Include the surrogate character count when updating text state

Include the surrogate character count when updating text state

The subexpression at the end of walkInlineText() should include the
number of surrogate pairs encountered - and hence match the expression
used to update the character data map. If these don't match, text nodes
following a text node which contains surrogate paris will have
miscomputed value list positions (and fail to apply 'x', 'y' etc
properly.)

BUG=597312

[pdr., 2016-03-30 23:09:33]

pdr@chromium.org changed reviewers:
+ pdr@chromium.org

[pdr., 2016-03-30 23:09:33]

I wrote the exact same patch before so you're definitely on to something!

The non-bmp-tspans.svg failure seems real though. I didn't look into this
further once I hit that.

[fs, 2016-03-31 09:05:20]

On 2016/03/30 at 23:09:33, pdr wrote:
> I wrote the exact same patch before so you're definitely on to something!
> 
> The non-bmp-tspans.svg failure seems real though. I didn't look into this
further once I hit that.

Yes, I noticed this, and have so far come to the conclusion that different parts
of the code account for the same thing using different units. I did not intend
to publish this (and I hope I didn't...), but rather close it - since this will
likely involve some shaving of a yak...

[pdr., 2016-03-31 23:31:05]

On 2016/03/31 at 09:05:20, fs wrote:
> On 2016/03/30 at 23:09:33, pdr wrote:
> > I wrote the exact same patch before so you're definitely on to something!
> > 
> > The non-bmp-tspans.svg failure seems real though. I didn't look into this
further once I hit that.
> 
> Yes, I noticed this, and have so far come to the conclusion that different
parts of the code account for the same thing using different units. I did not
intend to publish this (and I hope I didn't...), but rather close it - since
this will likely involve some shaving of a yak...

Yeah, this wasn't published. I just saw it in your queue looking for another
patch.