Add <b> and <br> tags checking in infering labels.

In |InferLabelFromPrevious|, <b> tagged element is not considered as inferred label. So <b> tag is added for checking. Also implemented to look for text node prior to <b
    
BUG=76296
TEST=1. Run grabber-bestbuy.com.html. 2. Make sure "6. ZIP 7. UNKNOWN"


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=88479

[honten.org, 2011-06-06 04:39:28]

Please review.

[dhollowa, 2011-06-06 21:19:34]

Looking at the gabber-bestbuy.com.html the long label is a symptom of bad
inferencing, not the length of it per se.

The html looks like this:

 <b>*E-Mail Address</b><br><input type="text" ... name="TxtEmail">

So why is "*E-Mail Address" not the derived label?  That seems like the real
bug.

[honten.org, 2011-06-07 03:45:54]

David,

Could you review?

On 2011/06/06 21:19:34, dhollowa wrote:
> Looking at the gabber-bestbuy.com.html the long label is a symptom of bad
> inferencing, not the length of it per se.
> 
> The html looks like this:
> 
>  <b>*E-Mail Address</b><br><input type="text" ... name="TxtEmail">
> 
> So why is "*E-Mail Address" not the derived label?  That seems like the real
> bug.

[dhollowa, 2011-06-07 16:19:04]

Please add unit tests.

On 2011/06/07 03:45:54, honten wrote:
> David,
> 
> Could you review?
> 
> On 2011/06/06 21:19:34, dhollowa wrote:
> > Looking at the gabber-bestbuy.com.html the long label is a symptom of bad
> > inferencing, not the length of it per se.
> > 
> > The html looks like this:
> > 
> >  <b>*E-Mail Address</b><br><input type="text" ... name="TxtEmail">
> > 
> > So why is "*E-Mail Address" not the derived label?  That seems like the real
> > bug.

[takano.naoki_gmail.com, 2011-06-07 16:39:34]

Sure,

On Tue, Jun 7, 2011 at 9:19 AM,  <dhollowa@chromium.org> wrote:
> Please add unit tests.
>
> On 2011/06/07 03:45:54, honten wrote:
>>
>> David,
>
>> Could you review?
>
>> On 2011/06/06 21:19:34, dhollowa wrote:
>> > Looking at the gabber-bestbuy.com.html the long label is a symptom of
>> > bad
>> > inferencing, not the length of it per se.
>> >
>> > The html looks like this:
>> >
>> >  <b>*E-Mail Address</b><br><input type="text" ... name="TxtEmail">
>> >
>> > So why is "*E-Mail Address" not the derived label?  That seems like the
>> > real
>> > bug.
>
>
>
>
>
> http://codereview.chromium.org/6993053/
>

[honten.org, 2011-06-08 07:14:10]

Please review again.

[Ilya Sherman, 2011-06-08 07:21:08]

LGTM

http://codereview.chromium.org/6993053/diff/8001/chrome/renderer/autofill/for...
File chrome/renderer/autofill/form_manager_browsertest.cc (right):

http://codereview.chromium.org/6993053/diff/8001/chrome/renderer/autofill/for...
chrome/renderer/autofill/form_manager_browsertest.cc:1139: "                    
      id=\"firstname\" value=\"John\"/>"
nit: The alignment is one space off on this line.

[Ilya Sherman, 2011-06-08 07:21:33]

(But please also wait for David's review before committing.)

[dhollowa, 2011-06-08 14:51:09]

LGTM.  Thanks!  Please make sure to run against try bots before submitting, or
manually running the FormStructureBrowserTest.* browser tests.  I've added to
the regression suite so you may need to update the .out expectations files.

[honten.org, 2011-06-09 02:32:24]

Thanks, guys.

I fixed output and now checked "Commit" box.

On 2011/06/08 14:51:09, dhollowa wrote:
> LGTM.  Thanks!  Please make sure to run against try bots before submitting, or
> manually running the FormStructureBrowserTest.* browser tests.  I've added to
> the regression suite so you may need to update the .out expectations files.

[commit-bot: I haz the power, 2011-06-09 03:02:24]

Change committed as 88479