SkTextToPathIter to respect linearText flag.

SkTextToPathIter to respect linearText flag.

[bungeman, 2013-03-13 16:16:20]

All other uses of SkPaint::kCanonicalTextSizeForPaths in SkPaint are guarded by
a check of isLinearText. It seems odd that this one does not.

The difficulty in general seems to be that advances, outlines, and
rasterizations of glyphs depend on the device matrix as well as the text size on
the paint. I'm not sure that this change is the 'right' thing to do, but at
least it makes getPosTextPath consistent with the other methods using
kCanonicalTextSizeForPaths.

[reed1, 2013-03-13 16:33:33]

lgtm

[bungeman, 2013-03-19 21:23:09]

On 2013/03/13 16:33:33, reed1 wrote:
> lgtm

The difficulty with landing this is that Chromium's HarfBuzz implementation is
using this to find attachment points. The current attachment points are wrong,
and with this change they may be somewhat less wrong, but they are still
incorrect. We may need a better fix.

[bungeman, 2013-03-25 17:23:05]

On 2013/03/19 21:23:09, bungeman wrote:
> On 2013/03/13 16:33:33, reed1 wrote:
> > lgtm
> 
> The difficulty with landing this is that Chromium's HarfBuzz implementation is
> using this to find attachment points. The current attachment points are wrong,
> and with this change they may be somewhat less wrong, but they are still
> incorrect. We may need a better fix.

An additional problem (on the WebKit side) is that the TrueType points do not
1:1 correspond to the SkPath points, as SkPath adds the interpolated on-curve
points of quad segments.