if a node was equal to focus_ every time the reference count was
> decremented.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager.cc:288: // Build a
> new tree, reusing old nodes if possible. Any nodes that's
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Nit: Grammar, Any nodes -> Each node.
> 
> Done.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager.cc:298: if (!focus_
> || !focus_->instance_active())
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Perhaps this logic can be moved into the loop above or even inside of
> > InternalReleaseReference. I think keeping focus_ as a non-refcounted pointer
> > would be simpler.
> 
> It seems wasteful to check if every element is equal to focus_ inside the loop
> when we can just check it once at the end...plus, checking it inside the loop
> would require checking whether the node is equal to focus_ AND whether the
> node's reference count just reached zero. This just seems simpler to me.
> 
> Another possible benefit: if focus changes and we ever want to "unfocus" the
old
> focused element, this keeps it around long enough for us to do this.
> 
> One reason to possibly do this: I'm considering storing certain attributes of
a
> node only if it's focused, to be less wasteful. For multi-line text editing,
for
> example, the line breaks only need to be sent if it's focused.
I see. I like this. The only thing is that we're using more raw pointers as ref
pointers. There are some good tests but maybe we can think about possibly
switching to refptrs in the future.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager.cc:304: void
> BrowserAccessibilityManager::RecursiveGetNodesAndClearChildren(
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Should this recursive function be a member of BrowserAccessibility?
> > void
BrowserAccessibility::DetachChildren(std::vector<BrowserAccessibility*>&
> > children);
> 
> Good idea! This is cleaner.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager.cc:337:
> instance->UpdateParent(parent, index_in_parent);
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Can UpdateParent call InternalAddReference in it's new child.
> 
> Hmmm, I'm not sure this necessarily makes sense to always do this.
Agree. The convention with UpdateParent and AddChild should be the same with
regards to the child already being referenced.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> File chrome/browser/accessibility/browser_accessibility_manager.h (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager.h:135: const
> WebAccessibility& acc_obj,
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Change acc_obj to src.
> 
> Done.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> File chrome/browser/accessibility/browser_accessibility_manager_unittest.cc
> (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:28:
> virtual void InternalAddReference() OVERRIDE {
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Why override this function?
> 
> Oh, that was for debugging. Deleted.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:32:
> virtual void InternalReleaseReference(bool recursive) OVERRIDE {
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Why override this function?
> 
> Done.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:83:
> checkbox.name = UTF8ToUTF16("Checkbox");
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > L"Checkbox" won't work here?
> 
> On Linux, wstring and string16 are not the same.
Ok. Nice to see these classes working on Linux.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:212:
> tree2_root.children.push_back(tree2_child1);
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Nit: Why not push_back tree1_child1?
> 
> I prefer clarity over brevity here - to make sure I didn't have any typos I
> wanted to make sure that the first section was only tree1 and the second
section
> was only tree2.
Make sense.
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> File chrome/browser/accessibility/browser_accessibility_win.h (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
> chrome/browser/accessibility/browser_accessibility_win.h:67: 
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Nit: Extra newline?
> 
> Done.
> 
>
http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility.cc
> File webkit/glue/webaccessibility.cc (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility...
> webkit/glue/webaccessibility.cc:390: // Don't add children are invalid thus
> preventing a crash.
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Unrelated: Should be "Don't add children that are"..
> 
> Done.
> 
>
http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility.h
> File webkit/glue/webaccessibility.h (right):
> 
>
http://codereview.chromium.org/6625042/diff/7002/webkit/glue/webaccessibility...
> webkit/glue/webaccessibility.h:201: // Returns true if |ancestor| is one of
the
> ancestors of |child|.
> On 2011/03/06 21:45:13, Chris Guillory wrote:
> > Nit: Nit: "one of the" -> "an"
> 
> Done.

[Paweł Hajdan Jr., 2011-03-07 20:13:30]

Drive-by with minor test style comments. No need to wait for me.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager_unittest.cc
(right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:66:
virtual void SetUp() { }
nit: Why put methods with empty bodies here, instead of just:

typedef public testing::Test BrowserAccessibilityManagerTest; ?

Also, you can just get rid of TEST_F and replace it with just TEST if you
prefer.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:100:
BrowserAccessibilityManager* manager =
Could you make manager a smart pointer (scoped_ptr or scoped_refptr)? This would
help with leaks on test failures.

[dmazzoni, 2011-03-07 22:02:42]

New version uploaded, ready for another look.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager_unittest.cc
(right):

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:66:
virtual void SetUp() { }
On 2011/03/07 20:13:30, Paweł Hajdan Jr. wrote:
> nit: Why put methods with empty bodies here, instead of just:
> 
> typedef public testing::Test BrowserAccessibilityManagerTest; ?
> 
> Also, you can just get rid of TEST_F and replace it with just TEST if you
> prefer.

Done.

http://codereview.chromium.org/6625042/diff/7002/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager_unittest.cc:100:
BrowserAccessibilityManager* manager =
On 2011/03/07 20:13:30, Paweł Hajdan Jr. wrote:
> Could you make manager a smart pointer (scoped_ptr or scoped_refptr)? This
would
> help with leaks on test failures.

Normally yes, but when the test specifically checks the state before and after
deleting the object (as it does below), I think it's clearer to make it an
ordinary pointer.

To put it another way: a specific leak checker is always more helpful for
debugging than a generic leak checker. This test explicitly tests for leaks and
could not pass if |manager| wasn't deleted.

[Chris Guillory, 2011-03-07 22:32:09]

LGTM! A few comments.

http://codereview.chromium.org/6625042/diff/1010/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager.cc (right):

http://codereview.chromium.org/6625042/diff/1010/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:289: // Build a
new tree, reusing old nodes if possible. Each nodes that's
Nit: Each node that's...

http://codereview.chromium.org/6625042/diff/1010/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:298: if (!focus_
|| !focus_->instance_active())
Shouldn't focus_ always be valid here?

http://codereview.chromium.org/6625042/diff/1010/chrome/common/render_message...
File chrome/common/render_messages.cc (right):

http://codereview.chromium.org/6625042/diff/1010/chrome/common/render_message...
chrome/common/render_messages.cc:795: LogParam(p.html_attributes, l);
Log p.indirect_child_ids.

[dmazzoni, 2011-03-07 22:38:49]

http://codereview.chromium.org/6625042/diff/1010/chrome/browser/accessibility...
File chrome/browser/accessibility/browser_accessibility_manager.cc (right):

http://codereview.chromium.org/6625042/diff/1010/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:289: // Build a
new tree, reusing old nodes if possible. Each nodes that's
On 2011/03/07 22:32:09, Chris Guillory wrote:
> Nit: Each node that's...

Done.

http://codereview.chromium.org/6625042/diff/1010/chrome/browser/accessibility...
chrome/browser/accessibility/browser_accessibility_manager.cc:298: if (!focus_
|| !focus_->instance_active())
On 2011/03/07 22:32:09, Chris Guillory wrote:
> Shouldn't focus_ always be valid here?

focus_->instance_active() will be false if the node that had focus disappears. I
don't think there's any scenario where focus_ could be NULL at this point, so I
changed it to a DCHECK.

http://codereview.chromium.org/6625042/diff/1010/chrome/common/render_message...
File chrome/common/render_messages.cc (right):

http://codereview.chromium.org/6625042/diff/1010/chrome/common/render_message...
chrome/common/render_messages.cc:795: LogParam(p.html_attributes, l);
On 2011/03/07 22:32:09, Chris Guillory wrote:
> Log p.indirect_child_ids.

Good catch, done.